diff --git a/.dockerignore b/.dockerignore
new file mode 100644
index 0000000000000000000000000000000000000000..4b034509f487b02bacb35e5294b183dae5b640c8
--- /dev/null
+++ b/.dockerignore
@@ -0,0 +1,16 @@
+.git
+.github
+.cursor
+.venv
+__pycache__
+*.pyc
+*.pyo
+*.pyd
+*.log
+results
+report
+tmp
+output
+Images
+agent-tools
+*.plan.md
diff --git a/.env.example b/.env.example
new file mode 100644
index 0000000000000000000000000000000000000000..6c902fb08c44242a3c70209b8fd79c4e32d6484a
--- /dev/null
+++ b/.env.example
@@ -0,0 +1,53 @@
+# === Docker ===
+DOCKER_IMAGE=dataclaw:0.1.0
+GATEWAY_PORT=3333
+TMP_WORKSPACE=/tmp_workspace
+
+# === Benchmark ===
+DEFAULT_MODEL=
+JUDGE_MODEL=openrouter/anthropic/claude-opus-4.5
+DEFAULT_PARALLEL=1
+OUTPUT_SUBDIR=output
+BENCHMARK_TIMEOUT_MULTIPLIER=1.0
+BENCHMARK_RUNS=1
+
+# === Web search (OpenClaw web_search tool) ===
+# Register at https://brave.com/search/api/ (Search plan), then set the key.
+# OpenClaw docs: https://docs.openclaw.ai/tools/brave-search
+BRAVE_API_KEY=
+# Optional tuning (defaults match OpenClaw docs)
+# BRAVE_WEB_SEARCH_MAX_RESULTS=5
+# BRAVE_WEB_SEARCH_TIMEOUT_SECONDS=30
+
+# method 1: use openrouter
+# === Auth: OpenRouter ===
+OPENROUTER_API_KEY=
+
+# method 2: use custom api
+# === Auth: Custom OpenAI-compatible API (optional) ===
+# OPENCLAW_CUSTOM_BASE_URL=https://your-api-url/v1
+# OPENCLAW_CUSTOM_API_KEY=
+# OPENCLAW_CUSTOM_MODEL_ID=
+
+# === Auth: Judge Custom API (optional, separate endpoint for judge model) ===
+# JUDGE_CUSTOM_BASE_URL=https://your-judge-api-url/v1
+# JUDGE_CUSTOM_API_KEY=
+# JUDGE_CUSTOM_MODEL_ID=
+
+# === Agent model capabilities (cost unit: USD per 1M tokens) ===
+# OPENCLAW_MODEL_CONTEXT_WINDOW=128000
+# OPENCLAW_MODEL_MAX_TOKENS=16384
+# OPENCLAW_MODEL_COST_INPUT=0
+# OPENCLAW_MODEL_COST_OUTPUT=0
+# OPENCLAW_MODEL_COST_CACHE_READ=0
+# OPENCLAW_MODEL_COST_CACHE_WRITE=0
+
+# === Judge model capabilities (cost unit: USD per 1M tokens) ===
+# JUDGE_MODEL_CONTEXT_WINDOW=128000
+# JUDGE_MODEL_MAX_TOKENS=16384
+# JUDGE_MODEL_COST_INPUT=0
+# JUDGE_MODEL_COST_OUTPUT=0
+# JUDGE_MODEL_COST_CACHE_READ=0
+# JUDGE_MODEL_COST_CACHE_WRITE=0
+
+
diff --git a/.gitattributes b/.gitattributes
index bed0738c7eeb449bca98b5d2f33c89a1ee56349a..9f037e8f2f3637b5bd9a90854c156a435568162d 100644
--- a/.gitattributes
+++ b/.gitattributes
@@ -58,3 +58,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+assets/database/**/*.csv filter=lfs diff=lfs merge=lfs -text
diff --git a/.ms_upload_cache b/.ms_upload_cache
new file mode 100644
index 0000000000000000000000000000000000000000..f068b2c932a1b7659eb8819db82e6157ac3e9178
--- /dev/null
+++ b/.ms_upload_cache
@@ -0,0 +1 @@
+{"version": 3, "repo_id": "GTMLlab/eipBenchmark", "files": {"Images/.gitkeep|1777384240.4363961|0": {"hash": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", "size": 0, "status": "c"}, ".dockerignore|1777384240.4363961|134": {"hash": "72c9c672cfe864aab14b1acbd24e0bf1a083a1b9e37df9a8f2ffe27040bafc8b", "size": 134, "status": "c"}, "Dockerfile|1777384240.4363961|661": {"hash": "e461aeb6e99da9c280551943b513fb411435689afb387260f39aa26421216cf2", "size": 661, "status": "c"}, "LICENSE|1777384240.4363961|1098": {"hash": "8098d70bfe4b0b8e4518f3e1fc2347a2ad654bfdf5d21f3e58c5d5b47e2d5a06", "size": 1098, "status": "c"}, "assets/database/bilingual_translation_english_chinese.json|1777384240.7668092|20059": {"hash": "092c4a1439b16c9cbd890a1b4f6bd76c6bcd441b6aa6b5336117774c4149fe87", "size": 20059, "status": "c"}, ".env.example|1777384240.4363961|1618": {"hash": "e1c82f3eed15cf81c649a708f1b10620db1d471fc5a53f3d1896e0645fa09b4c", "size": 1618, "status": "c"}, "README.md|1777435943.818469|18970": {"hash": "f69f9e976e6fab513fb71c595025db8243ffb956db4c55e58c4dcec87eabaa7e", "size": 18970, "status": "c"}, "README.zh-CN.md|1777435943.7831278|17446": {"hash": "27e4640c8a7264b59d2bb56cd4cb6068df0f787ae41651880c9777b878f77755", "size": 17446, "status": "c"}, "assets/database/enterprise/company_profile_as.csv|1777383639.1263497|242816": {"hash": "14488f5e36b9eec554bc1776513135206dc83b3fc834441bc7eff4ca5171a3cc", "size": 242816, "status": "c"}, "assets/database/enterprise/company_profile.csv|1777383639.12135|1783742": {"hash": "ecd68737f4f453746c3271b146ef5f6b6e741536aab4fc8ae2a32be89085e3c2", "size": 1783742, "status": "c"}, "assets/database/enterprise/company_profile_eu.csv|1777383639.1313508|228500": {"hash": "448bd2077692933fb41398dff37f7767c2cac726d852bf54d5ea9ead2b1c4a8e", "size": 228500, "status": "c"}, "assets/database/enterprise/company_profile_na.csv|1777383639.1353445|152112": {"hash": "83323ec3a56cb36abbb86263c95ee39e0ad5b2bff5101e90d95a69d5efe158cf", "size": 152112, "status": "c"}, "assets/database/enterprise/company_core.csv|1777384240.8871727|27416418": {"hash": "be24817f797d7bc84d1d9c9e3b4da9e727f790dafe0a47f00246f6c57add9cb4", "size": 27416418, "status": "c"}, "assets/database/enterprise/company_operation_status.csv|1777383638.5825517|34717208": {"hash": "2d30f8659da51de379500f0314d5d6322d534267e92377ca9eea47336e9fff94", "size": 34717208, "status": "c"}, "assets/database/enterprise/company_operation_yearly_status.csv|1777383639.1030643|65472033": {"hash": "df4aa7baff8cea1d9eadec7b6650d57a827ea0520cb3fe636d708afadcb0fd15", "size": 65472033, "status": "c"}, "assets/database/enterprise/company_operation_status_detail.csv|1777383638.8144367|70022647": {"hash": "016e39c51abb7f940506d2855bbc0419a6ba50dbbe14e62684285d6c7985d569", "size": 70022647, "status": "c"}, "assets/database/enterprise/company_profile_oc.csv|1777383639.1393816|79246": {"hash": "7f174f6be68dfd8ccfabfc50fed443faa3ffc339b23012a6e54a262b72d19981", "size": 79246, "status": "c"}, "assets/database/industry/national_industry_status.csv|1777383639.149343|1472926": {"hash": "723ce91a24f67b5a3d032a297d40417a1df32c2d584c8215047de92f7c204fd6", "size": 1472926, "status": "c"}, "assets/database/industry/national_industry_status_detail.csv|1777383639.1683435|3631383": {"hash": "3f040fb21b4208bc280acc3bb6c81feb286bbd63150a30ca22677ed802d529b7", "size": 3631383, "status": "c"}, "assets/database/industry/national_industry_yearly_status.csv|1777383639.1733482|55980": {"hash": "a6a2f88f64e19168d8adb70b0edd9dbeebd8df44c4db472c5b23af7a8b0d0f85", "size": 55980, "status": "c"}, "assets/database/industry/regional_industry_status.csv|1777383639.2763426|22504920": {"hash": "c6314f5198f67c8ebf90e03f891f042d8244a5cee5538bd4731678820b289d7e", "size": 22504920, "status": "c"}, "assets/database/industry/regional_industry_status_detail.csv|1777383639.2829518|287796": {"hash": "a08132df31975bcfbb56ea2d2a3b79dc3b052b97d0f88b63ebe1542a0d6f822d", "size": 287796, "status": "c"}, "assets/database/industry/regional_industry_yearly_status.csv|1777383639.285951|289546": {"hash": "ab96eb7f4ca90cf54cd75515beb6031ed72ff26d2df8d3264a969b1c2d8aab41", "size": 289546, "status": "c"}, "assets/database/internal_metrics.csv|1777383639.289951|2264": {"hash": "f2fb590a1da5ca23f92aa2da531e304c319fad52a98794acc5210522aa45d9c9", "size": 2264, "status": "c"}, "assets/database/policy/policy_release_status.csv|1777383639.2969472|1109101": {"hash": "a784a9301e6a8feb63d7b5e2fae7da0f4a3fee3d1871703cf1560d36163184ef", "size": 1109101, "status": "c"}, "assets/database/policy/policy_resource.csv|1777384263.9811273|20833614": {"hash": "041c1e89ed2a24164075e92c9b8e7f74490042ef52dcfbd2c0e8f5c17764a029", "size": 20833614, "status": "c"}, "assets/qa_gold/comprehensive_decision/easy001.json|1777383639.3009472|1375": {"hash": "46c41a28e26aaa10dddc4c99d1ab32d65d09c6ce19439da03dcdfd9d94e7361e", "size": 1375, "status": "c"}, "assets/qa_gold/comprehensive_decision/easy002.json|1777383639.3019476|1752": {"hash": "c500cdbcc40e224f6a37066e15c3231803d40d4e2224ea033b16b949330907ab", "size": 1752, "status": "c"}, "assets/qa_gold/comprehensive_decision/easy003.json|1777383639.302948|1601": {"hash": "6a057370348b65f718e757726a72fcd41b341a2c45674d771b1c14dc3805492a", "size": 1601, "status": "c"}, "assets/qa_gold/comprehensive_decision/easy004.json|1777383639.3049474|1777": {"hash": "4d3c1f0901ea6ab4b06c03fbc48a6e5887a8c49ae9938b6393730ebd0589e25f", "size": 1777, "status": "c"}, "assets/qa_gold/comprehensive_decision/easy005.json|1777383639.3059475|1922": {"hash": "b837e38fceadd7422577ee35675932a79e46d9d5fa896918c6f22d0bc949219b", "size": 1922, "status": "c"}, "assets/qa_gold/comprehensive_decision/easy006.json|1777383639.306949|1344": {"hash": "df24f1cd31c5ec24686c4129fabb0de7e0dc0986046da64232495238c04a6bc8", "size": 1344, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard001.json|1777383639.3089519|3587": {"hash": "7c264a522acb6639359908c1173d332f9526c85665c6b4859b525e4357aecdd7", "size": 3587, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard002.json|1777383639.310978|3468": {"hash": "03d095e908e0c9343c4e17ac23ef644d43ea1d5d4cf7fd59b69bd870ead1dbec", "size": 3468, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard003.json|1777383639.310978|4113": {"hash": "f25dcd97275eccda5199447431bf1277c780b6ab7d98207106ab6a2590c2a2e6", "size": 4113, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard004.json|1777383639.310978|3797": {"hash": "88a8f61909713ee6f685a09ebb9a33b4c34c4931c0a2394e306b31e13d8be207", "size": 3797, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard005.json|1777383639.310978|3369": {"hash": "15bec568cebc09748cf8a6fb702c0a5e6957dd6d64ec07a51060f476c98f85b1", "size": 3369, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard006.json|1777383639.310978|2962": {"hash": "c2b0bf148066bbb436c7aa7d8bb54908e87e8cf38bbd25a0c268fe2e3c53082a", "size": 2962, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard007.json|1777383639.310978|3333": {"hash": "7ce9a1f4fc3ecc8d3ad94c67d82fdd1ad115b728e3a518d10d4ccf4310539a9b", "size": 3333, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard008.json|1777383639.310978|3311": {"hash": "56b51e1612f66f18f328dc3c4e2c56b0af193fff70ecda05dec9f9b85d990167", "size": 3311, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard009.json|1777383639.310978|3413": {"hash": "de12425e0a9861b2c03144d1982d96b1b743a8d266d3f6b88c17ead212169e90", "size": 3413, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard010.json|1777383639.310978|3291": {"hash": "c6d9868e83909c23ba087b98bf5ecd32c7a314add54ecb62eb99ca7e0c8e5ff5", "size": 3291, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard011.json|1777383639.310978|3674": {"hash": "e07c14fc84e71633ae64ef0c356f1aeae3ec4d4d0364f194190d7a9af538c39e", "size": 3674, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard012.json|1777383639.310978|3376": {"hash": "75827fc8c6dc759e189b95a7466903f238ac9840e6c1a325ff3a32309da81c49", "size": 3376, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard013.json|1777383639.310978|2860": {"hash": "e8107d254af70f71fe41da391ab736b44b18666e4797da74429ee13e7b8d84b1", "size": 2860, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard014.json|1777383639.310978|3112": {"hash": "86686b1ab87813bd65ab0e6cf29d0c0d1db4db1baf8314ca4829f2f601b234be", "size": 3112, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard015.json|1777383639.3268986|3587": {"hash": "0eca4de4d53e0834ecf0dc125a8faddcc6fdeae75c411083ac264e625910cb97", "size": 3587, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard016.json|1777383639.3268986|3040": {"hash": "d1c28959089a3f80f68f6ef890dc56148e9f9915dcb1a53d768baead2db833fa", "size": 3040, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard017.json|1777383639.3268986|2480": {"hash": "d4603c65756048752e3cb10b5be9c880b0cee1291d216155f5465464963e7129", "size": 2480, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard018.json|1777383639.3268986|2573": {"hash": "be85057f61e8126f29797ad074e2ce78a0df59945a49caaf4270f216fa51537d", "size": 2573, "status": "c"}, "assets/qa_gold/comprehensive_decision/hard019.json|1777383639.3268986|3505": {"hash": "5bdab338cc8b9b04cb77a91c3801f2233f12dfefaec8de2554de82ddeb91bdca", "size": 3505, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium001.json|1777383639.3268986|2856": {"hash": "5b62419e526e2efc91fd9b9a3fd7712abf4338f5c8f29c0192554049a2e01058", "size": 2856, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium002.json|1777383639.3268986|2335": {"hash": "c9aea8523513439b9972d1269a40eb16119413152332995cad7d62680815a199", "size": 2335, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium003.json|1777383639.3268986|2394": {"hash": "39152f99d87d087cbecb0d6f2328f5fb3d26cc0a98f0a3fd19302298e1f400c4", "size": 2394, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium004.json|1777383639.3268986|2914": {"hash": "6573059a42bfa6fe6f784eea53804316fef97177d446ed0808d9403beac1e560", "size": 2914, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium005.json|1777383639.3268986|1878": {"hash": "7d7b63b7c5f0c1ecfeebd59839659311492a6a34dbcc6eef268186d5bc1edf8f", "size": 1878, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium006.json|1777383639.3268986|1542": {"hash": "7fc3ce48677b60ae32dffa25b742398abe9eee0d1d3678a9291420cfc8b708e5", "size": 1542, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium007.json|1777383639.3268986|2886": {"hash": "beb06c5b4129edcfa705723648b676d02726ccb116d1bc5bb72e8a9f0b902670", "size": 2886, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium008.json|1777383639.3429708|1732": {"hash": "ba0d3f6522ab7d55bd0cdac56b66a5bb19de0ade1d1cb197b32999095cdf11de", "size": 1732, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium009.json|1777383639.3449786|2053": {"hash": "abb77154dce1f24bb67f1487ade6edd100339865f95a47c7e981b8a29c8b174a", "size": 2053, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium010.json|1777383639.3469841|1953": {"hash": "571da7adde2b5a1e2d8e3136135fa16039c9bcf3e65e9dfc27ab57d09c7ee43f", "size": 1953, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium011.json|1777383639.3469841|1608": {"hash": "4626d0eb8b8a79f3a3235a906eb1c17c4ffe4e4e399e40f27d019020bd813572", "size": 1608, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium012.json|1777383639.3489897|2318": {"hash": "f2949c2ec139c2c81920179cea13b88586e7f59b129439ecff3490d32b2aa2cd", "size": 2318, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium013.json|1777383639.3509958|2979": {"hash": "5e2b989e97a124bbfcdf964a83910ba890b52d5173be6c0cc186aa79a2aafd56", "size": 2979, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium014.json|1777383639.3509958|3018": {"hash": "04259acf98d67180cb3eafc1c28d5fef8820ac64585c5cda1e201726af72f092", "size": 3018, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium015.json|1777383639.3530014|4229": {"hash": "d61787889870a0bbac3c9719c3d85dbf61e7b621e451edc4cb65d4568b2e0708", "size": 4229, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium016.json|1777383639.3530014|1962": {"hash": "305faa404e6f6c85595e68247abb6a02a80245b9815250231a346355b83839f0", "size": 1962, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium017.json|1777383639.3550072|2544": {"hash": "9a0fcbd290a7e39d853d72ef85ab81024f2eedab83c94341164eec74e9b6398c", "size": 2544, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium018.json|1777383639.3570125|2085": {"hash": "9fdf69e34e0fbc875ef9971a35b8c8001f82cf0720932913e8513eb7575be957", "size": 2085, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium019.json|1777383639.3570125|2101": {"hash": "3e5f1588c98703179b3e5952ff1512e7d720a6cc0959217a789a82a9ac0be3b8", "size": 2101, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium020.json|1777383639.3587596|3308": {"hash": "d355923a047115a3fca8c2e5519d61b3207c8019439bd06f2a17868d810812d3", "size": 3308, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium021.json|1777383639.3607676|2913": {"hash": "d4961ff8d85eb9c8889b9d18e7935b32caa58a218a60b8e9cfafde7f8e7681ab", "size": 2913, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium022.json|1777383639.3627791|1808": {"hash": "2ef0bca15e68e6987d7ab0c4e08c548bd2163f49af91061cec5a0f49b78e46d8", "size": 1808, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium023.json|1777383639.3647857|1780": {"hash": "b1c6b3d3f271b497c80d2f62e964453d9e4b8c3a0129b47aeb62b92708e71f00", "size": 1780, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium024.json|1777383639.3647857|2009": {"hash": "5ae5ac9def939a1cb5a599da2c6469dc4f38d136e60a8a364b2cb722d222dce0", "size": 2009, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium025.json|1777383639.366791|2342": {"hash": "e3cfeba2e2ae25a6f85a20a07b84b3d28fed78dbe35b2c53cf918d252a579970", "size": 2342, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium026.json|1777383639.368796|2109": {"hash": "900e129fe05e3e4ebd80c6adf322d2c6bd6bdb86c3dffb7d641745c7f505c608", "size": 2109, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium027.json|1777383639.368796|2817": {"hash": "cda7de97376b61b35d6e6bc84f161c597e90b5ad92967be499fed9501543a88c", "size": 2817, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium028.json|1777383639.3708017|4095": {"hash": "023230f150f66a9c1058d4e0eee886791abb9f9a4ad55cae32157afb13108cbf", "size": 4095, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium029.json|1777383639.3708017|2741": {"hash": "fb5e21ddf1e131f3b5cd203060529c4b93e3c31c55f1049e63f8bf176dc99c1b", "size": 2741, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium030.json|1777383639.3728068|2767": {"hash": "03b46ddf85df5b29039617b602bd2501a9d284411500ff2129614c3822d22cc9", "size": 2767, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium031.json|1777383639.3745532|1899": {"hash": "ef02ecd5d8b104eb9d9dd70680a3960c95673f315e7570f55f59348ee64fc379", "size": 1899, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium032.json|1777383639.3745532|2523": {"hash": "eb5fae54bf99fe29c9af12c16afe10e55a83248a4f75ae902bbff415b46e4b30", "size": 2523, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium033.json|1777383639.3745532|1931": {"hash": "ac2f61331c6fe64dce0b503fb2e89fea1ed89d3e9aa780ddac7b908776e3df76", "size": 1931, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium034.json|1777383639.3745532|3180": {"hash": "3d7113e53d139e254e824bc4fcd141515ec8f32f46000750daa0853353807913", "size": 3180, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium035.json|1777383639.3745532|2210": {"hash": "bdb976f1c937a9dec2fca22e710b30dd4c1013f192b5771eaa5e359498cb9da8", "size": 2210, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium036.json|1777383639.3745532|2278": {"hash": "8f703b9748ed92d42191c86aa90fd987cb09bc6f8465ec16ce7040113a095667", "size": 2278, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium037.json|1777383639.3745532|2274": {"hash": "c7b0710e53cb3018d8025e02dd5742cf9ddfbb78f68b6da60e5ac43e36c3d24e", "size": 2274, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium038.json|1777383639.3745532|1928": {"hash": "e3e54410f18224c1d4272064c94915704e15e21c7f0ba95a602b2306923507a0", "size": 1928, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium039.json|1777383639.3745532|4294": {"hash": "8421a10cd4832f29a303eb17c6dc5f5173cc475667f91740bca32101363ed149", "size": 4294, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium040.json|1777383639.3745532|2386": {"hash": "dbf57ce4dd531f1bc73c8215fc7d80c5534de1a0fa9754747bcee061d12426e2", "size": 2386, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium041.json|1777383639.3745532|2156": {"hash": "a960460ba0d96b2abb3aaa952a4051654fa8d79c65979ce6de33a145bf6bbe39", "size": 2156, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium042.json|1777383639.3745532|2192": {"hash": "c5f0bb647e16e0825a51d14ca50f5abde7566bd6b9460f82e8cc47e5ad3ebb2e", "size": 2192, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium043.json|1777383639.3906815|1481": {"hash": "4c940f1e858bf17793942feb86c011a2bd7b8b8f8e62fb70a0a194d2ab7dfa1d", "size": 1481, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium044.json|1777383639.3906815|2493": {"hash": "2abd7c6d704603743c2b06d0e88f8d360e7676c11d14a6c294e60d9aecb51559", "size": 2493, "status": "c"}, "assets/qa_gold/comprehensive_decision/medium045.json|1777383639.3906815|2320": {"hash": "e02ba2b6731505fe842a197249923735e15f9b73551cbbdce8d94743c1ccaa91", "size": 2320, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy001.json|1777383639.3906815|1392": {"hash": "1dcfc026ccc5a112746cc97bcb5f4b19ab574075553cd05f731a47ffb237a2c3", "size": 1392, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy002.json|1777383639.3906815|1447": {"hash": "8946693eee30317039c503ea3cee3f7ca6e6d442ad6523585d9cb46bb16c1bc3", "size": 1447, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy003.json|1777383639.3906815|1588": {"hash": "3439a592275eb82feac58fe11de9a462b3c1bb318a840d0975102ef340910e9f", "size": 1588, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy004.json|1777383639.3906815|1294": {"hash": "6cd267a453f0a5cf26a170d3c72876bc2fa3997c83956d9d301a78197d7085d0", "size": 1294, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy005.json|1777383639.3906815|1278": {"hash": "7dda31b2379ad52828475bf99bafe70610331ea11801430019a04c79b6ce88cd", "size": 1278, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy006.json|1777383639.3906815|1338": {"hash": "5c0abc8f6a6872640a359cb487ae4dbc24d6294129a37886cb994f5538f926cf", "size": 1338, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy007.json|1777383639.3906815|1269": {"hash": "1a162a31f46c32c6a4123ec43ef9153aa4e907d7039409272f8ba7869fc7dc2e", "size": 1269, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy008.json|1777383639.3906815|1362": {"hash": "44cb3084dd8ec86e5fb7aa982a26c160f24f4772fdaa28d0d3f8f55ec627e7d3", "size": 1362, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy009.json|1777383639.3906815|1184": {"hash": "65e3f51659ccbdf24d4452ce75097ed4131403556906014edb2513acd84dc785", "size": 1184, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy010.json|1777383639.4065244|1299": {"hash": "2948aab18fa5e5392b623c51d465a6f1bc701b8ac469f0481832f902a135c31e", "size": 1299, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy011.json|1777383639.4065244|1203": {"hash": "5b1f8f84a3ede197c33d16aea1de1be05b755f5ed9f006c6540f6fe3be932253", "size": 1203, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy012.json|1777383639.4065244|1271": {"hash": "918a502427d1befab1b498cfe7ce7b36b7429e1a4c7e4b41f64366b45a82aecb", "size": 1271, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy013.json|1777383639.4065244|1227": {"hash": "14906c2ef3db91ce45ca3764db35e5f895adf3a0ece57f4e99ebf978d1807070", "size": 1227, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy014.json|1777383639.4065244|1271": {"hash": "d02f7a9066c39a0b3c62136daaaf03c1d561618d1f3c26a479c3e26f6b2253a6", "size": 1271, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy015.json|1777383639.4065244|1287": {"hash": "2073c0fb85484e1354ef61de1128c7044c0ec81f6e14ed87d663f551b503e585", "size": 1287, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy016.json|1777383639.4065244|1392": {"hash": "04f9c0780dcca9c8a773c319df34afd2a660d14b19846051e6ec1a2bc9ae2d13", "size": 1392, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy017.json|1777383639.4065244|1433": {"hash": "b8c88ca1ae2118ef4e10849b03a0dca2026fecc8d4cd3bc4cf8f4afa73a6bad7", "size": 1433, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy018.json|1777383639.4065244|1359": {"hash": "bc4c012571697fb1c2039bddbd01ed138aa52ad03c0d74d781fab9489b953857", "size": 1359, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy019.json|1777383639.4065244|1253": {"hash": "64e0827337b921ea196731ed9fc52bf549c132eafff75198da76b5880521a95e", "size": 1253, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy020.json|1777383639.422225|1258": {"hash": "f5dbaffcab3a8c95fe006c03b8fbf641ffbe14b171618da4bf627cbe98ecc54f", "size": 1258, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy021.json|1777383639.422225|1241": {"hash": "1735e085a196cde2557b3709adb6f2023709a8ab302eaacfe8944499a40fba68", "size": 1241, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy022.json|1777383639.422225|1319": {"hash": "d3ed4f9956527d35fa77a75112b4d759b76253a1d741e94b1219d8c64d63fe95", "size": 1319, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy023.json|1777383639.422225|1424": {"hash": "663bcc2cfdb160492aed2fb3828dd9691d8086ee0583d78e6681de2868428764", "size": 1424, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy024.json|1777383639.422225|1310": {"hash": "29c61b2122c1619da560fc1ab69b0ee252bde168c48960011996a56f95b74da4", "size": 1310, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy025.json|1777383639.422225|1332": {"hash": "395110a6a7d99117499281c18fa037b932292f005454bf245a528c2265eb746f", "size": 1332, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy026.json|1777383639.422225|1252": {"hash": "30146346b34ac5124646693a803565d10cfdb177ab60d6b4a705704f50b91a08", "size": 1252, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy027.json|1777383639.422225|1373": {"hash": "dfa92e8e1b0352d2f45f61a9dd2bc73086800acaf2bafa4223a476d41d47658f", "size": 1373, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy028.json|1777383639.422225|1317": {"hash": "5b93e2c7e7690f87d2c14de41089d89559191edc3cdbc94af69925aad8d236d5", "size": 1317, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy029.json|1777383639.422225|1315": {"hash": "1d788b6e1c69c755a59a4b02124666ff20aa5e0c80ccc58dede2f5f6ae6a96d0", "size": 1315, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy030.json|1777383639.422225|1222": {"hash": "0ae7b9caefb85b28736c15f90a3c5adc50bccd2eef8185ae3422070f60068a46", "size": 1222, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy031.json|1777383639.422225|1319": {"hash": "28adc63554fce471654ba5ce99d82027f2c6c10c28cd9e97c7b9277a82189b3c", "size": 1319, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy032.json|1777383639.4379306|1252": {"hash": "56056323e6930ecbee3943ab499b4a685c2d4b9be1072d10cc667019c288105e", "size": 1252, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy033.json|1777383639.4379306|1266": {"hash": "511bc22c014a538e0fabeb66ce9090e63d05d187bde9e6668444f2dfaabe3db2", "size": 1266, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy034.json|1777383639.4379306|1366": {"hash": "ad72369f51790ae1b0682758769955838fb915774e90c0520965f74afed6e587", "size": 1366, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy035.json|1777383639.4379306|1075": {"hash": "7a0c2b553a5cfc165dc89b2d245f7e918f03a337c9f190b851b4ad693d8212d2", "size": 1075, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy036.json|1777383639.4379306|1161": {"hash": "583b006105481dae8865d71045b6d67c675a3e16b89901f89763b36ce37ecc1d", "size": 1161, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy037.json|1777383639.4379306|1087": {"hash": "52ee72f3414be28602765264ff2cde13f5792975e0d62ce46027fd7c6fd801ff", "size": 1087, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy038.json|1777383639.4379306|1129": {"hash": "9b40e79b94abbef968bc2b04382b896117eb29f54ae1a1c7f8629a7b8026d4f6", "size": 1129, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy039.json|1777383639.4379306|1182": {"hash": "f0f8c95149c9b03f0dffc8974bdeaf6cdc670e65f92f543f601af5e22ec4bc1d", "size": 1182, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy040.json|1777383639.4379306|1139": {"hash": "c548c8730e09e9aead994f9efedf847635a36a9aaf33a8d7fc5d0b5af3fac9fd", "size": 1139, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy041.json|1777383639.4379306|1333": {"hash": "c3afd1a1633edd91166114199f502cf3b40d206243f129e49366e00992ee0b4f", "size": 1333, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy042.json|1777383639.4379306|1091": {"hash": "90795b82855ad1f721a251aec1aecdaf984d780763ace4761ed73c3fd0866416", "size": 1091, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy043.json|1777383639.4379306|1333": {"hash": "15e3705d87a5514d52e0970388e759cc1a9a4c99845653a992a8e8297981dd0c", "size": 1333, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy044.json|1777383639.4537518|1179": {"hash": "8d7d70545ecf2b7a5abbfdb5b03cb85527783e687dd49bdcf2e273a52ca30cc0", "size": 1179, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy045.json|1777383639.4537518|1158": {"hash": "d444dbecc919eff6eece91c09f179375420107fcb997be56dd2994d61018b734", "size": 1158, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy046.json|1777383639.4537518|1120": {"hash": "e1406933f20ee9481eb80649704f8a9fc164b4f438a64ddaf00b9ed7d5b210dd", "size": 1120, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy047.json|1777383639.4537518|1156": {"hash": "c89f0d6b17d0cf25573ce382714a3a9d760c81e2c48232ab428a31928fc84cae", "size": 1156, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy048.json|1777383639.4537518|1148": {"hash": "c6ccbedd28e7ab2c0208575511b5fa7d90c7c7d4eb0a44402f35ac82f1f9db47", "size": 1148, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy049.json|1777383639.4537518|1041": {"hash": "48db602f3e9b75a6363e2ca1c5b7665d8642ddbe5b814bc98e5df4e04b7fa39d", "size": 1041, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy050.json|1777383639.4537518|1080": {"hash": "c315fe75b9ade62bfd8639a0eced1714b2cb16005141e7d7a53b38d891435272", "size": 1080, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy051.json|1777383639.4537518|1116": {"hash": "18be5fcf71f031ec68a757df1ca5fb62844004f5ac7460d48bcbfe2472f185a7", "size": 1116, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy052.json|1777383639.4537518|1114": {"hash": "08fad28e682443bdbc66a98795c4476943fc7982ee3e70398ade95d7548dbf07", "size": 1114, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy053.json|1777383639.4537518|1168": {"hash": "1265eb652586fecb4f12cc39da933e1b112f5d9b38c1840401fe8e85658d8f40", "size": 1168, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy054.json|1777383639.469393|1082": {"hash": "8533394a90f8c73f6fc920f76ecb3da1099e5299319a8581be7942754c9bf735", "size": 1082, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy055.json|1777383639.469393|1685": {"hash": "5b85a5c2fca328358305ce2633c8826d2f5704a7fd2728cc4c912dd408104666", "size": 1685, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy056.json|1777383639.469393|1736": {"hash": "aa9bf1c9ecdaa88e3fa010586b24eacb9c6985b6d402ee8e2f516fbf6b1dfa1e", "size": 1736, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy057.json|1777383639.469393|1645": {"hash": "30b7097463ae4beb1d407c6ce60a3c0e57118800667cb34a9ce6ac2e27f8dfa1", "size": 1645, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy058.json|1777383639.469393|1585": {"hash": "4b2c269adf95b3307b1a1e8158a02352547e12a8761e8394fa2e7da712cbac1b", "size": 1585, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy059.json|1777383639.469393|1553": {"hash": "1801e56f400d30d74ebe64998926f1e45c55ecd721e64eec9311deaa32b4970e", "size": 1553, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy060.json|1777383639.469393|1756": {"hash": "54a64dc87ca38cbebc9d7380e4ad24521f9562af2ce310cb7317d470c7ad897a", "size": 1756, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy061.json|1777383639.469393|1774": {"hash": "c340b1e18212b4b1aeb3745b704fcd211de8cae68aebad9ef45f2bd3463bd510", "size": 1774, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy062.json|1777383639.469393|1370": {"hash": "e49054991247e4d2cddf5220982b51f9bb47d72c779f3de370e09e9c034cf958", "size": 1370, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy063.json|1777383639.469393|1323": {"hash": "33d11583962644bd7027930780c1b5d5b5050faa241bc6868449d27d807f8df8", "size": 1323, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy064.json|1777383639.469393|1365": {"hash": "09979820df2fc45e8b28777a447609add81393c7dfc383f5cce0cc86b140f5c7", "size": 1365, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy065.json|1777383639.469393|1426": {"hash": "018149af73298f148dac15445b780ebad9486839b86a20af69eb6751776470b5", "size": 1426, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy066.json|1777383639.4853766|1278": {"hash": "c068c3a5ce7379082474fbe563ce96d9dfdc158e51e06eb7825b5813f48e0377", "size": 1278, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy067.json|1777383639.4853766|1408": {"hash": "1c9091fcf51ca38a8fbd029c34b89ad5b72f223435570887e8311a87c25e4fa0", "size": 1408, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy068.json|1777383639.4853766|1473": {"hash": "4081568f77a791fa668276ee1e2119c5cd36eb00ca8e9a585f0530d4d393f3e9", "size": 1473, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy069.json|1777383639.4853766|1146": {"hash": "3904bf593fe0607b9d36940e89a7ec5422d2a24dfd13322bd47b607892b230d1", "size": 1146, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy070.json|1777383639.4853766|1351": {"hash": "e9026b5e37188e7ef90386c95761e25f1649ab0bd7310632ab4660b745690358", "size": 1351, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy071.json|1777383639.4853766|1352": {"hash": "57aff529398f8bb458e5bb5dcccec327b246aa5ec48189d55e1c9ba08555e80d", "size": 1352, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy072.json|1777383639.4853766|1289": {"hash": "a57871599e6f89db5ef0460fbbcece98f661feb8ef67973a90e8c1b976fb19e8", "size": 1289, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy073.json|1777383639.4853766|1381": {"hash": "c6cc656b66bf619d42453fd3014534ab6a6ead15f801fa9ab5ba9a7db8e28216", "size": 1381, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy074.json|1777383639.4853766|1330": {"hash": "01fd0e41599925e57e65af3a4fb788e9baae70ec859c7b9ed3c68587484135e1", "size": 1330, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy075.json|1777383639.4853766|1155": {"hash": "dc57e40dab8b028e2a2282ea97ae4c8993caa20d91b17b8d36190b2c73bfe092", "size": 1155, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy076.json|1777383639.4853766|1230": {"hash": "a13a532be228759e1ad15e0b70d165c969af0424198148d93451c79902844d94", "size": 1230, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy077.json|1777383639.5013058|1206": {"hash": "5165530ea88a1a7ee9b7b6310232ef0b6cc8a8ca3cdcf9f7d969a6022b8251a3", "size": 1206, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy078.json|1777383639.5013058|1521": {"hash": "36422a027d0d0cb6d44a2e406c209bc847389fc8521a5c61e3e55b1e85182d47", "size": 1521, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy079.json|1777383639.5013058|1316": {"hash": "0d43ba08a3bdb45443212eef24e185682719dd0ad3e0435644ae3aab406dcde9", "size": 1316, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy080.json|1777383639.5013058|1393": {"hash": "700673f3e29b5462ca813c38b028883cae0f5be6c32f2397bf66e63ae6782a7d", "size": 1393, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy081.json|1777383639.5013058|1369": {"hash": "6c75b83506c7d02864b0f779e87e25f5ee585d4e2b21c7460a5e46ab60a38b71", "size": 1369, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy082.json|1777383639.5013058|1368": {"hash": "5be0647e2ae49f27d4af0285a354f003fe985f7be98097db6ecf2b0dba941e00", "size": 1368, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy083.json|1777383639.5013058|1283": {"hash": "675b83a852ba40830552a16fc26407e7d638955fe1da95374288d98389a5ef36", "size": 1283, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy084.json|1777383639.5013058|1384": {"hash": "fb705425863c21d1a8b46322a0211ee8eb6098764c21655cedf6dd0872569f41", "size": 1384, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy085.json|1777383639.5013058|1176": {"hash": "25bd9070a816748787f5a4942f3d061dc538efaa5fc81c79f743c13d17a583c6", "size": 1176, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy086.json|1777383639.5013058|1346": {"hash": "d90b6aa0389ee2432efbd1afff82de30955243cc581ab2b5ee8c967d60f0d135", "size": 1346, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy088.json|1777383639.5013058|1362": {"hash": "4a2413ef1e60972b173de09d93b2e8340d51101d9124df02453e98ef90c09bc0", "size": 1362, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy090.json|1777383639.5013058|1403": {"hash": "b713613cafff5aa23f15b0b2b0b32d5b20bf32ef813f087958c99c2d25c6748d", "size": 1403, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy091.json|1777383639.5171814|1072": {"hash": "f99cbea1db445e9061c34e2d55831451d45be0cf81509ba393bcab27223b8fd8", "size": 1072, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy092.json|1777383639.5171814|1212": {"hash": "6b3707af178d12766a4a0acdb9dcadcbd0f7198cf3e81cf7aa411797a5e35465", "size": 1212, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy093.json|1777383639.5171814|1249": {"hash": "20b4023316ae80b214d7ea987deab559ca8bffc19f48947606939e30ad71b23a", "size": 1249, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy094.json|1777383639.5171814|1262": {"hash": "e759a4e516328f2f3507a6440c2a889d699d59500c9f712b38bdfda23c756215", "size": 1262, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy095.json|1777383639.5171814|1050": {"hash": "d274688c4642070e263b80d81e8a213deb5f13f340773d588822a92d283f1912", "size": 1050, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy096.json|1777383639.5171814|1176": {"hash": "dad7827c39bc776b8f392013d22fec79b37237bbc029009985816014a2eb8059", "size": 1176, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy097.json|1777383639.5171814|1170": {"hash": "552759e832b7719a05b8d6dc2caa4d993b0ccfb9ac60039f4a011c68f582820a", "size": 1170, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy098.json|1777383639.5171814|1080": {"hash": "5d11402e0e167472b316e5e4c2612c1330e9904da3b0facd1b4e247727a1a7b6", "size": 1080, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy099.json|1777383639.5171814|1127": {"hash": "9ac27f9651fb19ce5981ddcdad956c64f032fd6e44c83f595d62d9a6f7a468e9", "size": 1127, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy100.json|1777383639.5171814|1179": {"hash": "ce0761daccf30da6f04191510542c096244158f8e4bd0a0f1ccd6f2007b09b03", "size": 1179, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy101.json|1777383639.5171814|1118": {"hash": "2fd5983d6d03e9cc2b027a6c18a7034f3d1f2d2d2773c14206801928e18101e7", "size": 1118, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy102.json|1777383639.5330975|1059": {"hash": "3707b0b3f5e7b400b728998f4652bc7561b46e1da7741d57b454bb2c8bb17c8c", "size": 1059, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy103.json|1777383639.5330975|1097": {"hash": "eeb7463f45fd3dcd8acd50bbe3d22410c1d4c970e84d5fdc56a2bef0f9ee79ad", "size": 1097, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy104.json|1777383639.5330975|966": {"hash": "3670a39e23f4d0de0d9961d5fe53c34545afcaf6e8145ddcdfc5daff508bfeb3", "size": 966, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy105.json|1777383639.5330975|1085": {"hash": "bf5749e8dc852b8f0d9d93f3fb9e34e65324ce1618255cd430a07d56962206ed", "size": 1085, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy106.json|1777383639.5330975|1338": {"hash": "52f3b2f71ca2effd97b3ee6120bc4e3e1f4b587ca946c9348d8eab993a4eccf0", "size": 1338, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy107.json|1777383639.5330975|1069": {"hash": "1e93fece1ac802c4e12268111ac724fc0ff06877bfb44690a85f33b6fe958737", "size": 1069, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy108.json|1777383639.5330975|1040": {"hash": "2262d03f54282e6452a90bd432dff343c14e1b9fe0049cd31810909c843ef7e9", "size": 1040, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy109.json|1777383639.5330975|1049": {"hash": "31e73cf99d6b09cbfca016855fc2a3ea68111127b00361623d55f9c820f230c6", "size": 1049, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy110.json|1777383639.5330975|985": {"hash": "f72d8b079a0381cc91ed664b98d54fc060b9e7b33f866216957af18e904ff87a", "size": 985, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy111.json|1777383639.5330975|1024": {"hash": "cbd68b3152508eae7bca83679ff46d300f0d486da9a46731f1294f8e3fba0a30", "size": 1024, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy112.json|1777383639.5330975|1073": {"hash": "172f9276b087a8c5a1407c80abfbf06f816c745b6959df0100e35cd0126a5aa8", "size": 1073, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy113.json|1777383639.5330975|1108": {"hash": "be33e5987f695c719ffc41e511ae3cd27db76748b9f57b94071e212742f4070b", "size": 1108, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy114.json|1777383639.549103|1148": {"hash": "593731238da32e8aefc1cb92a5ee8fdc24963ca983c0e39586467accb2284fc6", "size": 1148, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy115.json|1777383639.549103|1007": {"hash": "42269757e67622f9b66085b730d2af37a80c7e962907be7e2224055407081bf0", "size": 1007, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy117.json|1777383639.549103|991": {"hash": "b1f476721d7742bb38a0ee4753c4f4559a2ae240637eb8fc6f33926aaffae0ab", "size": 991, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/easy116.json|1777383639.549103|1092": {"hash": "0870fa182079a7563406c85903432c4f5bdaa621e3315fa01e91cb5dc7f945d5", "size": 1092, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium001.json|1777383639.549103|1403": {"hash": "6c9e5fa18be34e4ae3774215dfcf84c00dd43a092950bbb4b1a0a4a7abc8bfc1", "size": 1403, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium002.json|1777383639.549103|1422": {"hash": "42e5587507be174de257333d26a4d5875c9846f92dfbfd23af6cf9ee67493639", "size": 1422, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium003.json|1777383639.549103|1385": {"hash": "e7b11e26673decc804042ee66c60b2483e7adfd03cd8ee9b7f2c28355c4ca163", "size": 1385, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium004.json|1777383639.549103|1451": {"hash": "f85403df57f205130f8488853610e975c1540baf9a4ffa60945a4e08c073cdef", "size": 1451, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium005.json|1777383639.549103|1595": {"hash": "db547a5dbe7e6732b49c6be987f024481f3253609189dd15f23cd77bcc791da6", "size": 1595, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium006.json|1777383639.549103|1522": {"hash": "9b5c014187dcc03b09342ed9429391289bd1700475703b10e7764814512e3411", "size": 1522, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium007.json|1777383639.564959|1734": {"hash": "e67f69a51b4fbf2f66510d5808bc72dae4d94629d2cf9ab36fe6549e8e4a4716", "size": 1734, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium008.json|1777383639.564959|1814": {"hash": "2e614fee2dd7406e76ef6769932f0df64bb9e149ddee044794fb915793d01f03", "size": 1814, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium009.json|1777383639.564959|1616": {"hash": "e5c981fab329c04839e4d0635c7e539b1fd9072dcfe90bc2599bbce07346f75a", "size": 1616, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium010.json|1777383639.564959|1569": {"hash": "ab3ea89afb4359de647ee710e829a821a7f91fe2c545efafea47a6930f32da5c", "size": 1569, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium011.json|1777383639.564959|1643": {"hash": "35e913f8494079ba9a9d1ae91986eebc2d727d2ffefe712ae299e54a2ecdfd85", "size": 1643, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium012.json|1777383639.564959|1562": {"hash": "4ebe5bb90d02119bf3bed607740cbb469c9ec0212161b3afacbe6dc31fc7f11e", "size": 1562, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium013.json|1777383639.564959|1578": {"hash": "adf4d4e82ff03dde353b79037e96c814fa4aae78ff85e2d520bb2dcc61d30baa", "size": 1578, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium014.json|1777383639.564959|1965": {"hash": "dcc77bc44b6429a5cc6c8dbe53e0ddea1426e7c8b8d3efbebb80e2496613712e", "size": 1965, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium015.json|1777383639.564959|1827": {"hash": "7500b22690fe571e19c377e7974fe25278d28d39972d6ad498504bbcfc703975", "size": 1827, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium016.json|1777383639.564959|1825": {"hash": "c1e33b666f22321c093a5fa0d462bcfb1648e245f7aa9e11ec46853037299558", "size": 1825, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium017.json|1777383639.564959|1719": {"hash": "98e0cdab7e711cba57fb06bc75f0f973ad60c3c0c89b19942a51a446523a1a95", "size": 1719, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium018.json|1777383639.564959|1838": {"hash": "31806dad660b9d903a5cf80d63b4b7c0618c55c08061930c859d1e0e598633f1", "size": 1838, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium019.json|1777383639.5807436|1795": {"hash": "60d19b38e65622683d15412191837bf9b8ca3e90ea50f3981e0d6c7d70b6bda3", "size": 1795, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium020.json|1777383639.5807436|1713": {"hash": "a61c0de89e70ed3cb47212ab859045c79c58943dd01f3e803e55a81aa36de204", "size": 1713, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium021.json|1777383639.5829082|1630": {"hash": "e4944e3c8437a61ac02d2a60c656889beeb41378cfdaa6b1a0eb6262fe28c422", "size": 1630, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium022.json|1777383639.5829082|1724": {"hash": "6a1f676e4d0a21617afbef5775e1eb9039d3043905d15eb013d41d312e3104d2", "size": 1724, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium023.json|1777383639.5829082|1988": {"hash": "144134bb94f80ede79ed2a32f7ef3866f0db7619c77cd7e5eb8c572ef65b1fc8", "size": 1988, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium024.json|1777383639.5829082|1895": {"hash": "ae50e82632e5eb7068674bc7b79fbe3cecc848cbd9b994e9209246b25a44f17a", "size": 1895, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium025.json|1777383639.5829082|1842": {"hash": "f4c8ce7561f651bc0504f392b5889043ab603c0af65cb9a4a184d21ec57679d4", "size": 1842, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium026.json|1777383639.5829082|1642": {"hash": "d7b396756cb7b43556aa7424d18e457558c5aaf7832ba12b1bfd8f2733a431b0", "size": 1642, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium027.json|1777383639.5829082|1676": {"hash": "ea60a22c9f43079e80d70eb8423a68eefc5b874e16417c540ea7674c39ced9ea", "size": 1676, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium028.json|1777383639.5829082|1686": {"hash": "fc626790388baee1e03c2415a276d60dd9256c0fdd6cfb639fd456eaa8ce25df", "size": 1686, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium029.json|1777383639.5829082|1637": {"hash": "b6b8c464cad37bcb7b618c925420e635ebf2badc900bb63a81b55641d85dece3", "size": 1637, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium030.json|1777383639.5829082|1626": {"hash": "b0f6b4db66afae38f4cd75a94a9717fb0e9d884d14de017d14c51cf12850df5c", "size": 1626, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium031.json|1777383639.5964653|1737": {"hash": "9eb8eb9ad65c8dae3c54659166fa37441ceef8ee05795d6071a056e7b1a0a0b2", "size": 1737, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium032.json|1777383639.5964653|1727": {"hash": "ec3128b00443ee8e1485c7262cc5d253b0ee784db114508895a86b94b12b8b7b", "size": 1727, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium033.json|1777383639.5964653|1975": {"hash": "9e4c036a16779bb55d71183acf1016ba478bdb86a65e1906b74457deb79e189b", "size": 1975, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium034.json|1777383639.5964653|1827": {"hash": "bc9b3cfb3561baee1f50661b6663f38bfa509b4ce73b6e39c81d683cb9d0c4c1", "size": 1827, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium035.json|1777383639.5964653|1911": {"hash": "5082e8e4913015c31735a9184043af3107c398c276f560b259dc22e7db544d13", "size": 1911, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium036.json|1777383639.5964653|1748": {"hash": "2247a18086ca45d40041bad03ba1f6e1891940760067a655a320f4a39541e21a", "size": 1748, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium037.json|1777383639.5964653|1969": {"hash": "4f0ac5264fa5c8fb6074f703c6951bdda891fec6e6a018c48938467314dda02a", "size": 1969, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium038.json|1777383639.5964653|1745": {"hash": "1281ae00d2e8f9548590013905f1087a4b08f4e201612a3455052ae42087e8d5", "size": 1745, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium039.json|1777383639.5964653|1718": {"hash": "9a2a06cd3cf12933b52ac04bf12a4a7c021bd143555f3dbcd33f5e57030a3d9c", "size": 1718, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium040.json|1777383639.5964653|1761": {"hash": "fb0f03e4d0e2e85bb8502d552bcd86e4195f71af7417c7c33b5919777a544148", "size": 1761, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium041.json|1777383639.5964653|1806": {"hash": "aa8e9463a2369c0d5d4baead2e667ccb034d04acb599f0fdfad8846047077015", "size": 1806, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium042.json|1777383639.6121407|1950": {"hash": "3a63144f4ffd417c7fee119a0d87bebba02e75b3526aeb54e7e3c1046def3f78", "size": 1950, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium043.json|1777383639.6121407|1765": {"hash": "9fcdccf02ae02cbf0ed271f1d8df8293794263c8afca1918354c6bf5ad612bcb", "size": 1765, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium044.json|1777383639.6121407|1853": {"hash": "dea6ebedba6bb0130fa8c708817d298c29c8654e3e26fda5a48e0e93af4eec88", "size": 1853, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium045.json|1777383639.6121407|1737": {"hash": "263d780e73ae18cdc72d4aedede208fd2387e707b3691311b2d07e86c53bbcb6", "size": 1737, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium046.json|1777383639.6121407|1759": {"hash": "2ebc945d32f90249852480bb38acacdc06c980a12ea9df0f5a559ae963c9d902", "size": 1759, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium047.json|1777383639.6121407|1785": {"hash": "717570c5faf972859fd1487e0f967ac5dd776bdfee3f42bacd6560efdf35b2c7", "size": 1785, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium048.json|1777383639.6121407|1849": {"hash": "f557633e2dc71bea8512742272d0e039fae69c5f26bbdef67edb23f2ac888834", "size": 1849, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium049.json|1777383639.6121407|1685": {"hash": "08b20a4719bd3b40ecb032902aad6c9ea5ff385a1dd3ad8761e8038aac4ea27c", "size": 1685, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium050.json|1777383639.6121407|1724": {"hash": "97788e581ca90ccae4f2346fcc779c053f7542f7f3ff9835e496360b88923ff4", "size": 1724, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium051.json|1777383639.6121407|1692": {"hash": "6166ecfb45e796b40515ee3a73ad9d6655f4cc975b63182cdab45d7b1c55e5ec", "size": 1692, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium052.json|1777383639.6121407|1901": {"hash": "eb25896f5a5323244827c410f1ef25cd41dfd7701ffe9ba2c5e790e27feb2f3d", "size": 1901, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium053.json|1777383639.6121407|1844": {"hash": "553683e0617071ed375649ff170c53a77f27ad8807ef91d4e1352c2a35039f33", "size": 1844, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium054.json|1777383639.627825|1908": {"hash": "e619a31dd54671aa0c57659840432df67aa636a2bb499764faf65dbca67c2962", "size": 1908, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium055.json|1777383639.627825|1941": {"hash": "9c58a0329438698997fcd5a3f291f645cddaac7385d5071e47d4ca919a9b9d7b", "size": 1941, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium056.json|1777383639.627825|1771": {"hash": "4fbc843b52ba1a3f372738ff1742a96b4e418d1150fc2006182da00cffcf69fd", "size": 1771, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium057.json|1777383639.627825|1734": {"hash": "1d329825389a761a960096a10f81d2edc33918a1b6c7d61326ecc847b1f5b870", "size": 1734, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium058.json|1777383639.627825|1792": {"hash": "8466c65834a13d2df55f33801df8d75441d7d9953f021e53eed9d490ff8df368", "size": 1792, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium059.json|1777383639.6372535|1786": {"hash": "aa942860e3a966a13fa5be383ec81da5138122a0a75a20e717322d46f3f1e1dc", "size": 1786, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium060.json|1777383639.6392617|1759": {"hash": "007c9ece38c9b02c2248aec188f23b78077fe5afa4e91f7f72e6c2ba52a84819", "size": 1759, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium061.json|1777383639.6412704|1573": {"hash": "e308b7c1390bd3759853f5d0837d35f2ada90712a14b4cb77fdc4defb9fb9ea5", "size": 1573, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium062.json|1777383639.6432781|1419": {"hash": "112af738f861a85f55a84c9cf77f58ae5258d3ece1dbd9beb0e668d268a4ce53", "size": 1419, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium063.json|1777383639.6453216|1911": {"hash": "0b9823ea9e2d41435b1c8f19c1d5b110a39e23ff6f308a65c031a414491372da", "size": 1911, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium064.json|1777383639.6473317|1754": {"hash": "753f2564336aa00d94191442df1d7b4a936b46f23c8226aecd699c145719fb16", "size": 1754, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium065.json|1777383639.6473317|2302": {"hash": "f808aee5a6eb9ef01f1d99b995e31a98a61147c9393bbccba0c25dc98f592661", "size": 2302, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium066.json|1777383639.6473317|2022": {"hash": "5032d9e26d11942fc03dbdc5447129b4d1014f1400c963570642bae3ccd982e7", "size": 2022, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium067.json|1777383639.6473317|2185": {"hash": "21fac6353e360240d534ff5c5d46a01e63bd29e1e2e47b9b163c9effa77dee7f", "size": 2185, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium068.json|1777383639.6473317|1635": {"hash": "02496b51a9df3eb00d53746b461d2460fdb5f3876f43a83a3520492c2113a534", "size": 1635, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium069.json|1777383639.6473317|1938": {"hash": "f1fef73733f8d654165d8b3c03b5be1446f4586c204a3898aaa25dd6d7aa2967", "size": 1938, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium070.json|1777383639.6473317|1791": {"hash": "4887c2df8ca84c273a3f99973724f76ce65ee88d5e49c0c366765b46bea84947", "size": 1791, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium071.json|1777383639.6473317|2119": {"hash": "7d25b89d50f592a1544bd0d8fec0e6bec40def8e1ea0c3e78e40cf9efb904156", "size": 2119, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium072.json|1777383639.6473317|2122": {"hash": "822d9d7817b8dfe5863f4b931e5e9e547a939f54987733d0a32dd2f2897c6106", "size": 2122, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium073.json|1777383639.6591184|1519": {"hash": "82c4f42b706cfc7d3f02133947da4046ba361cf381c5d1450302091e5748ed0f", "size": 1519, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium074.json|1777383639.6591184|1552": {"hash": "5daa0280ffb26f5cd6ec1561722665925db00b5d63b65b307b5250ddcd4b569b", "size": 1552, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium075.json|1777383639.6591184|1516": {"hash": "1d57898e9baae4e52cabcd8feb3df1a630a003de01cdebc5d14a05204c6ba6fa", "size": 1516, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium076.json|1777383639.6591184|1685": {"hash": "33bd330ccafd692b4e16424559891ef565c3f4c501b2f55cc6c2dee6698db174", "size": 1685, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium077.json|1777383639.6591184|1902": {"hash": "f94f2e58257cc837359272e9434fe656588f2d367db8791ee5900ae0fb33e0b8", "size": 1902, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium078.json|1777383639.6591184|1809": {"hash": "c1b37ee1b1d1aa5dc41776e72d42777a221f23865454ba40a93375941d06ae13", "size": 1809, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium079.json|1777383639.6591184|1579": {"hash": "70c9188b719a862f421f91b13228be2e739861fe2f59954e544ea9aaeaf5c841", "size": 1579, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium080.json|1777383639.6591184|1578": {"hash": "f740888114c7a1b6afbee21b171180a7f8035cd6302604f3eb4979845fb479db", "size": 1578, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium081.json|1777383639.6591184|1467": {"hash": "229f57f6125c4e155286d90d11c83ebe205a6f275b980f533785086fccfcf0f8", "size": 1467, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium082.json|1777383639.6591184|1655": {"hash": "215ad373ec446662968f8233976e0dee8fbfc697f09bff8c8344b78849e6b7c6", "size": 1655, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium083.json|1777383639.6591184|1536": {"hash": "462db81368841ec01514125876269d379e358fcc554382f95d59b975e2d6dcd3", "size": 1536, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium084.json|1777383639.6591184|1967": {"hash": "8fa885c16179c53351e751955f4155dbef37b7a69983ccc6a5e05b03d3b29383", "size": 1967, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium085.json|1777383639.6750546|1939": {"hash": "faeae04adc1ce74ebaf4f8954b01b927772b8b40f77e97fb877ac8e3ed85d763", "size": 1939, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium086.json|1777383639.6750546|2129": {"hash": "8b61685c7f7e867bd6ecdb7c23b0752975dea9f85e269c39996d5619e30c8b74", "size": 2129, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium087.json|1777383639.6750546|1610": {"hash": "8d59d4934ec6b2f08240b05aa7b3c104022ebbeb35c0d139d4fd440c60a22653", "size": 1610, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium088.json|1777383639.6750546|1769": {"hash": "357854ca91c78a809ac019e0370c8e9be6293c3446796311bd002d1d4e4110d7", "size": 1769, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium089.json|1777383639.6750546|1561": {"hash": "4c8d011167846b01ca239b494d80881d97659cc33db16635600b1f49129da050", "size": 1561, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium090.json|1777383639.6750546|1763": {"hash": "9fe132024f6a5fd284b4611c51830a1ba45c982c68907535c546470ffd71f2c8", "size": 1763, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium091.json|1777383639.6750546|1505": {"hash": "2e438d8e8a737aac767d94d0b3f41d1b96fc73699cbed108af1b3b138be626a2", "size": 1505, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium092.json|1777383639.6750546|1503": {"hash": "ce12a0a47cde9ac9c36e9fede71c2cc5de9e78c05104cd7b06b5d073f28e0d82", "size": 1503, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium093.json|1777383639.6750546|1474": {"hash": "796cb3ab649d1435cfd8aa3c90fd64977a5304f50524787db3fda6a31faade27", "size": 1474, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium094.json|1777383639.6750546|1511": {"hash": "52931892bac288bf2ad9d497b54d120b6a77227c92f4bfa34a5134dec39cef5f", "size": 1511, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium095.json|1777383639.6750546|1625": {"hash": "b251e92341fe3edd316cf9061655b2cd9a97559488989e9f74aaf52462899a65", "size": 1625, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium096.json|1777383639.6908405|1868": {"hash": "8a55bd7bac10d796272992ea9e184ac8b97fdb38f784206b2a48a7b53fb04735", "size": 1868, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium097.json|1777383639.6908405|1934": {"hash": "3331ffa8e99104b81bb7c7082db58da66991efb254cf7d922bef64056fce3ff4", "size": 1934, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium098.json|1777383639.6908405|1517": {"hash": "c3c7c51e1f8397b18670ac6c9a4ff23f93e05a9b3c351553172e0ad5e17c6a6f", "size": 1517, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium099.json|1777383639.6908405|1316": {"hash": "dc8d5f6dd6a488db71385dd1ec9f8ffb7661c9a01a112ba4c3a23255fd01d1b8", "size": 1316, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium100.json|1777383639.6908405|1448": {"hash": "ecc661640eef240bf2e56c9077db92dab0d7eed6dbc556bbe8a136f9d4e9c97d", "size": 1448, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium101.json|1777383639.6908405|1559": {"hash": "21d1ce2cd87c5a3bfb7a706790a5160528eacdf7efcd3ad9ef8c1ac96d8a8ae1", "size": 1559, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium102.json|1777383639.6908405|2051": {"hash": "da64769b5f9e7546de4fbcd8032fb36c495a538eac5a129b4e40e5d8ae3415c3", "size": 2051, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium103.json|1777383639.6908405|2005": {"hash": "60513889d07ba06dea25ecba163669b0ddbd99a90c02bd0cbf38526e53ef2db8", "size": 2005, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium104.json|1777383639.6908405|1873": {"hash": "639e781520578e094e3a09c92b0540215ab33a539b9925e16261795ed97fc547", "size": 1873, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium105.json|1777383639.6908405|1723": {"hash": "ce1b5ab13f1b25d93005cb05aafa2517cf5ef3a7fbc94982c5b10fbfd61347a6", "size": 1723, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium106.json|1777383639.6908405|1662": {"hash": "1355a3434a4c0be245bc7f350720d84b5737158155bddeabca33532211dfbf49", "size": 1662, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium107.json|1777383639.7064755|1632": {"hash": "a9ca34ef1a64efa36c94630f64cac5cb1872fbcc88c465111df4d71323f48dc1", "size": 1632, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium108.json|1777383639.7064755|2167": {"hash": "d12bdc2f6b7d01c0fffd1c34a45123d11949ca475d841643458255e5816ab804", "size": 2167, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium109.json|1777383639.7064755|2329": {"hash": "68394600fdd29ac9927aad45463ac6504de401044e4994c699c4558578c8177c", "size": 2329, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium110.json|1777383639.713541|1704": {"hash": "002c0f2460dcd08cf7e5e4b6694513040277a13e5bd74606f16eb01074a5acae", "size": 1704, "status": "c"}, "assets/qa_gold/enterprise_industry_analysis/medium111.json|1777383639.7155638|2040": {"hash": "5759265ae14a340636c97ed34cb7c16f76622b796cb8697b43cecddf7909790d", "size": 2040, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/easy067.json|1777383639.7155638|1071": {"hash": "01b76cb7e7c6c4e621f7a3b7e7cd8758e6233622a9c70034819092489ea352ad", "size": 1071, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/easy068.json|1777383639.71758|1005": {"hash": "111e8da09b423069c674332f3105a1f81930c37f03d4f66e7a7ad6a544d8fe6a", "size": 1005, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/easy069.json|1777383639.71758|1268": {"hash": "d8de1f413e6068437511c93aa63a7b1641416423155b93ea0c1026670890be56", "size": 1268, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/easy070.json|1777383639.71758|1081": {"hash": "e26a348755762d2b482fef31096b37af1faf44c6be2e8b25cd04e9a0052cfd91", "size": 1081, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/easy071.json|1777383639.7223783|1133": {"hash": "6e500b148d4a5d42c9d31b7e74adc4d31104bacb109ae393b424e369314c433a", "size": 1133, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/easy072.json|1777383639.7223783|1107": {"hash": "a5824d88bbcd442bad3c47421b018394d619c542ee11e9873495ca079b78d074", "size": 1107, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/easy073.json|1777383639.7223783|1210": {"hash": "a96d56114894c6e04fdd7e223bbc32fb39e9211db8014160fc0042208926006f", "size": 1210, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/easy074.json|1777383639.7223783|1288": {"hash": "bef98230970d8613c00103ecb85506f1431311dce05d2abe07fd7e5a6a0eae6c", "size": 1288, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/easy075.json|1777383639.7223783|1488": {"hash": "227ab299bd38a3c35965297b9531d702cabc46dbb67ae6da3ec2d774c9d452af", "size": 1488, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/easy076.json|1777383639.7223783|1167": {"hash": "e9e5cbd5e2657b99a8444a94d326075fdfff3e6dde10f568ab5e953786ca8886", "size": 1167, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium001.json|1777383639.7223783|2166": {"hash": "5cb166ad98b638cd6e311772ebf37a47f164b761b7d961be8f03663ee502541c", "size": 2166, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium002.json|1777383639.7223783|2157": {"hash": "3484fe7b39e2a79140545f1aef6158b918381a5fb4365e3cbcdf4a584385014f", "size": 2157, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium003.json|1777383639.7223783|2136": {"hash": "ca5b6d4e0803605513cbaa54f87d94dcfab55e567c942568b6b460deeefe0a0d", "size": 2136, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium004.json|1777383639.7380877|2171": {"hash": "3b06aee849662ba94f8acf7731ae105aa9053fa4b9f56137f9d4ede187036baa", "size": 2171, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium005.json|1777383639.7380877|2156": {"hash": "d94888cdace62d2d72bf66eb5dd337636b2d819aa6be54df3ffdc58bb80c31c1", "size": 2156, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium006.json|1777383639.7380877|2075": {"hash": "e975446cec546b22d6251bbc1ea5beb2325ffe36415f9bccb348233d7d86ec4e", "size": 2075, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium007.json|1777383639.7380877|2003": {"hash": "3a6353003b42701f1c27d463f05105f40cb19ff5e134c023092fc1bc2be974d1", "size": 2003, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium008.json|1777383639.7380877|1832": {"hash": "d21dfab642be9ba83533b78f862c24a50ad71ee8fb68d013b0fd6e4c15ddada7", "size": 1832, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium009.json|1777383639.7380877|2535": {"hash": "f4777e10dc03a81591364d2f9218fc4222c83ff4353019309ace2e58ef4e59f9", "size": 2535, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium010.json|1777383639.7380877|2008": {"hash": "c605a412d1d2fbf74445f34799077302f0d4314d6586afe658a7067a77e37911", "size": 2008, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium011.json|1777383639.7380877|2272": {"hash": "f3c6103cff184394b7a57fa6548d857739010703379a1bde749964da3cb4b45f", "size": 2272, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium012.json|1777383639.7380877|2119": {"hash": "a1483df1754766c201b73b8b3ab399eee67f715998e4f04c1b5db877568210ff", "size": 2119, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium013.json|1777383639.7380877|2324": {"hash": "8a7f33be8a40011d84c40aa3a9f008d28dc796fcc9d57cb178bdeec7b2f46aba", "size": 2324, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium014.json|1777383639.754071|2251": {"hash": "01cd3bbab0a944a0e0e3a158bec9b6b6933bbe45f83887ee23de257b8dab0fc0", "size": 2251, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium015.json|1777383639.754071|2021": {"hash": "6fa0536fc8c0b21faad25221863c26495094180717846a6f05fdac8f9509e14d", "size": 2021, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium016.json|1777383639.754071|1905": {"hash": "4d004f183843cf876a01bc40e202e75983334ec17510e0c1b2f33fecf8f6abf1", "size": 1905, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium017.json|1777383639.754071|2179": {"hash": "32eddc159cdca8a5edd3519da43291a3b28abc6babe81fddeb9459ae6be1ae7a", "size": 2179, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium018.json|1777383639.754071|2255": {"hash": "cac3ed194d71fcdf3282e67c08da8d8a31f469c5c32c4788e19f2619d67a6514", "size": 2255, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium019.json|1777383639.754071|2002": {"hash": "d05809878b8983931046e86a9d35e15cf4a2bbe925318d63d980e2f192540e7b", "size": 2002, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium020.json|1777383639.754071|1935": {"hash": "00f0976df80593f8541b5a03ce8c2fa8a2d31ecafdab7c878c9ba6a397d6a003", "size": 1935, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium021.json|1777383639.754071|2097": {"hash": "ffa7fd777d0a355d31d5a5e777e02a52cdaed1e9a1be41868b8519062f8ec819", "size": 2097, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium022.json|1777383639.754071|1936": {"hash": "756f42227a3b315829a48d8245b6b6a93b0cbcb8f850fd0413127e9f25dec2be", "size": 1936, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium023.json|1777383639.754071|2015": {"hash": "b2a08b8a315e37611915c682325884c9cf5fd5ddb26b76e092904b0609d24f0c", "size": 2015, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium024.json|1777383639.754071|2132": {"hash": "2dd0fd16231f9b1427aa7418eaf25083c8d0eec5a585fe5ee1de54e8f83d829f", "size": 2132, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium025.json|1777383639.769922|1985": {"hash": "6c6d8330785f1f37d15eb9c0888a7ff1c6ff6c6dc2bf07ad5e586164b63597ed", "size": 1985, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium026.json|1777383639.769922|2185": {"hash": "a15581c8d6fc2a48f78e2835f2a3687d2dd45532113094dd2d5c187677eedd78", "size": 2185, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium027.json|1777383639.769922|2254": {"hash": "cbc40cc228898cb375fc403ea84db89c1600bf66089689ebd1f37d2015cc4f6c", "size": 2254, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium028.json|1777383639.769922|2109": {"hash": "48d7c1cc281822f61e739a663aeeda790b4768db318d3fbfab5dc0357ed46e1e", "size": 2109, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium029.json|1777383639.769922|1878": {"hash": "635a8ab4d3a5de4035716d7426692e92be0258451e5adbf5f71b665cb95c5753", "size": 1878, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium030.json|1777383639.769922|1921": {"hash": "a1aca41685e97fe96b936d1c061c0a9548c14950e41d430c5c1166dd0c93d5f4", "size": 1921, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium031.json|1777383639.769922|1969": {"hash": "6a20cdd2ce12d68e50311727efa8a141557db5e870aa77b7f627a950366a22b8", "size": 1969, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium032.json|1777383639.769922|1847": {"hash": "79c6fdd606f629bfcb2355cb06d8b9fe4cf9a0c1567b4164f83a8cf1fb612f28", "size": 1847, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium033.json|1777383639.769922|1930": {"hash": "3dce34aa8840603376201dd6eeeadaf974fdd0801081b07c7b54e70debea3f99", "size": 1930, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium034.json|1777383639.769922|2150": {"hash": "d9432b41a0f465798ef1b06d71a05afb7eedf68e8b5e561e91ea5e1f766a49c2", "size": 2150, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium035.json|1777383639.769922|1999": {"hash": "5286d074052bc92180f973fdc10efb2d18af8a8010f7102103312784301643aa", "size": 1999, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium036.json|1777383639.7860718|2048": {"hash": "5458a8695cd883dc775537323fbc5de71e5516bac43d3d376ed19ad8f0b45bd4", "size": 2048, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium037.json|1777383639.7860718|2409": {"hash": "fc4a0f2f591365cfa970cf837f93cfe058156b95d859d37f3f319f231b3a8535", "size": 2409, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium038.json|1777383639.7860718|2038": {"hash": "abe1b260e5c8dd14512998c0d9430c7549e24a856ff550b62eaf34b1c22204d9", "size": 2038, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium039.json|1777383639.7860718|2356": {"hash": "a24fc8bb992191a3e66042b3c70809145bf4ca2040e481bf562677a351a8f5e9", "size": 2356, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium040.json|1777383639.7860718|2130": {"hash": "28b09916052c39106c6fed46279f641685fa3da1c5ca403cf5fc9f7d605ead84", "size": 2130, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium041.json|1777383639.7860718|1879": {"hash": "d8277323b1866d868a07b8d3c0570a4120ebb778d1ee707d4be182b3edd15eb7", "size": 1879, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium042.json|1777383639.7860718|1922": {"hash": "0d1a229cab1cff61153b5cd1a18833e61ad7708f8e0b069a6d703888e0328e5d", "size": 1922, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium043.json|1777383639.7860718|1835": {"hash": "64b8cb0df9786decb288f3249be9f55c0fbf26c307a28f49f11bd17276bc42c4", "size": 1835, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium044.json|1777383639.7860718|2095": {"hash": "43b397e7d93365d79fdaae519d14020549057ef6865197c138b723b85946d436", "size": 2095, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium045.json|1777383639.7860718|1674": {"hash": "a374f19009a238a87be31db34163f01248fb7ec2e14726bc7f009c0d9ea28e81", "size": 1674, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium046.json|1777383639.7860718|1627": {"hash": "f4665e9ec5e44b927a7fe9fb13c441e2255caa0983f07ac1d1039bef7832bcbf", "size": 1627, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium047.json|1777383639.8018582|1880": {"hash": "3c70187585d965df23784f4dbb54641d01abf337e2ebce2d172819950f0705b4", "size": 1880, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium048.json|1777383639.8018582|1640": {"hash": "456debf17106cb31483df8bdc91e9d52ef0e7c551f393750883f97b65d1c734b", "size": 1640, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium049.json|1777383639.8018582|1780": {"hash": "35c4864539c21bc48f36c01e6215095848e325c40262db9f623ece073933287f", "size": 1780, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium050.json|1777383639.8018582|1430": {"hash": "52b029b418a5d11ba6e4256179e2916f9c1d89e599169b3b1a02d2de8be6e2ed", "size": 1430, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium051.json|1777383639.8018582|1308": {"hash": "3db2259db29abbd4840e496b6a1cb5e2aa9579ffb824fde9f20473bcd3443efb", "size": 1308, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium052.json|1777383639.8018582|1652": {"hash": "b868751d03c8a9006a7043ad2c8dff997d5df1dfa0df20eb988957de820bee0c", "size": 1652, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium053.json|1777383639.8018582|2035": {"hash": "4d851db5e13c49799482e7925fa691e12b04431aade8d037a9004e7d72808af4", "size": 2035, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium054.json|1777383639.8018582|1410": {"hash": "810a9b43c70a040300dd0ebfbe04884b9b6133dc77975c52cf86df465491bc8f", "size": 1410, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium055.json|1777383639.8018582|1714": {"hash": "a59a70908bfcf18d5345de15c4004250e2c91cbecc5afe90bdafb786a7767b9f", "size": 1714, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium056.json|1777383639.8018582|1579": {"hash": "99ddf85e615853d68f6b1cd91b892f3b81cd6cefe4ef50223c9263a31c16650f", "size": 1579, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium057.json|1777383639.8018582|1391": {"hash": "06b3d8cba9cd6e4924a7464c678bd3af6e98938b020132469b099f7e83797bc9", "size": 1391, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium058.json|1777383639.8178916|1571": {"hash": "80a1dbe89735cd4b8ad349d0b3e5092ca21b6228a7129568f6bdde8fe5e7c0b2", "size": 1571, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium059.json|1777383639.8178916|1938": {"hash": "48a2a357b0c75159b92a57ec7fbef9bd8e3b3dba87d7f86c3d10630f451842b7", "size": 1938, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium060.json|1777383639.8178916|1466": {"hash": "237e1c2688f9713329af1e6ee797b9e5beb3349dd8d2b0cc6148d3f4d1014250", "size": 1466, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium061.json|1777383639.8178916|1731": {"hash": "f93becd6d43ee5cf98f848eac6cd85570efd76ac46ce8176dbeb1d80c0e88c07", "size": 1731, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium062.json|1777383639.8178916|2076": {"hash": "87bc58d5106b93a34547338931ff96a356e6567f513820449502bdaf7cf56d6e", "size": 2076, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium063.json|1777383639.8178916|1501": {"hash": "058e3098891fcb05ed2754033760339380338a54d1a4b20b06e9ed51670134da", "size": 1501, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium064.json|1777383639.8178916|1893": {"hash": "d7826d7d4b0780be23e0e7e0180cb4370d468604ee951ac2e7f9eb71cedce723", "size": 1893, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium065.json|1777383639.8178916|1531": {"hash": "0b8888551870a1479cf509db031fccec4c49fef1fed4b9f9b58f1bad9b8d0dc8", "size": 1531, "status": "c"}, "assets/qa_gold/enterprise_industry_policy_analysis/medium066.json|1777383639.8178916|1592": {"hash": "88ad46eb0830bda62232595e23457eb6949e5ba70214c79f1ebf6a04fcd67508", "size": 1592, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard001.json|1777383639.8178916|3172": {"hash": "34f6f87b6aa7126ee9717edd2f860305f374f83176fda933e9f15c5c59888ac4", "size": 3172, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard002.json|1777383639.8178916|4233": {"hash": "9feeedf9269a14ce896808074aa549608e4e2147b1ccb339bfd2da56f936e9b7", "size": 4233, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard003.json|1777383639.8337624|3565": {"hash": "1963b0cf23c7dceadec5ac91e8da25c7baa4be4f66b1d36ba67fe929c94116d9", "size": 3565, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard004.json|1777383639.8337624|3721": {"hash": "3e14b802f688cd5f7295b169b5f824c7d6d3a188c1ba7b3d119b1a151aad1614", "size": 3721, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard005.json|1777383639.8337624|3476": {"hash": "7ae18c3fe13d08becce4eb9369719ee87f0962e8a1bebc77f49c1463bbdce422", "size": 3476, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard006.json|1777383639.8337624|3058": {"hash": "1512aa30d7cc01299f4134f7b5d3f5116f6fd512b7d4404bcc403010f7cb344b", "size": 3058, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard007.json|1777383639.8337624|3698": {"hash": "23c9166dded91f16a190318acf510d9806d2506667bf80bcacd8b6f22f6aa391", "size": 3698, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard008.json|1777383639.8337624|4356": {"hash": "a5dbd8c22a0da2f0dde8b9376a3baf580a847565834f3a0d3fbc38a5d7c7d0f9", "size": 4356, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard009.json|1777383639.8337624|2957": {"hash": "135dac15bf7dcc9d56f89d70f8dccdfd2a91a4e7d911f0ebff4607baa7bbccd0", "size": 2957, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard010.json|1777383639.8337624|4109": {"hash": "0a21f5fba7fc090ee54e8d11084dbc3cc362b367eebb0c6a60e497ba4a0196a4", "size": 4109, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard011.json|1777383639.8337624|3388": {"hash": "4b8425f4f175520a001aba87988a3e533eeb3fc5b8a38b7ffb0726430f7e5ef3", "size": 3388, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard012.json|1777383639.8337624|4841": {"hash": "2a5cd64fa53f96f44fb31fb933f099f38ed1a4906e273209185ccd6503dec458", "size": 4841, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard013.json|1777383639.8337624|5310": {"hash": "315af2d7a11d713a438ee35f10c41aefbc00cffb8ceaf76bd2f3af571cc97d28", "size": 5310, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard014.json|1777383639.8337624|4525": {"hash": "dbcf422becf30997d192930ace51448b0ab1ae239e2fd1143c751749b2ed8a1c", "size": 4525, "status": "c"}, "assets/qa_gold/hypothesis_verification/hard015.json|1777383639.8495376|4280": {"hash": "6583ccd567ec020883cd7cb9af5de82679149ec8364384270cb93f99bc574937", "size": 4280, "status": "c"}, "assets/qa_gold/hypothesis_verification/medium001.json|1777383639.8495376|2944": {"hash": "fbf39d57d4f29bc64951f28f74f6cf1e06394fb6aed30eb90604ea5e6a049b83", "size": 2944, "status": "c"}, "assets/qa_gold/hypothesis_verification/medium002.json|1777383639.8495376|2284": {"hash": "a1505a6bc12de53d9b2443570f81eb1b24df9a2b1b26bdc2501a11c9708109a0", "size": 2284, "status": "c"}, "assets/qa_gold/hypothesis_verification/medium003.json|1777383639.8495376|3160": {"hash": "07bfbc2f1dc3082187d05e56e45d9a2d3823278f591e2057f43aa36df6b52be5", "size": 3160, "status": "c"}, "assets/qa_gold/hypothesis_verification/medium004.json|1777383639.8495376|2673": {"hash": "dd613316008bb4d9c5b68625f57624ede3271538a2bc93a426404a262f5051b8", "size": 2673, "status": "c"}, "assets/qa_gold/hypothesis_verification/medium005.json|1777383639.8495376|2718": {"hash": "c93b732ebc152235bc00f3ef4d64dc6fa39432019e7da97254226e5509595954", "size": 2718, "status": "c"}, "assets/qa_gold/hypothesis_verification/medium006.json|1777383639.8495376|2036": {"hash": "ffb46e37bb37b3abcce177f3750b8a433f09561815d5cf99972697afb2d495d2", "size": 2036, "status": "c"}, "assets/qa_gold/hypothesis_verification/medium007.json|1777383639.8495376|4051": {"hash": "48df8a9e18bf52372777c9320e77d240c78e84ad35a6ac1f7e4ff59409b3b748", "size": 4051, "status": "c"}, "assets/qa_gold/hypothesis_verification/medium008.json|1777383639.8495376|2949": {"hash": "81fbc496f014a7a892c6470a74c7c8fe0c2a588a6a9ca7e6580e86111b87d0c9", "size": 2949, "status": "c"}, "assets/qa_gold/hypothesis_verification/medium009.json|1777383639.8495376|2812": {"hash": "e27624e212a9724e913b09ef1fc9e0e12c09af1ebdcd57386730bdec99cbf160", "size": 2812, "status": "c"}, "assets/qa_gold/hypothesis_verification/medium010.json|1777383639.8495376|3623": {"hash": "91e8430618fb91d8818fa087b5952d0eb8da4d150e93c727247be9518b12bde6", "size": 3623, "status": "c"}, "assets/qa_gold/hypothesis_verification/medium011.json|1777383639.8495376|2505": {"hash": "50265004e0e0c5ebc93836f6f1eaa5c037c137bbdccbcacd88c1e1c82289b16e", "size": 2505, "status": "c"}, "assets/qa_gold/hypothesis_verification/medium012.json|1777383639.865567|2490": {"hash": "e1e1373a389b9699618575b9c37cd8ee85ee40f5b6a20e842e674eb9552eabc0", "size": 2490, "status": "c"}, "assets/qa_gold/hypothesis_verification/medium013.json|1777383639.865567|2810": {"hash": "a43ab8bb5b1ce0ed85facd91f0b1d7cab56487fd85b410286a3b121c362afd5b", "size": 2810, "status": "c"}, "assets/qa_gold/hypothesis_verification/medium014.json|1777383639.865567|3475": {"hash": "5975b318b93118ec05bc7f6dde7528a22bca3aab23d601c2c21b72250783a7bf", "size": 3475, "status": "c"}, "assets/qa_gold/industry_planning/hard001.json|1777383639.865567|3852": {"hash": "8658c002edaa16fbf75ae5791ef7cb95988bcc454b86b88205d400a91d905214", "size": 3852, "status": "c"}, "assets/qa_gold/industry_planning/hard002.json|1777383639.865567|4092": {"hash": "1047b000fdd90c2c6dca372e0222b536ec331525640284204da3c7cdcee57b21", "size": 4092, "status": "c"}, "assets/qa_gold/industry_planning/hard003.json|1777383639.865567|5074": {"hash": "7f415244eb5380abe7a32d7adee9605f1580fa34e777ac9d293598621a4d7cb6", "size": 5074, "status": "c"}, "assets/qa_gold/industry_planning/hard004.json|1777383639.865567|4080": {"hash": "4f6d51625a53f22f3f35461689ec183463fd4acebb074525ef8cbdf036d709db", "size": 4080, "status": "c"}, "assets/qa_gold/industry_planning/hard005.json|1777383639.865567|3029": {"hash": "6dcfcd48c512ee903e5890bf65d870f2e352c1ebb0f4fed964138b8eb7a1b5e6", "size": 3029, "status": "c"}, "assets/qa_gold/industry_planning/hard006.json|1777383639.865567|3767": {"hash": "d63b3896cc1c074737f4230da1a8c8ab61b757d40686d65da129546b86f87eb1", "size": 3767, "status": "c"}, "assets/qa_gold/industry_planning/hard007.json|1777383639.865567|4545": {"hash": "ea1f1ab036199c49746476d1c2e390b3cc2b16736c7cc3b9bc010e2f2f98f289", "size": 4545, "status": "c"}, "assets/qa_gold/industry_planning/hard008.json|1777383639.865567|4208": {"hash": "f9acd48319cdadffc00a8cde11bacb71b7be89c31f7229f6d656f8779a46cff2", "size": 4208, "status": "c"}, "assets/qa_gold/industry_planning/hard009.json|1777383639.865567|4136": {"hash": "47e6df2749bfbe300f83565f42a7a49d9fb1c0bb4841315c24a3ea570fec1ced", "size": 4136, "status": "c"}, "assets/qa_gold/industry_planning/hard010.json|1777383639.8815129|4207": {"hash": "c472fc37a720154c83160945f6cb6b563ed5567d00127f6397324e8b31f51737", "size": 4207, "status": "c"}, "assets/qa_gold/industry_planning/hard011.json|1777383639.8815129|4045": {"hash": "d752e9c281e3bba37b97932e5c8c7c060ca4b8feb4277d292298a05093f72c85", "size": 4045, "status": "c"}, "assets/qa_gold/industry_planning/hard012.json|1777383639.8815129|4022": {"hash": "424a1d0d5392ee95f62845054f06742fbf39a55b915d939ae5fe0ca65771c3f8", "size": 4022, "status": "c"}, "assets/qa_gold/industry_planning/hard013.json|1777383639.8815129|3775": {"hash": "f98268689a50edde58b677de62a92fbd1bc78d2fbdff7bd5365161c9010b7803", "size": 3775, "status": "c"}, "assets/qa_gold/industry_planning/hard014.json|1777383639.8815129|3359": {"hash": "c073bcae3b42efc8ba03f90b0da8370000df0c291e5deb0388a9e83a24e8fccd", "size": 3359, "status": "c"}, "assets/qa_gold/industry_planning/medium001.json|1777383639.8815129|3130": {"hash": "ce2313161de7700c3287d983588533aee94cf7c0d7ac2b3ca740dd2657e9ee2b", "size": 3130, "status": "c"}, "assets/qa_gold/industry_planning/medium002.json|1777383639.8815129|3099": {"hash": "69f4021bec58cee29e72d6897c9fbe44e20711b11a80a22ae17abf67c41fcf63", "size": 3099, "status": "c"}, "assets/qa_gold/industry_planning/medium003.json|1777383639.8815129|3633": {"hash": "017c9ab9ef74abfafc83e49026a3790be4b4d3d19110a0ce41977794ceae9336", "size": 3633, "status": "c"}, "assets/qa_gold/industry_planning/medium004.json|1777383639.8815129|2668": {"hash": "a85a446b5e24e66790177d4fe90d4dd08dbe3fa9da4f8b14e242a0cfa9b3e893", "size": 2668, "status": "c"}, "assets/qa_gold/industry_planning/medium005.json|1777383639.8815129|3901": {"hash": "91d075fa53e417a9afe90e95fc844d1e9d18b0ddf8f97f1a3e573ec0b9b742a6", "size": 3901, "status": "c"}, "assets/qa_gold/industry_planning/medium006.json|1777383639.8815129|4482": {"hash": "c31b302050fb2d7f3a4a4b664bd6d22a28318bddfbb81bfec8b5e84e89613bf0", "size": 4482, "status": "c"}, "assets/qa_gold/industry_planning/medium007.json|1777383639.8815129|4718": {"hash": "af0c00b37a9afc96e7491db5e1bd9769f2c6cd385ceaba55c0040b689d89541d", "size": 4718, "status": "c"}, "assets/qa_gold/industry_planning/medium008.json|1777383639.8973494|3398": {"hash": "a10b0064b4edcdcc94a08a423b42d07503f6620d286ca18ebde2cd08e43019ba", "size": 3398, "status": "c"}, "assets/qa_gold/industry_planning/medium009.json|1777383639.8973494|4409": {"hash": "d64c5c600a8c84ade603308b4de0e8ace72b2c20ced5d2311f69855b3f8ab1b9", "size": 4409, "status": "c"}, "assets/qa_gold/industry_planning/medium010.json|1777383639.8973494|3412": {"hash": "812b8675d4ef569f0774216537ee9d51e0be1b17dfeb33ef6e7dc789e5c0fe25", "size": 3412, "status": "c"}, "assets/qa_gold/industry_planning/medium011.json|1777383639.8973494|3813": {"hash": "74cb95bd888eb365e203b7a8ab427e3da7ea9447bc4fa45b8e588f953d5e1470", "size": 3813, "status": "c"}, "assets/qa_gold/industry_planning/medium012.json|1777383639.8973494|2954": {"hash": "5744350bbece2c75f531138b3d13d488cdc1b15832a0bb240f2934e068fa520e", "size": 2954, "status": "c"}, "assets/qa_gold/industry_planning/medium013.json|1777383639.8973494|4798": {"hash": "29ac22c721414121cf382776ac51a90e1d35c0d114f2b6f83d692f11aa686e77", "size": 4798, "status": "c"}, "assets/qa_gold/industry_planning/medium014.json|1777383639.8973494|3941": {"hash": "9ba5e32a17aa540300e09517da2d95aacd2e8d6e0a2088d80510aebfc43a3f39", "size": 3941, "status": "c"}, "assets/qa_gold/international_comparison/hard001.json|1777383639.8973494|3184": {"hash": "f85d07026edefd8552601ab09823b5737377a27679408e393f59256805706c2e", "size": 3184, "status": "c"}, "assets/qa_gold/international_comparison/hard002.json|1777383639.8973494|3368": {"hash": "aded51b8a1ddd154d2619e71ba35b02ef5e26f3dcdb49826b7a72c9768f6c2ac", "size": 3368, "status": "c"}, "assets/qa_gold/international_comparison/hard003.json|1777383639.8973494|2988": {"hash": "529280196216612219b842c33be6c6d97f716efb257407f9967fcf253ac0eab7", "size": 2988, "status": "c"}, "assets/qa_gold/international_comparison/hard004.json|1777383639.8973494|3152": {"hash": "65186ab4d01b4352c6796f584a65f58a014f9c00424132da6e9886879957faee", "size": 3152, "status": "c"}, "assets/qa_gold/international_comparison/hard005.json|1777383639.9131937|3456": {"hash": "587ac278c9234b4a116513278d44e06ce657cca618e282831b4056a81e71a298", "size": 3456, "status": "c"}, "assets/qa_gold/international_comparison/hard006.json|1777383639.9131937|3334": {"hash": "f85233bc856688f360e0b353fe8a00dc1da334b90e82487bb7c68f205cd516ef", "size": 3334, "status": "c"}, "assets/qa_gold/international_comparison/hard007.json|1777383639.9131937|4090": {"hash": "a14a594a2b4505671716f7e363c963005cc8e99a8885bbb778d7da4f9054fd57", "size": 4090, "status": "c"}, "assets/qa_gold/international_comparison/hard008.json|1777383639.9131937|3405": {"hash": "e8b370e7073c8db80a6204ed3214d934b37748ba07d0400eb05168642b3acc24", "size": 3405, "status": "c"}, "assets/qa_gold/international_comparison/hard009.json|1777383639.9131937|3577": {"hash": "bb8abaf2c270dcd5a2f36b07b71c3971025762bd8cad56b6e841c8fbe07524ed", "size": 3577, "status": "c"}, "assets/qa_gold/international_comparison/hard010.json|1777383639.9131937|3043": {"hash": "56734dfcbb6ecdc049616ab634d26133bd235a021db92e8ce8288ac6ab67b43f", "size": 3043, "status": "c"}, "assets/qa_gold/international_comparison/hard011.json|1777383639.9131937|2751": {"hash": "75d996e3d3decefa750f0653ff6ec3c4894e0ba44e033cf8c8532551493d9636", "size": 2751, "status": "c"}, "assets/qa_gold/international_comparison/hard012.json|1777383639.9131937|3607": {"hash": "00634baa9e266365d3d9d7ff849c19f6e3ec277c67dcf5212bc88edee8f03e47", "size": 3607, "status": "c"}, "assets/qa_gold/international_comparison/hard013.json|1777383639.9131937|3333": {"hash": "a38f07de31597fac6c1b9abfcf4e4263c4b6938911d0a3e226b980036da1a210", "size": 3333, "status": "c"}, "assets/qa_gold/international_comparison/hard014.json|1777383639.9131937|2756": {"hash": "39d98c5f6a75dbcd02da342896c20a3983e24cc0334b9653bb5233a549ba7ddb", "size": 2756, "status": "c"}, "assets/qa_gold/international_comparison/medium001.json|1777383639.9131937|1810": {"hash": "0b0f3e846bdef2afaa73874a2fc3ca45fd128e9eee3f84e6253247018661afe9", "size": 1810, "status": "c"}, "assets/qa_gold/international_comparison/medium002.json|1777383639.9289045|2091": {"hash": "dd4510f2edcb3c3cc4c080c6cff1c2a2a5237140100bd33af2ca547af1edc1f5", "size": 2091, "status": "c"}, "assets/qa_gold/international_comparison/medium003.json|1777383639.9289045|2174": {"hash": "c9ac4381026aef55e8757aa649438e416bc4afa078bdf4988828add53c071e0a", "size": 2174, "status": "c"}, "assets/qa_gold/international_comparison/medium004.json|1777383639.9289045|1968": {"hash": "6e75ba8fdf68b620cce5edefbb5c3049f378c7b4764a0aae30600ec4fe4c707b", "size": 1968, "status": "c"}, "assets/qa_gold/international_comparison/medium005.json|1777383639.9289045|2137": {"hash": "ec507e0fa04071b09161b8209a8682842c5ab3f67dd0766b32349565c5214dbe", "size": 2137, "status": "c"}, "assets/qa_gold/international_comparison/medium006.json|1777383639.9289045|1806": {"hash": "c5e8c0802e847b480c58660ba49bd7cccdf22252abd86aa5f392ebfad1f35396", "size": 1806, "status": "c"}, "assets/qa_gold/international_comparison/medium007.json|1777383639.9289045|2043": {"hash": "157b0522afe7c9a6a63a105c56621cf56afee7a3816af9e6efe4ee208dbda925", "size": 2043, "status": "c"}, "assets/qa_gold/international_comparison/medium008.json|1777383639.9289045|1984": {"hash": "89dd41060a68e268a6fef189e4f251cb32e689b72e287003207a292cb18d4b03", "size": 1984, "status": "c"}, "assets/qa_gold/international_comparison/medium009.json|1777383639.9289045|2565": {"hash": "8dd7d0aacdcd5dbe2f4872fccd6f3c866b8969d4a09c8740969933c94385d96a", "size": 2565, "status": "c"}, "assets/qa_gold/international_comparison/medium010.json|1777383639.9289045|2091": {"hash": "b9513bd7b7a5e80935fb4b696dfb46606effa42298db08dd1da21a5618bfb540", "size": 2091, "status": "c"}, "assets/qa_gold/international_comparison/medium011.json|1777383639.9289045|2011": {"hash": "25ccb9aad1f206765d61e8e28a4d2e852b443d18b07ef651c97f4ef44ee7eb5f", "size": 2011, "status": "c"}, "assets/qa_gold/international_comparison/medium012.json|1777383639.9289045|2110": {"hash": "6548734617a874dff7dff5b229041096131f8ce53f626cd2d88da35c015b3927", "size": 2110, "status": "c"}, "assets/qa_gold/international_comparison/medium013.json|1777383639.9289045|1904": {"hash": "1787c4b9efae251f96c0060ae2a6ccc99cd07ccff8726e3989a311136491c643", "size": 1904, "status": "c"}, "assets/qa_gold/international_comparison/medium014.json|1777383639.9446163|2628": {"hash": "c8d38f10986c1e063ec4588d9a1bf7f293b980a95a3e37a9f969cd692da07c74", "size": 2628, "status": "c"}, "assets/qa_gold/international_comparison/medium015.json|1777383639.9446163|2794": {"hash": "814b25af6bc939d80f855399bdd0b20ce0407a2c6f12b634765f1f3d8bf82192", "size": 2794, "status": "c"}, "assets/qa_gold/international_comparison/medium016.json|1777383639.9446163|1912": {"hash": "836da6d04ca6fff2715be34cd2f7d14073d71bc07334a809db3b5cc612b22d16", "size": 1912, "status": "c"}, "assets/qa_gold/international_comparison/medium017.json|1777383639.9446163|2942": {"hash": "d9fdd001a14ac2ba81850eed350088a4d1fbd9b931f3d72efef86f42d38934ba", "size": 2942, "status": "c"}, "assets/qa_gold/international_comparison/medium018.json|1777383639.9446163|2055": {"hash": "7bf376705fcf31b694195e555043caaff823679908dfcf9688f8cbeea897dc98", "size": 2055, "status": "c"}, "assets/qa_gold/international_comparison/medium019.json|1777383639.9446163|1994": {"hash": "5f75f7b63112e870064cfaa37096dff1780cfb8e707d1d2d6f0f7e6dd239b630", "size": 1994, "status": "c"}, "assets/qa_gold/international_comparison/medium020.json|1777383639.9446163|2642": {"hash": "fb7fd6b062822720891238b6a71775727e98f48d5b4b2b0866e5c0f67be366db", "size": 2642, "status": "c"}, "assets/qa_gold/international_comparison/medium021.json|1777383639.9446163|2359": {"hash": "f56de92a17c727551452fcc13ac8d0c75e2754a3b4c635d9d9c9de6916cc8d4d", "size": 2359, "status": "c"}, "assets/qa_gold/international_comparison/medium022.json|1777383639.9446163|2709": {"hash": "3a7719d5926999e3b2fb356042de0203c089d90e6802faecc87358d294527873", "size": 2709, "status": "c"}, "assets/qa_gold/international_comparison/medium023.json|1777383639.9446163|2607": {"hash": "31873cbdbe94820eb5e2c7c181eeb0d69fe58ff8d8b68440edaca3b173615e9e", "size": 2607, "status": "c"}, "assets/qa_gold/international_comparison/medium024.json|1777383639.9446163|3019": {"hash": "c8fa4d95c409c88752fee95a66f1ac3f3aad2ddb825d1689a753267aff5d3418", "size": 3019, "status": "c"}, "assets/qa_gold/international_comparison/medium025.json|1777383639.9446163|1948": {"hash": "de4e6f87655636f8b69e2cbaa2685617e35070e04c89e83aca6ffdae08da9216", "size": 1948, "status": "c"}, "assets/qa_gold/risk_assessment/hard001.json|1777383639.9603226|2671": {"hash": "b28a2fa7b324e56eba1361c448d2655ab0dcf5598ebc9b18bf7a79395929536b", "size": 2671, "status": "c"}, "assets/qa_gold/risk_assessment/hard002.json|1777383639.9603226|2670": {"hash": "1821cf7f3b9c9c63993687a0dbef1ffc39c910aeb6902781efee253fdf316658", "size": 2670, "status": "c"}, "assets/qa_gold/risk_assessment/hard003.json|1777383639.9603226|4155": {"hash": "c44a8096e8784a4b5a415d63d01eb7348d1f9fcdb63ca22efe7db44a6e6ce064", "size": 4155, "status": "c"}, "assets/qa_gold/risk_assessment/hard004.json|1777383639.9603226|2816": {"hash": "09c96cfc59e7297d288a808839e15aecd7fc0d5ec3a06dd2b141895719c8df4f", "size": 2816, "status": "c"}, "assets/qa_gold/risk_assessment/hard005.json|1777383639.9603226|2991": {"hash": "0fe463ae1b80e9c25405d4afbabfca1afb9fa2c73205d3c308796a9f236bf0f2", "size": 2991, "status": "c"}, "assets/qa_gold/risk_assessment/hard006.json|1777383639.9603226|3412": {"hash": "7beb9cbff45cf24c19c4bfe6250255257ae383bd7886037774b875743be9e13f", "size": 3412, "status": "c"}, "assets/qa_gold/risk_assessment/hard007.json|1777383639.9603226|2400": {"hash": "2980a40e455c25dfe017cfacedbc7c31d8786fb3460671d8f3a38d07abf301ed", "size": 2400, "status": "c"}, "assets/qa_gold/risk_assessment/hard008.json|1777383639.9603226|1986": {"hash": "6f543b1a4ee8b7ba73d3ad86e4ed1a5c6bca87457776694b1532b332d0e772fe", "size": 1986, "status": "c"}, "assets/qa_gold/risk_assessment/hard009.json|1777383639.9603226|3084": {"hash": "3d92bad27a2d1838ad96bcbb3e3427d4faa6a4ff50d797761860c79f639834c5", "size": 3084, "status": "c"}, "assets/qa_gold/risk_assessment/hard010.json|1777383639.9603226|3582": {"hash": "dda899a4145a12bf2d7e3318e76909f8008014589212da75522853efeb666e20", "size": 3582, "status": "c"}, "assets/qa_gold/risk_assessment/hard011.json|1777383639.9603226|3652": {"hash": "717d12db0d79004a9b9beba6349fd0bb96fd67a71f224c7527a31c0428bd98df", "size": 3652, "status": "c"}, "assets/qa_gold/risk_assessment/hard012.json|1777383639.9603226|2554": {"hash": "204284a486939a2061e3b724b1066ec9fa1da8a75c21959f3e2953b765952693", "size": 2554, "status": "c"}, "assets/qa_gold/risk_assessment/hard013.json|1777383639.976122|3148": {"hash": "b0d613125190b2d7ba7816ef863b8431836b52f60016729c13290149f89928e6", "size": 3148, "status": "c"}, "assets/qa_gold/risk_assessment/medium001.json|1777383639.976122|2981": {"hash": "c9f383cd21791ea2a67602472a3925bc2ef99bcc43b14a8d1d7252980804481b", "size": 2981, "status": "c"}, "assets/qa_gold/risk_assessment/medium002.json|1777383639.976122|2871": {"hash": "4b1b33a208f16e8ae0756fdb73bb1dd5586741b14b0a06abe6e0fbf9a82c416f", "size": 2871, "status": "c"}, "assets/qa_gold/risk_assessment/medium003.json|1777383639.976122|3246": {"hash": "ebb2bfbcfb1eaba2b7397d69ee6ee0d696b02483e051314b33d8bc86876daf6b", "size": 3246, "status": "c"}, "assets/qa_gold/risk_assessment/medium004.json|1777383639.976122|3322": {"hash": "cea07744ccb86ea4577651d617dd5f9bc6c55012c2974a9efa1e754fb4256761", "size": 3322, "status": "c"}, "assets/qa_gold/risk_assessment/medium005.json|1777383639.976122|3310": {"hash": "21acaff94d0efbbf514ecfeb1d8e6f05b940a4030719b545ac287115574e47ac", "size": 3310, "status": "c"}, "assets/qa_gold/risk_assessment/medium006.json|1777383639.976122|2944": {"hash": "869657e4528188a49b66b14f6ce0d6c1c3d161e314563cc0903bf32cf434af5f", "size": 2944, "status": "c"}, "assets/qa_gold/risk_assessment/medium007.json|1777383639.976122|3622": {"hash": "38d81b6440bea074255edf918ad2ce312009a3573b4564542af226d428bde431", "size": 3622, "status": "c"}, "assets/qa_gold/risk_assessment/medium008.json|1777383639.976122|4419": {"hash": "a4ca5770ee168d5f952e56b5e08a82589dcc4a636861d89825147b12486dcf1b", "size": 4419, "status": "c"}, "assets/qa_gold/risk_assessment/medium009.json|1777383639.976122|3129": {"hash": "17060248453cc62224e0b0a15268bf35c67ca3957e8541f05d6449fb10fbd328", "size": 3129, "status": "c"}, "assets/qa_gold/risk_assessment/medium010.json|1777383639.976122|3184": {"hash": "c5cae1132861e2df4f74f00a4a0eaa45e51beed8fe68c77f38a34d64441a9e29", "size": 3184, "status": "c"}, "assets/qa_gold/risk_assessment/medium011.json|1777383639.976122|3359": {"hash": "5a124fa2321bbd441148a350852efb3e413ba96e1cd021c116890336f3f2ed58", "size": 3359, "status": "c"}, "assets/qa_raw/comprehensive_decision/easy001_result.json|1777383639.9920673|1583": {"hash": "b59af64fcc88e94c1d32130a02e29f55ee5ebeb230ffa824fae5a65b092b5ab8", "size": 1583, "status": "c"}, "assets/qa_raw/comprehensive_decision/easy002_result.json|1777383639.9920673|1754": {"hash": "bf38744d7b01a9b4af2eea341741d9018b423740620381b7f466272f6f46eba3", "size": 1754, "status": "c"}, "assets/qa_raw/comprehensive_decision/easy003_result.json|1777383639.9920673|1603": {"hash": "4b3e108d6903c40d8a80c4aec9454c149d7bebf052d3bbf433b8573d7aa3aa5c", "size": 1603, "status": "c"}, "assets/qa_raw/comprehensive_decision/easy004_result.json|1777383639.9920673|1779": {"hash": "4899700d1c723537071879009ed164b309d62ca71ab45809122b7c34f37b475d", "size": 1779, "status": "c"}, "assets/qa_raw/comprehensive_decision/easy005_result.json|1777383639.9920673|1924": {"hash": "2681d28724d4fd695071d904c6594b9009cf64f24c2d8c8003fee5b704935257", "size": 1924, "status": "c"}, "assets/qa_raw/comprehensive_decision/easy006_result.json|1777383639.9920673|1346": {"hash": "595de82b4d316c44950e6214e33df5e0ca5666ec6df4b5af957374a12ec6e259", "size": 1346, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard001_result.json|1777383639.9920673|3968": {"hash": "2f3f1274665e83212326c2ebea28cd1aab898781b238315394d278ead88d300f", "size": 3968, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard002_result.json|1777383639.9920673|3784": {"hash": "db87a7ce4bd216f28903c4368c165e94b9bfbc7a28b384490ae840d9a91e4bc6", "size": 3784, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard003_result.json|1777383639.9920673|4470": {"hash": "2a927110487ba6f292ed5d6cafe41c405fd17fe8ae3e39954579f69a6f9f4d50", "size": 4470, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard004_result.json|1777383639.9920673|4256": {"hash": "b804b610be24f1cfc40e77be59475e0b1e088591d5c93bb5ec02116b32bd4b36", "size": 4256, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard005_result.json|1777383639.9920673|3702": {"hash": "e7429fb3f58986958d9f5e125749c1819457ff971d224de355b78dc226cf355a", "size": 3702, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard006_result.json|1777383639.9920673|3336": {"hash": "5224f6f5d84d17acc9329aeb238b0c2b3047b28c47afd7fd79cdf6d7fa3bc805", "size": 3336, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard007_result.json|1777383640.0081546|3688": {"hash": "0386af2bacf3373866c20937a62e2bcb5d05d45af033c54c0e0c00d5999915d2", "size": 3688, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard008_result.json|1777383640.0081546|3644": {"hash": "67150da31256ebc21ae669baa4dbccb8ca30bd7e5db6d6346b214c39b4145539", "size": 3644, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard009_result.json|1777383640.012844|3825": {"hash": "da4fbdfa34a5667b5c7d400a8eeea38770bf15b73b81f98e4948fbd8ae18fb76", "size": 3825, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard010_result.json|1777383640.012844|3684": {"hash": "9f2117e518ad8961ed656c48e4cf74e98d463c5f6d19aa21012931a5b876d060", "size": 3684, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard012_result.json|1777383640.012844|3841": {"hash": "3d756f395f48bcda8c0d46a2d0832fd3816ac6f2bf507ec7a9e5b087b5352674", "size": 3841, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard011_result.json|1777383640.012844|4067": {"hash": "9c1e8d8d2d492a80b20ff05db1556284b9e303a4bb2c387e24df74bc5c303f30", "size": 4067, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard013_result.json|1777383640.012844|3282": {"hash": "2b3e652259078495ed0cdca1e255a075f89d0339f5f28bd54842d4585219426e", "size": 3282, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard014_result.json|1777383640.012844|3519": {"hash": "e4b1a965aa8d9ae686d6c4733aa805e1385cb971c7dee2aa2b039eebeed84064", "size": 3519, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard015_result.json|1777383640.012844|3952": {"hash": "dcf8e5fd9b4b14f8b01ea88021d277f883e37836d59d2f9afe8b236cbc5fa8d5", "size": 3952, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard016_result.json|1777383640.0238006|3240": {"hash": "02176a2ba1842aa7c5a7ada348e83e58be9f3470f5065f0e55101511a02d06a5", "size": 3240, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard017_result.json|1777383640.0238006|2865": {"hash": "89487ccf34dad114f8b4e3412d3c72445f3ec1c6efaf8cb84de3568600cb7795", "size": 2865, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard018_result.json|1777383640.0238006|3053": {"hash": "82e18799d179527ab58fc31257d00a888ea8ee1183fd7de80bcf646dfa0e0868", "size": 3053, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium001_result.json|1777383640.0310652|3291": {"hash": "29429e8746f85b5752ee4397601968b6652e328a01feca5745b8283db80f14c0", "size": 3291, "status": "c"}, "assets/qa_raw/comprehensive_decision/hard019_result.json|1777383640.0238006|3723": {"hash": "eba2536895517199908109d808fed8a8b7047e816df221f950dcbe1e5328f45c", "size": 3723, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium002_result.json|1777383640.0330737|2712": {"hash": "3161d79c3d13cfae3b037067a71cc34c7db87e407d21adbb90b5bffaee37fa73", "size": 2712, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium003_result.json|1777383640.0350826|2848": {"hash": "83fd4bd4a091d3583e5bb9989ccd830a6431146564fe15aa2679cfd39291b510", "size": 2848, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium004_result.json|1777383640.0350826|3419": {"hash": "64b795407a762303793925904bedf10c4ea5d9fba84bf46c14b24e1dca8bd00a", "size": 3419, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium005_result.json|1777383640.0350826|2108": {"hash": "115f5cd12e9b2556b23bc8c23622245dcd593fe61c946ea88578367af2ccfce0", "size": 2108, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium006_result.json|1777383640.0350826|1837": {"hash": "d7fd61985c188da2b1b284d7ec1a5b17e4b4a17a8a5c987c4d3cd65be999b9fc", "size": 1837, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium007_result.json|1777383640.0395706|3165": {"hash": "6b15fde52349bcd52a955a7c0ead6345a35a04b03d50c19c5d6374176a62f8e4", "size": 3165, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium008_result.json|1777383640.0395706|2015": {"hash": "d36d737dd25cac80c2368a43b53cfa6f59caada5d1d8b4de4ddf09fb1922cbb1", "size": 2015, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium009_result.json|1777383640.0395706|2248": {"hash": "1d8e3f55ff0ccf15dd5a1788bb190ed1f8178650b577565fcdcf45abd3e19475", "size": 2248, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium010_result.json|1777383640.0395706|2206": {"hash": "4425fc0bbbd06475b23393f1d0adb492b864f77d386c9aa824e9eb7c33388155", "size": 2206, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium011_result.json|1777383640.0395706|1818": {"hash": "2a3c76759c4fff72af5a9342f52942f4f51aa5e0d717cccbb1b19a2dfd0f0cd8", "size": 1818, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium012_result.json|1777383640.0395706|2843": {"hash": "ac4a0a4968a916ccd80cbba440e05804ba8ae72419bbca395bef9e778efe09ed", "size": 2843, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium013_result.json|1777383640.0395706|3586": {"hash": "fba64ac68604a4aaf04b74aa4854845415567f0830674bc9195fa8bccf44aa0e", "size": 3586, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium014_result.json|1777383640.0395706|3526": {"hash": "205e456963a99bfde2b211493254dbcc3a8a52b13cdd960ca36eae7ffa47578f", "size": 3526, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium015_result.json|1777383640.0538294|4231": {"hash": "e0952a5d91c863ff5f0377115c31414fc5ddb435ca363566d68ddc6b1d82b6b1", "size": 4231, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium016_result.json|1777383640.0538294|1964": {"hash": "24bf6da9117c83a4589694477e80ee8502682b791cb944ab7d7300c75d611dbe", "size": 1964, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium017_result.json|1777383640.0555632|2546": {"hash": "610c31ffe028bda5f9fd5db33c7648af072504b8793d6dcb44e21d750c3e5c38", "size": 2546, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium018_result.json|1777383640.0555632|2087": {"hash": "9915648518a3acf848defa93aa20f7b4655197e8ff6a607a67853e3d4fec3032", "size": 2087, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium019_result.json|1777383640.0555632|2628": {"hash": "13c49256193489eb5dc0030d805a999843ee8582406266a34ad5b3639b0e43da", "size": 2628, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium020_result.json|1777383640.061738|3895": {"hash": "53bb220cdf98c4a62e1f09af727ae648b59317dd7621fef723b0648ee29b8fa0", "size": 3895, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium021_result.json|1777383640.0637474|3500": {"hash": "81dcce295ceb3db8ec26dd149fe9f808d6cb329cc1f713159792e4d0570f73df", "size": 3500, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium022_result.json|1777383640.0637474|2038": {"hash": "934c87623236edddcddabfd3f39618e2560b876184a46d9fbedd934d92480223", "size": 2038, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium023_result.json|1777383640.0657544|2148": {"hash": "fd0a13d054687c7b9cfefe81a888d189f9e44f34d885bff58c04f6541ecd37c4", "size": 2148, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium024_result.json|1777383640.0657544|2210": {"hash": "a576c6475b09c543798c151172dea99f0c08805d420be12121fe3509159b3592", "size": 2210, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium025_result.json|1777383640.0657544|2542": {"hash": "c05c4800d4303f1a315fdae531addb357b0994dddba35385923880f1019b2a8a", "size": 2542, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium026_result.json|1777383640.0657544|2373": {"hash": "4c42f26ee503c9714ad69feacd8aa6da1b7b98572b7b9deb1c040ba8ded49f82", "size": 2373, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium027_result.json|1777383640.0713518|3082": {"hash": "ba19380c95f8cf2130b06350adc35ac7b2d68b30ff3d853903efc78594d86f52", "size": 3082, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium028_result.json|1777383640.073991|4438": {"hash": "5a7108cfec0edbdbc4b37f18b9e00e8305e96e17d67d8ffe567bf8f71bc2e2a0", "size": 4438, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium029_result.json|1777383640.0778377|2953": {"hash": "dbc9f62235ed88c5d27c151b5737aa0d147e8f58224e1588436952c7e34bfb90", "size": 2953, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium030_result.json|1777383640.0778377|3067": {"hash": "3b4857386ad29ed2721f86a9aafd20479d876a34d923f20ad7bdf975917a4897", "size": 3067, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium031_result.json|1777383640.0778377|2207": {"hash": "e14da4d93d7b62b67a1c67e82df30ce59b1e6018466a7e60a912971e2fd1e5e3", "size": 2207, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium032_result.json|1777383640.0778377|2990": {"hash": "694c5dc7f24cd9abd644f5c409e58e5e1f87595c7393c7aa0c7c2ab1554e2219", "size": 2990, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium033_result.json|1777383640.0778377|2278": {"hash": "e099d08b56a6ee882ab64e5b1208680235170c361a6cabefb47af1a24e883e1c", "size": 2278, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium034_result.json|1777383640.0870187|3483": {"hash": "90f01627dfddd7284ad2bb2f5ae6707df5d09b0e4adc09d9cd9f1d04e3cb316e", "size": 3483, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium035_result.json|1777383640.0870187|2464": {"hash": "1691d03edc769a7d00dfdbe31ba21dcb8a1eb98496dff2297cef2b9086760a02", "size": 2464, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium036_result.json|1777383640.0870187|2280": {"hash": "0df77cfb601d6738a322d02ea0134a8949c234022498fcc44b44562eb014f815", "size": 2280, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium037_result.json|1777383640.0870187|2276": {"hash": "6bce83074471ec0fd76ffe92e83ced72669f420c3499e6b97af439fde76b7b80", "size": 2276, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium038_result.json|1777383640.0870187|1930": {"hash": "5f096fa5b078ec2b21383bedb089514a43a72a8a8aad1752731d23c0cc11e8eb", "size": 1930, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium039_result.json|1777383640.0870187|4296": {"hash": "2cff050bc5cb85a6ec88efda5a43017a031f572ed7510b02270a4e2afa5fec9e", "size": 4296, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium040_result.json|1777383640.0870187|2388": {"hash": "edc83cdb4e87000a59d3de2189a01dba3ae6fdfa3fbe6670c7863b9676b7f7fc", "size": 2388, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium041_result.json|1777383640.0870187|2158": {"hash": "b429e0b7889b07aaf941065ecb629dae7f4a86ee15c4f192719e5569bcc7d752", "size": 2158, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium042_result.json|1777383640.0870187|2194": {"hash": "55ae1344d890d7be1244e788913d0bcacc77697ab7f4c6fc78f3fa4562d5130a", "size": 2194, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium043_result.json|1777383640.0870187|1483": {"hash": "2ecb4cc91814862bb402873b59437c2e11df0a072b543b55bdf6070b05489a68", "size": 1483, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium044_result.json|1777383640.0870187|2495": {"hash": "71cc0307dca90240a222c47ef17587abf4384c1c86b3a22946df6af3fa63d9a9", "size": 2495, "status": "c"}, "assets/qa_raw/comprehensive_decision/medium045_result.json|1777383640.103081|2322": {"hash": "caa2b5385304eb941572b023b06f38a753c0775f9eaf9b7a075b054b65a0684e", "size": 2322, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy001_result.json|1777383640.103081|1954": {"hash": "8b72a75a3466b494f22c30f3a8d146f15162c3c56f9896b9d774722f458e48d6", "size": 1954, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy002_result.json|1777383640.103081|2009": {"hash": "e05cc104773b6c53f793d17441a175431f6b069b996982dc79dff1e2f298d31a", "size": 2009, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy003_result.json|1777383640.103081|2196": {"hash": "cb29a12c3f12e1442acc000e2e0f2395ecb1596ecffa95617a21da4c90a4e382", "size": 2196, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy004_result.json|1777383640.103081|1910": {"hash": "57ad2292cb328431178eda71d174460f25a3b0e55097b50ece33e01e6b60079a", "size": 1910, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy005_result.json|1777383640.103081|1896": {"hash": "e623fd00ec15e62550f7878dd608421984bcf7229c72b25caaa0f6efb4dacc6c", "size": 1896, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy006_result.json|1777383640.103081|1976": {"hash": "54b25988cdaf894802850f3ae3b6fdeffe0b5e7ee31306c5b982b45cd6960a33", "size": 1976, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy007_result.json|1777383640.103081|1905": {"hash": "6c96f461549dc32e0fd61ada64927474714f48b9ff645914e4a064b3cf97dd11", "size": 1905, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy008_result.json|1777383640.103081|1962": {"hash": "d2d30e62366d1c462067407221ed64222b3da2c84b23754c84319d04293747ad", "size": 1962, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy009_result.json|1777383640.103081|1755": {"hash": "1df89d2084d1a0508291669cf7d941ab5fe7eabfeb01374a83392bdd4bf5a559", "size": 1755, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy010_result.json|1777383640.103081|1921": {"hash": "7688262205f9d9631caafe1c51d8b8143d38ad080ce2820b580c1c137648f991", "size": 1921, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy011_result.json|1777383640.103081|1786": {"hash": "7d2e451ee6dbd0afdec8d8759a8f8bf9aedc97eda4004c0cc8af408b58728aad", "size": 1786, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy012_result.json|1777383640.1187773|1858": {"hash": "6570d1c54ef9be698aa88a31fd387c232e8157c5aefb1048d7d7bfb61fe8faff", "size": 1858, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy013_result.json|1777383640.1207995|1833": {"hash": "58b0bfe10d8a7fd9dba3455088aaa8382d3119bae3b08d83ea7815b03c262730", "size": 1833, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy014_result.json|1777383640.1228132|1859": {"hash": "d89bb202ff34b5637363e9d65130af4ff9d78aa6074b9de7b3de24e1dd4bfd1c", "size": 1859, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy015_result.json|1777383640.1228132|1905": {"hash": "1978955710188dd431fa321d04a22a3ce9dcbade42b8f10bb595d89638d9ad18", "size": 1905, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy016_result.json|1777383640.1248188|2030": {"hash": "a6e63963f0bad0e1991a48ca4a2c9f48ff5293f623144616da9c325eedc367c1", "size": 2030, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy017_result.json|1777383640.126825|2109": {"hash": "1e19dc1bb587682ce2338a558a2fc521dc18d333f7c35fb64eec302fa7d79ead", "size": 2109, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy018_result.json|1777383640.1288526|1950": {"hash": "b348cca0df3edb411e63befebe0a1ba46fde16fa87ce1d39f2e1c55a12e26bd3", "size": 1950, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy019_result.json|1777383640.1288526|1849": {"hash": "bb96d96e5e0254466821ffc58a244a2a624066281993476796d274c4c6c53214", "size": 1849, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy020_result.json|1777383640.1308582|1854": {"hash": "e0cc1794bdb7dea84205df59e33a1d7fdd5c0f0633b3be9fe37855c80e98f480", "size": 1854, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy021_result.json|1777383640.1308582|1835": {"hash": "300f9d9ddbdc1ea3bf74b763f69c1d0a722db65e14021ef88b7bab67ac0927bc", "size": 1835, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy022_result.json|1777383640.1328633|1938": {"hash": "4e77caccb97e30b5a45b16d167754aa15ffe1420c0d5720d186523e92336f8b4", "size": 1938, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy023_result.json|1777383640.1346092|2040": {"hash": "b7198e19c4360e83674b7b636ddbb033efb051087fadae34ea3954af9c4d0b70", "size": 2040, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy024_result.json|1777383640.1366239|1928": {"hash": "9bf87c365337491f8fb1e251e0a9655afa94225be909002966c93ed023155795", "size": 1928, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy025_result.json|1777383640.1366239|1958": {"hash": "6ef71c09fd88bbcb6fd53d42c878df7f2e377e336e30a22a8961373c9a95b0d8", "size": 1958, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy026_result.json|1777383640.1386285|1847": {"hash": "ab6b10cb2218c98e268696eb272bc6ea84e244378e4220611a362f2183ba7664", "size": 1847, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy027_result.json|1777383640.1386285|2017": {"hash": "b61847293f26a7ddfdaa20a8c5c163bbdff2cd02412952ac73ab49901b2692ca", "size": 2017, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy028_result.json|1777383640.1406336|1941": {"hash": "d9b0639c0f8b5f33b5e02d3d97e8cf78c64ef972b8ec87c0e9e7fe9cd39b33eb", "size": 1941, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy029_result.json|1777383640.1426382|1944": {"hash": "6d6170f826daa35ad0b36159c1375f02463b758bce835409e8629700201b5369", "size": 1944, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy030_result.json|1777383640.1426382|1806": {"hash": "f8f4230aa691462cbff8f7f8e456c748d365dcf7926a98a2593b02dc2708a700", "size": 1806, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy031_result.json|1777383640.14465|1915": {"hash": "9ba40a4187283b167ce0c519221714c42488d41aba257bbe0b67fa5c28279321", "size": 1915, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy032_result.json|1777383640.146663|1842": {"hash": "2ed7bd8df00e726778e4487f2ae011dd724349c2fa1106841b2f07de0349109a", "size": 1842, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy033_result.json|1777383640.1486683|1856": {"hash": "ee644c57340b2c420b4ff308fe55c145a605621c94036e89ed937a5873de181f", "size": 1856, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy034_result.json|1777383640.1486683|1968": {"hash": "b5c5e12b1f0e0304bbb46916a94b5b48ecc07ef8f44a416b74edb4e1fa55af27", "size": 1968, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy035_result.json|1777383640.1504195|1470": {"hash": "3fb5096ad5e3b15dd29e1efc641e209f508e49b826a618374c49b98473db2b96", "size": 1470, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy036_result.json|1777383640.1504195|1544": {"hash": "595ae7e8600951392457dc6a0eca96ed7519ed7a70cf93f2c5abfe3e7608aa05", "size": 1544, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy037_result.json|1777383640.1504195|1482": {"hash": "d49ada3c2ced1c75d5b72d6d2abbe2f1867823280b82fd050f0c7a4cac9e66af", "size": 1482, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy038_result.json|1777383640.1504195|1553": {"hash": "b3d6a0b7cd39b44d4665de4b80ed644eed5575e5ae768afcd601960cef1f85f8", "size": 1553, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy039_result.json|1777383640.1504195|1641": {"hash": "f4398f5540073174bd08ab8502ca0a60bdfe7bd9a9a62fc27d9c05ebdd51da3a", "size": 1641, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy040_result.json|1777383640.1504195|1568": {"hash": "7eaa3ebec44050cf611352442c37ceea6dfaa1f1fa2d7ed23a3f5b502f46396d", "size": 1568, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy041_result.json|1777383640.1504195|1759": {"hash": "b64c7e113c70ab9c73c56aaaddbce925e756384582fdb4034915cf7292d9cde8", "size": 1759, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy042_result.json|1777383640.1504195|1499": {"hash": "e76e8df580f114d086d0d1c5bb5c0eb705b742efda54715669bbc1e4361c3f0b", "size": 1499, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy043_result.json|1777383640.1504195|1759": {"hash": "04023af2290fd1fd1990aadbe2e5d47621d833c1b16ba13741af6c8afd31dc99", "size": 1759, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy044_result.json|1777383640.1504195|1623": {"hash": "57560c2df18a0049c8ad54d6981e021240c5f2d6daf5f1a3642d6ebb8d6b7c2e", "size": 1623, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy045_result.json|1777383640.1504195|1599": {"hash": "422d8df42addd53188825c0668f3f569fb6c6d7ca80bdae6a52ee02e5b4abade", "size": 1599, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy046_result.json|1777383640.1664648|1536": {"hash": "38c9d8d4eb7fee6bd9f10c0579f33b49b87a48dad361d3e16b4752ca18ac6ab1", "size": 1536, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy047_result.json|1777383640.1664648|1602": {"hash": "648dd740501604e11af5eaecda368fa729ead3537d8938e15abde2e00c1a9e0c", "size": 1602, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy048_result.json|1777383640.1664648|1580": {"hash": "4a1760697eeecd28e262ef731d1fe0831beaa6aed5aa9fe454eb96d42fe0835a", "size": 1580, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy049_result.json|1777383640.1664648|1429": {"hash": "8773ea88ba45ceedda5c20b59a2bc278eb6321ef416a55eae5ef71d169ef2799", "size": 1429, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy050_result.json|1777383640.1664648|1486": {"hash": "71271aeefeaf33c3debf45d5aad5fd9dea263f1e06b6839579abb07281ce0d1a", "size": 1486, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy051_result.json|1777383640.1664648|1535": {"hash": "edd028f40ce0173f85effda8614de8d3bdd80f6b6005553a2c959072ba24dc73", "size": 1535, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy052_result.json|1777383640.1664648|1534": {"hash": "e851eca11e7e26f7e53606f74c91ce10d35ffe0e98bb661e062d0119ab25b868", "size": 1534, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy053_result.json|1777383640.1664648|1595": {"hash": "9377f611112cb22a985bbfbd05e9be6910c702cbdbee5ad5c5e7bc437a9ad478", "size": 1595, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy054_result.json|1777383640.1664648|1491": {"hash": "d002ea4b91abc8e6db1f9b12ec0ed73e920dbbde0654b91840e1c900f542be7e", "size": 1491, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy055_result.json|1777383640.1664648|2525": {"hash": "e7d2585a261fc4d7b818a574c65f9b48ad266b22c8a2d9286bafe1cd77ac4b46", "size": 2525, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy056_result.json|1777383640.1664648|2540": {"hash": "ddf3a441590c422d5b682c074c84a93ee4f7aa92fbdf1e2cbdb766ea70658fca", "size": 2540, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy057_result.json|1777383640.1664648|2581": {"hash": "7c98673d381406b90c7f90c85f3cb556da12264a886a908c72d8305c8aeaf8e4", "size": 2581, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy058_result.json|1777383640.1825044|2534": {"hash": "6557c5e288e9333283ed293063a8906ea21ab88f2f4dd8d24f63455dbbf44478", "size": 2534, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy059_result.json|1777383640.1825044|2600": {"hash": "19c387bf01629e193c2b5f9abf626d9d6dc70c94466e1e4377ae8feaa31d50ef", "size": 2600, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy060_result.json|1777383640.1825044|2693": {"hash": "7cec902e06465cee26a3955727f6ab80ff9c2059acf4556415816f717b1bfccd", "size": 2693, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy061_result.json|1777383640.1825044|2707": {"hash": "9fb18c4bf944c3b13c62ad2a9b601024ffdb89f2fdb77c3b5af43cd4f9e57e32", "size": 2707, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy062_result.json|1777383640.1825044|1846": {"hash": "eea77f4d4b25477854562e482badd053de8d10768a08bec71483199d38671769", "size": 1846, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy063_result.json|1777383640.1825044|1805": {"hash": "6cfd9642174cd378bbef98a783da753de641747649ab4ccc38a9145ce08882e0", "size": 1805, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy064_result.json|1777383640.1825044|1852": {"hash": "f30f74ea9aed3d7191f3a8163e297b7690670ad0f4f94f0640181940b6aa5faf", "size": 1852, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy065_result.json|1777383640.1825044|1946": {"hash": "489831d60bcb6d459df8799dc9b0e9a16504c89a24426b1ed84295c4796872e7", "size": 1946, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy066_result.json|1777383640.1825044|1754": {"hash": "19ba02a9744dcf996808a33e4ff0b62b6f4ee58ebcda9baec09d0f28d870016d", "size": 1754, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy067_result.json|1777383640.1825044|1927": {"hash": "d5c423384b50ee9af58cedfec3673441d20c2d048f496a27e1b53cf6f015510e", "size": 1927, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy068_result.json|1777383640.1825044|2052": {"hash": "54b25c2bee180cacb664f4e42e00c939d7bff803bf9a068bafe3b6e8d9f596f6", "size": 2052, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy069_result.json|1777383640.198326|1598": {"hash": "cc06d5e636256362a4ced4c5caa8eca9e8cb0ba896b4d143ee70b11a68fa7626", "size": 1598, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy070_result.json|1777383640.198326|1854": {"hash": "382f8aa69f3585c8c11039c9b9b3d50c8a8a19f806953ffedf6b392878ccde2b", "size": 1854, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy071_result.json|1777383640.198326|1896": {"hash": "2e564d660eceaf307291bb4ef2144b5f460112719763f90917c533d5f133cc73", "size": 1896, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy072_result.json|1777383640.198326|1788": {"hash": "01a47ca69dbd5e7641a0039e13490f00acc3077cab10bd222c0797a249cc96be", "size": 1788, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy073_result.json|1777383640.198326|1874": {"hash": "84c76de47b225edb491ee3d91671610b78cb8bfde6dc7e35258090574e66cf54", "size": 1874, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy074_result.json|1777383640.198326|1866": {"hash": "8ad517cc0e68698648507fcd4a7a088459bf14f441d7d99a7170ae384c5ecff3", "size": 1866, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy075_result.json|1777383640.198326|1625": {"hash": "6d261e014bd517909a4a0d604ec6d04b5b9f3c58f6cd9194145e745629fa261f", "size": 1625, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy076_result.json|1777383640.198326|1700": {"hash": "edaa5a9557c12839c9959df43772ff511db29ebcdb7ec6c074a9b90c96ffc648", "size": 1700, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy077_result.json|1777383640.198326|1693": {"hash": "3900b888e2dd80219b0a7c359a5c4a9da5a3501ff2072aa4d3f09fe3324a057e", "size": 1693, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy078_result.json|1777383640.198326|2065": {"hash": "7b38dbef48edd603a68c8794828205319d67450f2380eb5389be19ad84d0d359", "size": 2065, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy079_result.json|1777383640.198326|1815": {"hash": "dd345d61b8da22aa31493f27a646eb2090fbc08f0a234ec4c859a115756b0e68", "size": 1815, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy080_result.json|1777383640.2142148|1909": {"hash": "18e8d9eba04d1a0cc17e943982eb0cf86e804e16223cde6fcf3c371cac21d049", "size": 1909, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy081_result.json|1777383640.2142148|1905": {"hash": "274aea34816d7c1301b347f136b64ad1ab5efaf18f90b6d217370eacaf0c99b8", "size": 1905, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy082_result.json|1777383640.2142148|1910": {"hash": "1780b1bd2aebf1949b5b385d93f8d96a9fc616db4ceb368f6640fba896472b62", "size": 1910, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy083_result.json|1777383640.2142148|1824": {"hash": "25a0ec3a0a6f9982cdcdcbd0e35bbcdd70c2ad24ebd08552b05fe1f31044700b", "size": 1824, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy084_result.json|1777383640.2142148|1866": {"hash": "b2348c6cbc8b2da31c2538ca4b443f01b414cfbf5b8615a78d0802acfe70ac62", "size": 1866, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy085_result.json|1777383640.2142148|1668": {"hash": "df9a9a39fe27bf51c722fbf2318dad1b60e4c6d6dcb14ed29a65d9525012b8ea", "size": 1668, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy086_result.json|1777383640.2142148|1844": {"hash": "0fc9e131887c76e5a8711036e5196c65cbf0db2e36d2b0d5f84b7b2f3eeb3eca", "size": 1844, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy088_result.json|1777383640.2142148|2174": {"hash": "b1b182a89e07f5f04be2a966ccaa8f7e44139658d9357531e683eb707eecbf00", "size": 2174, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy090_result.json|1777383640.2142148|1901": {"hash": "75cf6d702bd3108b4816acd7020cbe69dbec2a66b81e2573d2ef5427d1e3cf4e", "size": 1901, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy091_result.json|1777383640.2142148|1494": {"hash": "3c6fe90bb13acbde42fa12eed2d596fa026c1b032a8fea7cdc52752878ca0b42", "size": 1494, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy092_result.json|1777383640.2142148|1705": {"hash": "76029a2c07c910bcbc67f37fea8279701db244a9e6132000fe3442acb7f49086", "size": 1705, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy093_result.json|1777383640.2142148|1748": {"hash": "6a38ae61205aae9f38cdc6e5d9d5ff29c1a5dcefc9e2f0a78b01dd38de442c6c", "size": 1748, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy094_result.json|1777383640.2299395|1751": {"hash": "b2a7bca73b00fb4a7fd06e75b5b0369621b2cf097474f44dd45bc98a82cdb84e", "size": 1751, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy095_result.json|1777383640.2299395|1463": {"hash": "d1864177e37216a893d3ce3028db268e6bd86e5e98bd8e9d2ccd60e740b70839", "size": 1463, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy096_result.json|1777383640.2299395|1638": {"hash": "b3c82b135a5e31e87a7bff2f5df077efaca57a53dae1163a4eb82d865cb937de", "size": 1638, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy097_result.json|1777383640.2299395|1625": {"hash": "8ab6e83672dab42c24e7f660788fac3833f66475a06b8cd7d73dd84fcd3e9847", "size": 1625, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy098_result.json|1777383640.2299395|1506": {"hash": "d25faef95f4063f55867dff49fef755dc94ebeea88c1eb89ef64f242160ba218", "size": 1506, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy099_result.json|1777383640.2299395|1469": {"hash": "a6095946247f4d9bfd8755d5ac98db763b3dbb866811b59fc1e569b910cf603c", "size": 1469, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy100_result.json|1777383640.2299395|1513": {"hash": "c7bb41d58dc68224cff7fc06c0b30ab7fe58b26bf24f9a569e2eb9de9696b30d", "size": 1513, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy101_result.json|1777383640.2299395|1508": {"hash": "dc41b9063d9c37ff77c388e4cc3423873395d114b4392d6101e542ccaa5ed1ee", "size": 1508, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy102_result.json|1777383640.2299395|1365": {"hash": "c525ce14a1b2092d76e79f0b8ab03e40023c8e8bf9073df8a59180831d0ccb31", "size": 1365, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy103_result.json|1777383640.2299395|1481": {"hash": "8187784ac8f790dc1ae8e538c9269705af8b2fbb56fbd143bbdee5a864ddf116", "size": 1481, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy104_result.json|1777383640.2299395|1257": {"hash": "0464661eeef1ec9f990ad0c71707f8108e284ddafd2c50b2dd5662afb100fece", "size": 1257, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy105_result.json|1777383640.2456493|1505": {"hash": "b92b423ecc2708bcbb9e2520366c1aa6759402d84fb4e86044aa6a53d6056fa6", "size": 1505, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy106_result.json|1777383640.2456493|1769": {"hash": "a8152f562920b91b6d40ea991ce5206a11bddb9eace5350262cee4ebfcc5d32e", "size": 1769, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy107_result.json|1777383640.2456493|1422": {"hash": "a1a60357136b4dd5047747978c69eec9112e215954908d0e5b56a76392437dcf", "size": 1422, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy108_result.json|1777383640.2456493|1341": {"hash": "d9b00e93f2324310c18d4a4274d0969c23725c027d54ffa1ceb503a886aa3c2b", "size": 1341, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy109_result.json|1777383640.2456493|1349": {"hash": "8268f776649f7bed4ee379151767dbea30461b518a39e189f450bb449fe33639", "size": 1349, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy110_result.json|1777383640.2456493|1285": {"hash": "e98d36431e91eb6e1fece19210d1b46b25dce84d102f9f02b1885eed5effa430", "size": 1285, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy111_result.json|1777383640.2456493|1325": {"hash": "1789a7c2fcdaa964b6b44316413115dbad8953b400f7c967b284e487a42bb963", "size": 1325, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy112_result.json|1777383640.2456493|1390": {"hash": "8bf3701e65097acf44694a1ca267cfc2bc596a5b49f8dba6e5aed1c9155b7629", "size": 1390, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy113_result.json|1777383640.2456493|1403": {"hash": "460b0af18de8e3e69805cc67df3daeb8f56c2c1d0b7fd8e2299b0e2064a68da3", "size": 1403, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy114_result.json|1777383640.2456493|1455": {"hash": "84fb0c6c009985910da950d9878d5f36bc09425feadc9eb0b01175033429d991", "size": 1455, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy115_result.json|1777383640.2456493|1294": {"hash": "09851fbc07252f3beb8452847201175714e6115df6e8aea5e7125ad786b95a95", "size": 1294, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy116_result.json|1777383640.2456493|1387": {"hash": "fc818e4e9e102d784285f08a4c5564a9bd986a656fbfa8a3be79270ca0030e83", "size": 1387, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/easy117_result.json|1777383640.2616017|1286": {"hash": "a1dc1cdf995d19e737fbfcbaf94805168f65cc0b978d0dc3f4701f842a3046e2", "size": 1286, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium001_result.json|1777383640.2616017|1993": {"hash": "bf316faf3b1760166c44ba2c68952c6d209f84592afe1b60fec895a1a9ea45b8", "size": 1993, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium002_result.json|1777383640.2616017|1987": {"hash": "3c54bd9145198a7de68404632ecb93e7af674e78be31f11435df2b070d60d155", "size": 1987, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium003_result.json|1777383640.2616017|1949": {"hash": "70825ec20987f552cde82273e54e8530a6c59e593886ee21ac367eb2d698f98f", "size": 1949, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium004_result.json|1777383640.2616017|2065": {"hash": "81a32b35ffbacb2d8d6f04069e11ac8ea17c3359d28af5d95cee38a57e43390b", "size": 2065, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium005_result.json|1777383640.2616017|2252": {"hash": "0b610a27067b9750686ef54bfb724604eadbf8a825d3f6bdbde38d97d7472b1b", "size": 2252, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium006_result.json|1777383640.2616017|2182": {"hash": "e6c90cd90989b43c30bc0da243b0df1f6870e625d659c629a12242211684825f", "size": 2182, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium007_result.json|1777383640.2616017|2422": {"hash": "cfd802b71df365822828eb3a1bc43f2e5375a21ea6c3d5aaae78720b5e6aa82f", "size": 2422, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium008_result.json|1777383640.2616017|2545": {"hash": "b1e5335329591622328f7b1d314f52d84856caafe0dd4963a281cf971c966660", "size": 2545, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium009_result.json|1777383640.2616017|2244": {"hash": "5d29d22053e4be4b50e1e3aba75da04110c0d99646a704b96e12e4ed54ca79b4", "size": 2244, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium010_result.json|1777383640.2616017|2201": {"hash": "b1b7f38d09ebea389f2dfb0cc3a2bfc9652e8ddb04de168970c81636ecb64aba", "size": 2201, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium011_result.json|1777383640.2616017|2322": {"hash": "1cdd5e0ff7a8fd9db697f7e9ed9de2484bee668684a29a749a4a2cdac910ed59", "size": 2322, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium012_result.json|1777383640.27765|2202": {"hash": "73a7b8c7c453ebda794facde55fe9de8c79a6ee699c5bceba0d3483024eddd16", "size": 2202, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium013_result.json|1777383640.27765|2189": {"hash": "604ec541efc81171bd7e8f4b12f9296936599dd001287fef10eb54d5c96daaa5", "size": 2189, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium014_result.json|1777383640.27765|2785": {"hash": "541a8f0759abe719099d4979da9b7fee5eeb63c9fa89e5eedbedf499082b381b", "size": 2785, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium015_result.json|1777383640.27765|2609": {"hash": "1f1e89595f2459d413658c690e6baca948614520000bd28c14138c7e3a2b6a1a", "size": 2609, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium016_result.json|1777383640.27765|2597": {"hash": "a4cf9a4143ad1038bf20dc40274540ec76fb016f3be3e95f776eaca7597e02c0", "size": 2597, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium017_result.json|1777383640.27765|2450": {"hash": "9564a8238ca7481b00763dd0ad9998ac749f357ed2457358bfb55735a3579ce8", "size": 2450, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium018_result.json|1777383640.27765|2578": {"hash": "b1e709b8447c5c7e072ece3c325afaf4e32fcf3c49d5d7825c72a29697144374", "size": 2578, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium019_result.json|1777383640.27765|2520": {"hash": "32e7f57708081b2254e59c39b868e4c4edec47bc82e8166c8ebf5362c2762673", "size": 2520, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium020_result.json|1777383640.27765|2409": {"hash": "a4accdc01bf4862606a1cc76a40377457e5b660dbc01356d2ccee00c6f42309c", "size": 2409, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium021_result.json|1777383640.27765|2327": {"hash": "3a505eb74b9c304fb8ec6a474e0cac426c72b637bb415676d7b3bd91f9c5f293", "size": 2327, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium022_result.json|1777383640.27765|2413": {"hash": "a061731962d1240fe20caf09fba11c352e9a70fee89a3441f57b65ef649d2ed3", "size": 2413, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium023_result.json|1777383640.2933118|2803": {"hash": "97dd2409d37c7940e08e429698c9aed2c73c9d66b4218733d22759187f396ecc", "size": 2803, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium024_result.json|1777383640.2933118|2748": {"hash": "3c80e01d826ed572f9f62767caada7a4f1129975b6f047abe07bc63803d152f4", "size": 2748, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium025_result.json|1777383640.2933118|2680": {"hash": "43ee49cfbadd8cd310bc6258c68c0ba3d937271bc0bed2e832bc32020f1932cd", "size": 2680, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium026_result.json|1777383640.2933118|2403": {"hash": "72c2161086046bf24eb4f1e97b25d0061b05a2235e281b19d9f6103ba83ae6dc", "size": 2403, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium027_result.json|1777383640.2933118|2441": {"hash": "97d119270aa6eb67bfcad26372c711a2f677cc28aa05b41199594f72b6b17dd6", "size": 2441, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium028_result.json|1777383640.2933118|2463": {"hash": "99e4161ccbdfb0a5d0bec5324613e46b70e8128508c6fad3778f69ce814aa93e", "size": 2463, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium029_result.json|1777383640.2933118|2309": {"hash": "7c85e9535345f61b34ff8e27bf84f4d8b6fb25dcd989ad9288b1b58a9f977bb5", "size": 2309, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium030_result.json|1777383640.2933118|2297": {"hash": "818d88ae898e68340d8e5d1fe70d26bfb003d4fecd15e9baa14df7c62e5da947", "size": 2297, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium031_result.json|1777383640.2933118|2444": {"hash": "08967ddbc6f8624b1baee67794c9a457f94d2681cb57a9604db8197d262d257b", "size": 2444, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium032_result.json|1777383640.2933118|2410": {"hash": "170fe6931d7fa13d3e90b3637bc4e40cd46d878150aef324d3c3d3959c5cd4f0", "size": 2410, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium033_result.json|1777383640.2933118|2825": {"hash": "2d690f72aa5daf674b33c8aa8e2a7c62994777177f04cf9d7efd6b721b4a4614", "size": 2825, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium034_result.json|1777383640.3091712|2685": {"hash": "be7413218760ef83bc2551f839f7d51d537c937660d5dfd5dd2f010cf2d0ac93", "size": 2685, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium035_result.json|1777383640.3091712|2692": {"hash": "d7c808da6485db525530b733b5d7fc03dd909813d59f08fa264e2ad183220276", "size": 2692, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium036_result.json|1777383640.3091712|2524": {"hash": "ac31e84c6353a652b4010211a50cecb55a4396ba02800d71be2f9ebb9ebee734", "size": 2524, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium037_result.json|1777383640.3091712|2689": {"hash": "91b1577ea1d38e6a158b1fba4a71718370410aa9e22712fdf3c356c6cf53c71d", "size": 2689, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium038_result.json|1777383640.3091712|2488": {"hash": "80f636e408e701f6847493d0e607c959192b8dd54f21eaf9efaa975b056e5cc7", "size": 2488, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium039_result.json|1777383640.3091712|2461": {"hash": "2b1a6a17df3d452a887b8f604a5497283d6c36d0f8497b137c1f432473817c7c", "size": 2461, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium040_result.json|1777383640.3091712|2478": {"hash": "2ac871f5be0c234b3741807d1fe1ce2fda74f137e59aa3ab893531eeadc7ae4e", "size": 2478, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium041_result.json|1777383640.3091712|2552": {"hash": "0904f72e1fd87f6bf3fdb2503aa4a23a3d3f87500c3275425542454daa788a60", "size": 2552, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium042_result.json|1777383640.3091712|2689": {"hash": "213a683d0f597f708527d646468be92080729c9cb263061a3f849f21afb08744", "size": 2689, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium043_result.json|1777383640.3091712|2487": {"hash": "9f554fced432720d2795af8a5a21dee6f1d63b0f1d5734f819089ca67b7a165b", "size": 2487, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium044_result.json|1777383640.3091712|2579": {"hash": "d380d34c63504368c13b528927c7ab31644ce5205abf3c3178b4880c4d6c97e5", "size": 2579, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium045_result.json|1777383640.3250296|2447": {"hash": "2357e3aa6997ee07a7c286eac2968412e530230719ff6f7074241007035d2175", "size": 2447, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium046_result.json|1777383640.3250296|2483": {"hash": "884cb7d2f11fb587faa7d44ca07a69ed5b427307b99e8ae4af790aa26abdf316", "size": 2483, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium047_result.json|1777383640.3250296|2521": {"hash": "889e182b840b33a7212671a986c5a08aca2aa976072c5b236d9c0b39df8ca46e", "size": 2521, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium048_result.json|1777383640.3250296|2594": {"hash": "4d1e1197c2198f82b23e68e45362181da3d53354c6b4cb731fcc869641262470", "size": 2594, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium049_result.json|1777383640.3250296|2428": {"hash": "123d905aed7e1d4670d9b2df34d972a7ce0c1d14907cc6a72b4e4f5a5ce60d57", "size": 2428, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium050_result.json|1777383640.3250296|2478": {"hash": "0cb60fa28427fa5b73f4229068c9af3bc57273fe9f1d400ba30cad01ee17eb0e", "size": 2478, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium051_result.json|1777383640.3250296|2443": {"hash": "a1239b8d90e31f7216f1d150c592619e8ceaaf8ff5455493364f8401108dc715", "size": 2443, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium052_result.json|1777383640.3250296|2733": {"hash": "b2b6925e1c5d1dc8abe036644ca5ed1f0cf5a3a8fc142b168fdcf28312253068", "size": 2733, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium053_result.json|1777383640.3250296|2616": {"hash": "9ef5be561024285e643000059ea7fc71350181034b87f29238c0292968f3b37c", "size": 2616, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium054_result.json|1777383640.3250296|2672": {"hash": "601abdb34d95e793c94cfb463d9b2e76a7fc40256ef45e54c44a2c9d8c80b6a1", "size": 2672, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium055_result.json|1777383640.3250296|2718": {"hash": "b4611d1215139c25f0c5d0041638c0b8325f859cdf5b972f72bfab14346bd8ce", "size": 2718, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium056_result.json|1777383640.3407829|2524": {"hash": "3bcc2b3afeea0e1af11b7e92421f42fb60e1399381d4e9775e77da63b8d6975e", "size": 2524, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium057_result.json|1777383640.3407829|2535": {"hash": "c3c1f96e010b96a34d119bc90cfaa9ad1f03ec5c99b68d599a771e84ff04e77a", "size": 2535, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium058_result.json|1777383640.3407829|2572": {"hash": "ba29ff281a523c695cb4a4907790ac207aab48606ee8bee6654109a16e7ae66a", "size": 2572, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium059_result.json|1777383640.3407829|2493": {"hash": "b11b6f1e697db0406717ddd7481aa0e0c7fd951d9f1aea4cd89ea33c0d827826", "size": 2493, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium060_result.json|1777383640.3407829|2560": {"hash": "a1922a47c1d758837dfb2539a411051ff08e3ce2dda654b59d4626325d681b72", "size": 2560, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium061_result.json|1777383640.3407829|2275": {"hash": "db9298d672a155a47d506269bc1c7931e84988ca0b796ddbedb863db72ac5410", "size": 2275, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium062_result.json|1777383640.3407829|2085": {"hash": "70e8c9f98bf991866f4a1f9801a350f42a2eefc7a4fb0c9f93e4a97c254653c2", "size": 2085, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium063_result.json|1777383640.3407829|2598": {"hash": "97a6ee544bf266aefbc5516b5db7aa2130bb4fd24d9de17a54d9b25c0c1d1a9c", "size": 2598, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium064_result.json|1777383640.3407829|2345": {"hash": "231d94034b88affe3bdc2dd1814983b035c0b8c1c94f6f7ab259fb30882aa05e", "size": 2345, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium065_result.json|1777383640.3407829|2915": {"hash": "e3f5001a4e61395a24fa9ff85a9658447639cc69a21a0f0717b8bbe02abc22c6", "size": 2915, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium066_result.json|1777383640.3407829|2623": {"hash": "beb6a85d7c5b979afe870ffdf1ad1867632cdba835b202736356e51f0c880f0c", "size": 2623, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium067_result.json|1777383640.356525|2770": {"hash": "bbdb569d8b366a20afd5d340525cbdff15768fadceca2b368878d9b586f98f5e", "size": 2770, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium068_result.json|1777383640.356525|2229": {"hash": "142ff179528d53c751e18a9183803fa7ff2bbd334a9f315091f09ab2b67018b8", "size": 2229, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium069_result.json|1777383640.356525|2622": {"hash": "ee5bfc247bda4bb88c6171ec051c48b403dae93acef125bd4a95bd6d0f1a46a2", "size": 2622, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium070_result.json|1777383640.356525|2384": {"hash": "2fa4fdec8e7454244aab2b9f4f2770ff5178e15d3ddb26da6ab526502d6224f8", "size": 2384, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium071_result.json|1777383640.356525|2692": {"hash": "0cc230fbb3b21eac41d0ab5a7d5d13675250a2473a6f18523941106fa4fd0c1c", "size": 2692, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium072_result.json|1777383640.356525|2691": {"hash": "9a579ec57eb9c682a2ca86fe8fe21ea4fb13ff9b9fc6cdad8d72baded6d7a8c7", "size": 2691, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium073_result.json|1777383640.356525|2116": {"hash": "8bbbf12a04a492ffb94fc626c591cecdfd053d8d394f617e1a16679b7a66256f", "size": 2116, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium074_result.json|1777383640.356525|2160": {"hash": "01edc7289e55177d7dea218f12f7950ed6e6bfb914572886e83348c1bfb4925b", "size": 2160, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium075_result.json|1777383640.356525|2103": {"hash": "9ed7abd6cedd9355e17fac36b579f418b98c28b86085f23e9f8dc983632726ab", "size": 2103, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium076_result.json|1777383640.356525|2305": {"hash": "4064a60aaf9c765d546c2005bf03b88c3f8f968baf2744e705ba1e14bf1fa089", "size": 2305, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium077_result.json|1777383640.356525|2504": {"hash": "816af7a1693ee5c7f58ca0c22cf58cd7732c5cb37bf3044d935915c9664a5410", "size": 2504, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium078_result.json|1777383640.356525|2465": {"hash": "d2131b46c117622ee6d5a1735865364ced0419b05749b82bbfe25f575ae9ab12", "size": 2465, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium079_result.json|1777383640.3723128|2175": {"hash": "63b92dd17724611d918b22eb3aabd185c745ebe897f596e8146928fdd6ff79d4", "size": 2175, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium080_result.json|1777383640.3723128|2147": {"hash": "88aff46b1e33177d3d66327e7848a4e20caf37bafa638385816e3a71f1f47c91", "size": 2147, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium081_result.json|1777383640.3723128|1999": {"hash": "f0b7d8e161177be5ae4923a0263ac3affaecf6992edaaad9a17c949789b8ad73", "size": 1999, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium082_result.json|1777383640.3723128|2210": {"hash": "48c08d8486396ab5f74ba740e1a8693bee430d98ea7391242fb50c272f95090c", "size": 2210, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium083_result.json|1777383640.3723128|2113": {"hash": "5073da4bf5015f019d4c6501bb96f787b56dc3a4a461ad5dcb75a9fd268fbc84", "size": 2113, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium084_result.json|1777383640.3723128|2538": {"hash": "25da99044b318802c8e95c224ef0736705885ad553cc14f0ad1f34fd1efb57b6", "size": 2538, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium085_result.json|1777383640.3723128|2557": {"hash": "e06e1b7e5713b3eab33fe686fb3fd2354c2d0305568af8065b288348d05526bf", "size": 2557, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium086_result.json|1777383640.3723128|2750": {"hash": "f59771ab50c17875014325ba6fde2c0cf7871462d3a78a3b71d7662aeaa97937", "size": 2750, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium087_result.json|1777383640.3723128|2176": {"hash": "3213d1a2574170897aa9a8380020826688caa4c7b67116339037c4247f9cd532", "size": 2176, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium088_result.json|1777383640.3723128|2333": {"hash": "2b5fa431f404adca4c12b92c34d9cdbfff65fe069548a4d43fc3047257f8e8e4", "size": 2333, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium089_result.json|1777383640.3723128|2151": {"hash": "64acf0c41acb3622a14bed28bd3f67302389fc12f82249d597246b447d8446c1", "size": 2151, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium090_result.json|1777383640.3879454|2402": {"hash": "41f0f5ee2cd6adf473cfc8a072ee9c4b71c892604c895c56bb854d6c891b8969", "size": 2402, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium091_result.json|1777383640.3879454|2068": {"hash": "e68e79eb6c752439ea14f9840061ec618233d2cceb0d9d8b0c4f4109f3340410", "size": 2068, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium092_result.json|1777383640.3879454|2079": {"hash": "f0e48f268d5034b532daf0e1b66b945adad9be59c84e16c37e7027a83c01c0fd", "size": 2079, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium093_result.json|1777383640.3879454|2062": {"hash": "26ed6d2a4214efe463f96d3a4179bcdf80f390bd63160fb0c57ad057ef9f1941", "size": 2062, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium094_result.json|1777383640.3879454|2091": {"hash": "c6eceb24305066a66fbe90a5438daadfc1c1b9aaacf01956c94ea139384a00f4", "size": 2091, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium095_result.json|1777383640.3879454|2202": {"hash": "89aa64acda12c077453ad29e06cc9ad57747d61b82d92a477b00a98e4d408ca9", "size": 2202, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium096_result.json|1777383640.3879454|2615": {"hash": "dea3c4e3eb7d35251d7d4c03dcd70d6e3598acc391852426f8788e0cceaeab94", "size": 2615, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium097_result.json|1777383640.3879454|2667": {"hash": "6f7b1f15cf17192a4ce59b6428289ae5d33a3968e2ed660194c99693b061a915", "size": 2667, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium098_result.json|1777383640.3879454|2058": {"hash": "f1b1084b2a8e86230df61888926aaa8d97c6dd67563670316d91840be7dfd26b", "size": 2058, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium099_result.json|1777383640.3879454|1848": {"hash": "adfaf65cc9ee82c2c5de64125dbb41162dbb9eaaeb35e5b88315597195932e69", "size": 1848, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium100_result.json|1777383640.403621|2001": {"hash": "ce8a6a5371cc3c2fd8d2763e2c70a3d98f4eed6e0582877b0abd84c021937a8a", "size": 2001, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium101_result.json|1777383640.403621|2136": {"hash": "094ed8d4a33a987ecc2a71fd4725997a3b4a9fc6447564c5ef80cd1778610cc9", "size": 2136, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium102_result.json|1777383640.403621|2616": {"hash": "f42a8d2ad62b78f79ba44992f6646ce362a41e4013406e0ef7f352193069c66b", "size": 2616, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium103_result.json|1777383640.403621|2564": {"hash": "e0c022da3c6a1229842325647fdb15066c8342c69acd6cdf2981cb5337a42f79", "size": 2564, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium104_result.json|1777383640.403621|2474": {"hash": "6129d2b34762cdb5abf52b1e9aca72ec63468156ad326e03daba8ef4e4962539", "size": 2474, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium105_result.json|1777383640.403621|2382": {"hash": "36475ed92bf3c4bcfcde2d831238d136cd3ecfb30ce484e38b737ee749461b8f", "size": 2382, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium106_result.json|1777383640.403621|2278": {"hash": "b2b13cd297c26ab97b2ac59428699febb5e4e348dbe24d7835d3bdd0c7c50508", "size": 2278, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium107_result.json|1777383640.403621|2237": {"hash": "20cdc63f3a07548d6f65a375eaa1130ef8f9d584105d56dd6f2dba45068ebfb8", "size": 2237, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium108_result.json|1777383640.403621|2783": {"hash": "e6ca88a08d19ca38f2d87ef777e95de79f516bfec50d2390277ca02f4a926288", "size": 2783, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium109_result.json|1777383640.403621|2958": {"hash": "5d5fdc0aadf4b072eb57cbec9806bd3c656fb1493c6bc744bd577dafc47be64d", "size": 2958, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium110_result.json|1777383640.403621|2345": {"hash": "33fd4b7fac9d47b0ad688eb700d3df6622d2beea0ec9ca36f301bb3151875d92", "size": 2345, "status": "c"}, "assets/qa_raw/enterprise_industry_analysis/medium111_result.json|1777383640.403621|2736": {"hash": "f16676ccbcda7098ac6a3c3818edd97ea306b137d4ef271e824a4f14316dbf01", "size": 2736, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/easy067_result.json|1777383640.4192727|1479": {"hash": "4f3e9039e3ae73dded41ad7d13e6c652a21a071b860d70c68c633c7b98c0829d", "size": 1479, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/easy068_result.json|1777383640.4192727|1371": {"hash": "b4419ac2ea1d7ca9ca57f7b10a3477251c9996334aafd4fa5e4b6aeda70a9755", "size": 1371, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/easy069_result.json|1777383640.4192727|1765": {"hash": "21388e28e6a92052f12c5977d28f2ccdb2c809eda0b433e74fe9f4c33e1ba48d", "size": 1765, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/easy070_result.json|1777383640.4192727|1491": {"hash": "41e531014c6dba08a917669542fc9cba6baa55e71ac6d4e714e5a5307a06ba2c", "size": 1491, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/easy071_result.json|1777383640.4192727|1572": {"hash": "d96b45a26623e9fd0b4b39e4155eab9462d858e2aecedf3bc9bb589b80520338", "size": 1572, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/easy072_result.json|1777383640.4192727|1530": {"hash": "6c933a908ec862d8c19e0940561bc60c4cdd9e267fc48ba73715c19c2aacf898", "size": 1530, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/easy073_result.json|1777383640.4192727|1605": {"hash": "b203aef9be851a4c3bdd58f293a898b19119671fa27f03223c6ef2ff6e8602e6", "size": 1605, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/easy074_result.json|1777383640.4192727|1788": {"hash": "25b89f7ee73de362a6d3bdf287d6356fd85797f0cfc19b3546ec21d5bc588c7d", "size": 1788, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/easy075_result.json|1777383640.4192727|2091": {"hash": "1fbc06f1e47318d1ef6a2fa41dedf1b380c750a32aeca3e8dae9cc1058ff140d", "size": 2091, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/easy076_result.json|1777383640.4192727|1653": {"hash": "da9515d50bbd7542948f044d05408ba46cedcdfe71036da402f5c9a27d7f08f8", "size": 1653, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium001_result.json|1777383640.4192727|3016": {"hash": "12b67ddea398a18df8d9195d555f2e62c0385e40450a15eae699106ad13a9f69", "size": 3016, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium002_result.json|1777383640.4192727|3007": {"hash": "a0f9771e06e9dd8402de62faf1e2e095dbabec73bb42147021cb6bc0ac7b3e73", "size": 3007, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium003_result.json|1777383640.435121|2967": {"hash": "977f83522e4c253e3bedca0c1950d83b24c276557d761629dcd8860d82a72c0e", "size": 2967, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium004_result.json|1777383640.435121|3025": {"hash": "05bc66f69723c9303520b0c53ce4a7c4bdf6ea8883fe65cdf26d62c1ae39063a", "size": 3025, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium005_result.json|1777383640.435121|3002": {"hash": "d57fcdf99722629b3684f1c72d722d77d867b524b4499be8bfc3106ce5ca07d8", "size": 3002, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium006_result.json|1777383640.435121|2926": {"hash": "dd4fc116258da3336e63991ab5ece389aeedc19562f63f8a29ea40a5ec1ad045", "size": 2926, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium007_result.json|1777383640.435121|2783": {"hash": "bc876064514946b0d20e3879ba40ed62ce07d0a5eebed31d6a819978dcd8e907", "size": 2783, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium008_result.json|1777383640.435121|2566": {"hash": "c6d1b4d07ce626c528d5bf82f74d6a618d931bfdbd95828c363035d328064aa7", "size": 2566, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium009_result.json|1777383640.435121|3438": {"hash": "d04e618886d22d2d44be19496efbffc88a2a4e7718741b7aa5892a47abd9f210", "size": 3438, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium010_result.json|1777383640.435121|2860": {"hash": "cb79a286441ffe6e9babd6fcbdb2c4e58ed3da9ea908c5e8fd68c6b0f141c03a", "size": 2860, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium011_result.json|1777383640.435121|3109": {"hash": "daff29cdeb25db14524ec1413cef695baf8f25bd4ea27ecc27abd2705422340c", "size": 3109, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium012_result.json|1777383640.435121|2938": {"hash": "d3df168f3e77ff16f49a29cf64b05b1694ae34b3a9e151c306205f44c3900854", "size": 2938, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium013_result.json|1777383640.435121|3209": {"hash": "cc4676bba9586017d873c04d83e731dc134ed81fcabf120653d757038cc6aa1f", "size": 3209, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium014_result.json|1777383640.4510615|3118": {"hash": "1e0456e1ccbe88caa566ed3afcc5b33dad73041a9204abebb5c92d99f9ff001a", "size": 3118, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium015_result.json|1777383640.4510615|2782": {"hash": "6662447401f5335ec46f59be7efbe56de2a5232bff6c5856c381fe2c9a0f3ef6", "size": 2782, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium016_result.json|1777383640.4510615|2649": {"hash": "265e998d83afcb2c851154dd4447c3d14191d9983a9b377e88b14ff2a9765842", "size": 2649, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium017_result.json|1777383640.4510615|3025": {"hash": "8b3f5f4cf2186e37b8d99f923a0bd1a7109f80bd902520cf8a6baab9b87ff975", "size": 3025, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium018_result.json|1777383640.4510615|3125": {"hash": "a7016b36f93741babbfff94cb735a022de49a00e820b73a3b791a3556e8cecff", "size": 3125, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium019_result.json|1777383640.4510615|2793": {"hash": "f64acb9e67cd2772c1900b82c03fdb9b3a4dec384c9de8e7a679dfb06e8c9197", "size": 2793, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium020_result.json|1777383640.4510615|2730": {"hash": "d503a118a490ce906d82d818073fe53cdfc7cb2727ee2393d136ed10b23ad855", "size": 2730, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium021_result.json|1777383640.4510615|3037": {"hash": "8fd6a0b53eade95bf56356bb9a4edaf99c1c5111fa5db447ce7d2fb784a8f80f", "size": 3037, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium022_result.json|1777383640.4510615|2895": {"hash": "e30ed516e23c6ce72685c55ba85eb01a7c7a9cea515dc899d430e51487c6133a", "size": 2895, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium023_result.json|1777383640.4510615|2824": {"hash": "12c88d704d2338ac786cb11d0fc0a6f7d1d09b712efe84758a5b63e07469a770", "size": 2824, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium024_result.json|1777383640.4510615|2934": {"hash": "c604894c4ae7f8d95d0f2ec958a98c584c10b3c18b33180016a3fb6c9ea8a1a8", "size": 2934, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium025_result.json|1777383640.466971|2794": {"hash": "e9a35ec504d37b4b50c6bd09e6c03a8e9c352e0a2520065208d3368319ec2a20", "size": 2794, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium026_result.json|1777383640.466971|3037": {"hash": "a05c1782cdd3a7621e665f0a8f1dda663a52a63af1f541e9c21491a2ed4c9e25", "size": 3037, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium027_result.json|1777383640.466971|3151": {"hash": "f66cada32427089fbdf2a09841ffce335d5c669805668315a8e537414caaa919", "size": 3151, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium028_result.json|1777383640.466971|2955": {"hash": "04277bdb099030da2c1ed83bd37b6bc8599a9ade491827096dc2654feea8b7c7", "size": 2955, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium029_result.json|1777383640.466971|2586": {"hash": "38173fd2e557e0394b5344fc36e9ad536b7781be6a29753b6b7cfc97727de3c9", "size": 2586, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium030_result.json|1777383640.466971|2666": {"hash": "473bbd737795418e351e04c3b3025ce7f1b9dad2e1786afaa9e060766582288e", "size": 2666, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium031_result.json|1777383640.466971|2623": {"hash": "dc9aaf9b8771684b234a464106080f3354965d01b019143f656d4d62f03d54c3", "size": 2623, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium032_result.json|1777383640.466971|2520": {"hash": "79322ffd3d21a7a5b5267230433ff1ebc537a5681e1cb3df1863bfd6bffdf173", "size": 2520, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium033_result.json|1777383640.466971|2596": {"hash": "373c6afa2b0f4db44b53c19eb61779112ecb614bda7f62d27ee3aa9427d76de6", "size": 2596, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium034_result.json|1777383640.466971|2791": {"hash": "54724a036a6af765ecfb733d3744a4e034c042a668b1d34058b355523f26fb82", "size": 2791, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium035_result.json|1777383640.466971|2672": {"hash": "27f78639ea76876079d346237bc3e5c929684b88d7915b0ae5376e391733cc24", "size": 2672, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium036_result.json|1777383640.4828696|2731": {"hash": "cdf44ae315f85c1cea11ebf82538a46385cabdd8f63a04f5ede62ef8232fc616", "size": 2731, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium037_result.json|1777383640.4828696|3112": {"hash": "9ceebecc5a9e28b3b229a486df012095013bd4b62b048dafbe867f6cc31967ae", "size": 3112, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium038_result.json|1777383640.4828696|2732": {"hash": "e3a5a0726135a68920ac8644e45a092d6f46335f2083259329221a2c40005fb8", "size": 2732, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium039_result.json|1777383640.4828696|3036": {"hash": "5daf5b908cea22667024bf24494f4662f3451b852277e6d9b98517c71a66dcf9", "size": 3036, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium040_result.json|1777383640.4828696|2853": {"hash": "0d124c3e3111ae55578d6b8c1a1c8522812fe0715ba117be91cf732910f80d4f", "size": 2853, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium041_result.json|1777383640.4828696|2488": {"hash": "9d39536c0d6941142199d3633df71630a26779cb56547b5ff53ce4e8688ae91f", "size": 2488, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium042_result.json|1777383640.4828696|2529": {"hash": "36ffbbd962ba7afb5577421c8fadd4867f33cf58c368a1c9baa01dcf29cd4897", "size": 2529, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium043_result.json|1777383640.4828696|2421": {"hash": "2f10bceca563f4ea0ac4428b2c1803955f60a4117a06dd555fe0bc2041ad38c8", "size": 2421, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium044_result.json|1777383640.4828696|2754": {"hash": "182bae18a024bd9978d509c731da17099b860f9648ea41746905886bd171ee5f", "size": 2754, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium045_result.json|1777383640.4828696|2249": {"hash": "8394d932bb4f76bb71414d13d373c76b14bfadcae22004e3894d08a17a9f0969", "size": 2249, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium046_result.json|1777383640.4828696|2235": {"hash": "af59f13f678959ef57dca39d3a62dbeff7e0fc575a5f1f283cdc8f6f3aa97fbb", "size": 2235, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium047_result.json|1777383640.498546|2559": {"hash": "06fe1575b1c5c8be93a7d247b3c96df4413a2a0fe974853799135f1c7c9a5fc9", "size": 2559, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium048_result.json|1777383640.498546|2241": {"hash": "c97fb1aec5c5c81c43b4c669647daffccd512eb98842ab62656580e1559bbfb9", "size": 2241, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium049_result.json|1777383640.498546|2374": {"hash": "2f34edc8287674aa1fdb01542571f2513155695c46533a3fbb6f97416bb23376", "size": 2374, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium050_result.json|1777383640.498546|1993": {"hash": "ce451a29156b345595e259ffdf2b245ad3b2cb241edbf050ddafd687130438ad", "size": 1993, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium051_result.json|1777383640.498546|1823": {"hash": "3a6eec2149077b075f1531055bdfd4f0ab49fbefb2cc46c491b916ba8d25ec81", "size": 1823, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium052_result.json|1777383640.498546|2202": {"hash": "518c196d53072fd066c2c0a0af86fd3479802b818fd15efb8a77f6289f02c32e", "size": 2202, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium053_result.json|1777383640.498546|2665": {"hash": "3bbb6255df09d816e7ca58cafb553f223ee4ec054b287d925c16a04e612eda41", "size": 2665, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium054_result.json|1777383640.498546|1989": {"hash": "381d214043401d1861f582c8bca65412197962218306bc2392c359dd10317dd2", "size": 1989, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium055_result.json|1777383640.498546|2316": {"hash": "02fb3ea64bd46eda8d3d5797a739173988e46da817a5dffe1fc335abfef728a0", "size": 2316, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium056_result.json|1777383640.498546|2175": {"hash": "965ae4c3c740ba16606726134607fc9e3411c024ce23098624f3e6591aca6e1b", "size": 2175, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium057_result.json|1777383640.498546|1963": {"hash": "faa998fea00fc89c8f92df7c67d6750872aae51d89d2de5afb5d1fc5fe2c9dce", "size": 1963, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium058_result.json|1777383640.498546|2173": {"hash": "9043d1d673c0fe5a5c10e5ad7f2cc42764c63e57f5663d71bea0fe4c6a05dad1", "size": 2173, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium059_result.json|1777383640.5142512|2609": {"hash": "06124200464bd0f79f1758f7efeafc50339d55799a81a15173402751566760f4", "size": 2609, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium060_result.json|1777383640.5142512|2015": {"hash": "78c2bd6dde7f4556fbabc5d9e0981337147fd728692bd685de4430cd47a921d4", "size": 2015, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium061_result.json|1777383640.5142512|2359": {"hash": "a8469ca9edc0c4d0047f8240437a68c23b459f777f6de6e9a047148829f6284b", "size": 2359, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium062_result.json|1777383640.5142512|2764": {"hash": "32df4d368d6bd529557b2cb75a6150209e7fed52d11bd58edc0d91de1c2b34b4", "size": 2764, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium063_result.json|1777383640.5142512|2091": {"hash": "8188e992b5981655ba37ff405c8febe1bb7b2ccb8d2ef9d81a37214d914e0115", "size": 2091, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium064_result.json|1777383640.5142512|2620": {"hash": "6274582e9c7b930bf433b443b8b2af8b9dd317dfe5fb4fa34594fedb6fee43b4", "size": 2620, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium065_result.json|1777383640.5142512|2128": {"hash": "913c34deda9212f5207e50b35aaed4cedaf0b039ae3f636b668f28dc8be48406", "size": 2128, "status": "c"}, "assets/qa_raw/enterprise_industry_policy_analysis/medium066_result.json|1777383640.5142512|2190": {"hash": "ba8007219eec0875453b066af57fd267f90faaab428c37665a860d1ad8814120", "size": 2190, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard001_result.json|1777383640.5142512|3640": {"hash": "53d5f1269c74ca17858ee94139cd69c5b503c2e318c2d412ee086cc6236f76ab", "size": 3640, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard002_result.json|1777383640.5142512|4609": {"hash": "62ec378b7ea9128a7a5bfa2828061986efdbe515c10e14d23f2e0fec90dfb0c4", "size": 4609, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard003_result.json|1777383640.5142512|4172": {"hash": "17aec99a5521478374b0a0d0cd90c3d2fec0e9f5a82c79e208b46f5260229cef", "size": 4172, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard004_result.json|1777383640.5298924|4234": {"hash": "9df70463921b6c1371d7d1d77e0c24cabdc29b1316b6ae4dc1b2e64f980441a4", "size": 4234, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard005_result.json|1777383640.5298924|4005": {"hash": "b99e1ba4f6a1061c58f3a096be91fe3bbe69bf88832da5cd8ad882c46c84a11b", "size": 4005, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard006_result.json|1777383640.5298924|3640": {"hash": "ec7bf583c873a8353a7f7849b237347b27cff5d56727a0d02b9b1da2fc3f4ff3", "size": 3640, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard007_result.json|1777383640.5298924|4156": {"hash": "d4ea3421478cd09386f587da060da80c98db112b171941de68f4f77c1b453ecd", "size": 4156, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard008_result.json|1777383640.5298924|4900": {"hash": "d805427809f14d37d32b7b3f8192a886e47559878990ccb6cebb65ac740cf9e4", "size": 4900, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard009_result.json|1777383640.5298924|3419": {"hash": "09ae42e042703629c59287faa15932808b11901aae024e4679590a80cf7a97aa", "size": 3419, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard010_result.json|1777383640.5298924|4780": {"hash": "3ffca9d6df95e819dd5dbd769c1087784026c178338eece225d6c8633dd82f9b", "size": 4780, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard011_result.json|1777383640.5298924|3999": {"hash": "fccc3287ca97c0713f6d9fb30f4a73bd81b1c4d7a079ff887b4a704ad97519a3", "size": 3999, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard012_result.json|1777383640.5298924|5734": {"hash": "1ec0e7e8912a5b8dd390038d4463711e2d2447327e5b60c03dd5a4de24c59079", "size": 5734, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard013_result.json|1777383640.5298924|6141": {"hash": "4ece8331c7affff7f3ce1561b375f52b6802577668e197dad577d2537500c00c", "size": 6141, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard014_result.json|1777383640.5298924|5218": {"hash": "090de85347083b6b60d1b0a1a602579393be0ff1b1cf679bff3ef23b37247900", "size": 5218, "status": "c"}, "assets/qa_raw/hypothesis_verification/hard015_result.json|1777383640.5298924|4945": {"hash": "bb97ac9be5787a9c4b00be16c7acc6ea4fe1acb87001c6f05ae563623bb8405a", "size": 4945, "status": "c"}, "assets/qa_raw/hypothesis_verification/medium001_result.json|1777383640.5298924|3324": {"hash": "7fabf0fd0524be938815bce1c2ea55ed797cd1c598e353ecf7b883a610948606", "size": 3324, "status": "c"}, "assets/qa_raw/hypothesis_verification/medium002_result.json|1777383640.5460262|2603": {"hash": "d09264ddeb9c48d47f33e9c8fb398676ca9fbd1cba647640de66441da5ef68e3", "size": 2603, "status": "c"}, "assets/qa_raw/hypothesis_verification/medium003_result.json|1777383640.5460262|3438": {"hash": "3831abd6aeb5151de13334d64ede289518f8a8d775970d5486c320c5ea180710", "size": 3438, "status": "c"}, "assets/qa_raw/hypothesis_verification/medium004_result.json|1777383640.5460262|3181": {"hash": "2f1d1ebec245ffbea9e2d025e9e7bca54ef343b5952554b9540645735524b23e", "size": 3181, "status": "c"}, "assets/qa_raw/hypothesis_verification/medium005_result.json|1777383640.5460262|3359": {"hash": "3feeab1617adce1f880ed9c1e25a111774564f820218a90fe463ee13a94257e9", "size": 3359, "status": "c"}, "assets/qa_raw/hypothesis_verification/medium006_result.json|1777383640.5460262|2367": {"hash": "543abe5c62e63ed3ef1551ea611f84e73837fab2172f09331e6e465487ca834f", "size": 2367, "status": "c"}, "assets/qa_raw/hypothesis_verification/medium007_result.json|1777383640.5460262|4562": {"hash": "db4838eccd661391a1ad739c4e62efea285ecfb111c1dfb01574fa1f72fca034", "size": 4562, "status": "c"}, "assets/qa_raw/hypothesis_verification/medium008_result.json|1777383640.5460262|3459": {"hash": "c77e895596042760bb775b6898b58cf04edf9122d3816c997cebdeda370933f4", "size": 3459, "status": "c"}, "assets/qa_raw/hypothesis_verification/medium009_result.json|1777383640.5460262|3293": {"hash": "f5cd1e53591ae578e59d3848ca51d8bf1dfcaa7cc907f95a00df799a87c0ce87", "size": 3293, "status": "c"}, "assets/qa_raw/hypothesis_verification/medium010_result.json|1777383640.5460262|3935": {"hash": "0d2ff728ec4c04f54309d2146caf6e05eff202de41fc3e747921b37021003d10", "size": 3935, "status": "c"}, "assets/qa_raw/hypothesis_verification/medium011_result.json|1777383640.5460262|2769": {"hash": "503310c25e42cc876576bbf45b3735a8de92e1f4dd9b2af7edb691c1e76a7eaf", "size": 2769, "status": "c"}, "assets/qa_raw/hypothesis_verification/medium012_result.json|1777383640.5460262|2847": {"hash": "cfc86520f1aa1e4150d44862a926781c6c3fb2f6920d5fe67db64b459805a040", "size": 2847, "status": "c"}, "assets/qa_raw/hypothesis_verification/medium013_result.json|1777383640.561872|3193": {"hash": "d71717b97b34e78260b1adaa32fb1ff1e2aeb788ba861c49120fdfbdeca127c0", "size": 3193, "status": "c"}, "assets/qa_raw/hypothesis_verification/medium014_result.json|1777383640.561872|3897": {"hash": "0d25b35e3890cb745a6718d192f820f39e0ddd7ce90b2a8779e820894a6c2e9c", "size": 3897, "status": "c"}, "assets/qa_raw/industry_planning/hard001_result.json|1777383640.561872|4385": {"hash": "c8f9fe557437eb216845bf9b6f30013d2ceb5d99dbb44439e4bbb3932e14a054", "size": 4385, "status": "c"}, "assets/qa_raw/industry_planning/hard002_result.json|1777383640.561872|4799": {"hash": "2dffce5d42fcaa0aaf41365efdf162ddde928db1413955104b580736aecff61a", "size": 4799, "status": "c"}, "assets/qa_raw/industry_planning/hard003_result.json|1777383640.561872|5522": {"hash": "6b365c7dfc38a45f7392382038197502a4fa8cedde20aff59b6174103c3d2cfe", "size": 5522, "status": "c"}, "assets/qa_raw/industry_planning/hard004_result.json|1777383640.561872|4587": {"hash": "4a3d32bdb3124489033ad73189a00e61d000711e4291527b6dbc8932726ebb48", "size": 4587, "status": "c"}, "assets/qa_raw/industry_planning/hard005_result.json|1777383640.561872|3514": {"hash": "36f3189ca74715883f7fa72cda9a767a2cac0f6b3a4cdb34a9f0ae740f5e6782", "size": 3514, "status": "c"}, "assets/qa_raw/industry_planning/hard006_result.json|1777383640.561872|4330": {"hash": "d877ac364f100224009fe681c8eafc001aa700b69e5b55c29ecf1b83c5968ffb", "size": 4330, "status": "c"}, "assets/qa_raw/industry_planning/hard007_result.json|1777383640.561872|5267": {"hash": "82e17dc82fba1144aeffbad5cd29bc715f35156ca5a3eb896aa67e17cdde46c9", "size": 5267, "status": "c"}, "assets/qa_raw/industry_planning/hard008_result.json|1777383640.561872|4929": {"hash": "0d1a37c9488b9c93e8c763a5b056a354576368c3758c2c6043629bfb0ed768cd", "size": 4929, "status": "c"}, "assets/qa_raw/industry_planning/hard009_result.json|1777383640.561872|5234": {"hash": "ca2b4b82a846252ae0ef563f5e287535d5b3ec084032dc4c5b971fc5726483fc", "size": 5234, "status": "c"}, "assets/qa_raw/industry_planning/hard010_result.json|1777383640.561872|4936": {"hash": "785ccf3d24e1af1880de1e2857e6d0a654a92749b56d643b87f0e5dbe716b1cd", "size": 4936, "status": "c"}, "assets/qa_raw/industry_planning/hard011_result.json|1777383640.5776587|4921": {"hash": "deaf7278fd00c8a3c700a24e98013198db53524a696dced3280b96aadba83d38", "size": 4921, "status": "c"}, "assets/qa_raw/industry_planning/hard012_result.json|1777383640.5776587|4824": {"hash": "d2f4dad40a5f51461bae127fdc520a096ded86bb9b10491ce5854eb605e0ba0a", "size": 4824, "status": "c"}, "assets/qa_raw/industry_planning/hard013_result.json|1777383640.5776587|4430": {"hash": "69e4043019500eb1fd01f6d773d8e69059fc6d3adf74292ffe9c2f1558073ca7", "size": 4430, "status": "c"}, "assets/qa_raw/industry_planning/hard014_result.json|1777383640.5776587|4016": {"hash": "8c2bf368a095ab2fbac1cc440678d688466b79d40977550fc10d5200f71ccc31", "size": 4016, "status": "c"}, "assets/qa_raw/industry_planning/medium001_result.json|1777383640.5776587|3535": {"hash": "bc40bd75815e378ab1c6cdc79f80fb2bb144f5a6052e30a84465045c68f8eb0b", "size": 3535, "status": "c"}, "assets/qa_raw/industry_planning/medium002_result.json|1777383640.5776587|3484": {"hash": "ffc1a9ea10ce86d65c67f7c8653fc48e0cb494fbc6772bdddfcdc458551fb4d3", "size": 3484, "status": "c"}, "assets/qa_raw/industry_planning/medium003_result.json|1777383640.5776587|4036": {"hash": "5d1d5790099b7c264bf834420a0574c74f1175ac3b1af44942b4cd662393e554", "size": 4036, "status": "c"}, "assets/qa_raw/industry_planning/medium004_result.json|1777383640.5776587|3055": {"hash": "631c491a04e152c0ab789bc721261fd8bda975f107cfbbb5fb1625a997f75ad4", "size": 3055, "status": "c"}, "assets/qa_raw/industry_planning/medium005_result.json|1777383640.5776587|4405": {"hash": "6220c3857316097d719521a168cd371ad66dc30913826443506ed427d0df46c0", "size": 4405, "status": "c"}, "assets/qa_raw/industry_planning/medium006_result.json|1777383640.5776587|4780": {"hash": "9d0e9c87e9c06ec19ddddcad5a5a5af5a5c602cd9ccee1d8fa7c134d455ed66b", "size": 4780, "status": "c"}, "assets/qa_raw/industry_planning/medium007_result.json|1777383640.5776587|5418": {"hash": "483105999334cc9edd858313509be773fd7f59cddb1565932085098d05c9a739", "size": 5418, "status": "c"}, "assets/qa_raw/industry_planning/medium008_result.json|1777383640.5776587|3907": {"hash": "bc843031e65b1b7eba915baced770ad1b438ddb4c2ead29df1706cb79f28ae04", "size": 3907, "status": "c"}, "assets/qa_raw/industry_planning/medium009_result.json|1777383640.5933943|5024": {"hash": "fd25bcd586d487fceae2e22eab426dabed409b1e478100acaacefe35ffd5f8db", "size": 5024, "status": "c"}, "assets/qa_raw/industry_planning/medium010_result.json|1777383640.5933943|3838": {"hash": "a8bb1cd46634772ece29a5591b48c868f5ac98f4d1d7205151c1a89ecbcb955e", "size": 3838, "status": "c"}, "assets/qa_raw/industry_planning/medium011_result.json|1777383640.5933943|4292": {"hash": "977f9b2a29b209f665da21d340f8e4ddc472a3d724f50bcbe013901cf35bf0c8", "size": 4292, "status": "c"}, "assets/qa_raw/industry_planning/medium012_result.json|1777383640.5933943|3412": {"hash": "b5cf2420624428d2c5a7071c6753d5cc002afcc456998af9a1b03faba1f48845", "size": 3412, "status": "c"}, "assets/qa_raw/industry_planning/medium013_result.json|1777383640.5933943|5355": {"hash": "e8badeb4827f6825d27d18e5150e0bff6ca778226da4261c16c648b1847fffc0", "size": 5355, "status": "c"}, "assets/qa_raw/industry_planning/medium014_result.json|1777383640.5933943|4362": {"hash": "d985430dc076409f10be11ae29686371d892ddc217f6132d7b6bfee8bca5d546", "size": 4362, "status": "c"}, "assets/qa_raw/international_comparison/hard001_result.json|1777383640.5933943|3561": {"hash": "6d15a79491cabeeb521476036d28e2aae43899f5baee8004d142d70f03aa4807", "size": 3561, "status": "c"}, "assets/qa_raw/international_comparison/hard002_result.json|1777383640.5933943|3885": {"hash": "23decde3e97d8bf1ca3a48bf31a9a1ed132306bc404b417f87d2459459396cef", "size": 3885, "status": "c"}, "assets/qa_raw/international_comparison/hard003_result.json|1777383640.5933943|3430": {"hash": "df5f8101d65e7c09fcd6a3d923bbf02e042439433225d142a6e818dc259425e6", "size": 3430, "status": "c"}, "assets/qa_raw/international_comparison/hard004_result.json|1777383640.5933943|3564": {"hash": "cb87f6ad06917cea1e319a4ac6713855e5e69951f5e73c17e9cfc40b6a3596c5", "size": 3564, "status": "c"}, "assets/qa_raw/international_comparison/hard005_result.json|1777383640.5933943|3776": {"hash": "4393e297c3370af23bccb2c9f772a0429d466e545939f9274addf764caa788ac", "size": 3776, "status": "c"}, "assets/qa_raw/international_comparison/hard006_result.json|1777383640.6095273|3794": {"hash": "c72855563014c1cd1401d045889dd87a6d501f3d687dd0cb9773fd8de3656d6d", "size": 3794, "status": "c"}, "assets/qa_raw/international_comparison/hard007_result.json|1777383640.6095273|4902": {"hash": "cf38f84655e3e0a98d28241646a1a1e7dd570d1aad222b293ba929209749d83d", "size": 4902, "status": "c"}, "assets/qa_raw/international_comparison/hard008_result.json|1777383640.6095273|3856": {"hash": "056433415512743ff68964bdb168408fa428d0ba935a1343482296663e7dcb0d", "size": 3856, "status": "c"}, "assets/qa_raw/international_comparison/hard009_result.json|1777383640.6095273|4141": {"hash": "ae003002bcc4c2affd7f426b908f1c7f608d47f679d75db37e901ed09c8fcc5b", "size": 4141, "status": "c"}, "assets/qa_raw/international_comparison/hard010_result.json|1777383640.6095273|3449": {"hash": "9fceedb48b40f0b8fb4cdfeaf4dcd3495688b7d12b175b54ef913503eab93b8e", "size": 3449, "status": "c"}, "assets/qa_raw/international_comparison/hard011_result.json|1777383640.6095273|3109": {"hash": "760cd895d90e31a84a0d80ca67ff645789ea6f0b90db922a0c7905d407e3f453", "size": 3109, "status": "c"}, "assets/qa_raw/international_comparison/hard012_result.json|1777383640.6095273|4294": {"hash": "a201effebd89fff0e835d2ce6dce2f550c0678c24272713659122b5d4da9c01e", "size": 4294, "status": "c"}, "assets/qa_raw/international_comparison/hard013_result.json|1777383640.6095273|3876": {"hash": "34674a53c3a41d254a5796d1697dd0b01e1b429590e69063b07d2acdd3051615", "size": 3876, "status": "c"}, "assets/qa_raw/international_comparison/hard014_result.json|1777383640.6095273|3125": {"hash": "dec3bd2510f90de94d4688d63f53c35f1e97567b1eddd1049f6d16b861cbf03d", "size": 3125, "status": "c"}, "assets/qa_raw/international_comparison/medium001_result.json|1777383640.6095273|2177": {"hash": "63567dca03d19d1541de888f0d4aae222609901e9e91756eee05cfc33f5c1599", "size": 2177, "status": "c"}, "assets/qa_raw/international_comparison/medium002_result.json|1777383640.6095273|2495": {"hash": "193a6f98fc99e4081114208ab5548d3f59f82ed894edf4c96bb91c33480c86a1", "size": 2495, "status": "c"}, "assets/qa_raw/international_comparison/medium003_result.json|1777383640.6252823|2589": {"hash": "65ecf9c549cc7a74617bbf6466b699edc17cfa61def70af6441685fac45cb721", "size": 2589, "status": "c"}, "assets/qa_raw/international_comparison/medium004_result.json|1777383640.6252823|2357": {"hash": "b02c9853f1d12b178e46dd0716308e55143820fe418c153b724e501687f07146", "size": 2357, "status": "c"}, "assets/qa_raw/international_comparison/medium005_result.json|1777383640.6252823|2515": {"hash": "d207c30132b2d425b8fe2d9e188f40ed386269e1cb86cd11595be264f5f56f37", "size": 2515, "status": "c"}, "assets/qa_raw/international_comparison/medium006_result.json|1777383640.6252823|2201": {"hash": "89d70defc34324336982d1e1318c44eb03e5cb864f7cb520414c1a01caf5661c", "size": 2201, "status": "c"}, "assets/qa_raw/international_comparison/medium007_result.json|1777383640.6252823|2409": {"hash": "37b233d12401ecedd9d78d2301bd695dbe80f25a39e34ac51f3856dd6cb069cd", "size": 2409, "status": "c"}, "assets/qa_raw/international_comparison/medium008_result.json|1777383640.6252823|2338": {"hash": "b63f5447485ca81d956f0475beabb93e037d0f007eac14835f7ffe8c18c16527", "size": 2338, "status": "c"}, "assets/qa_raw/international_comparison/medium009_result.json|1777383640.6252823|2972": {"hash": "a59d03e24473958f7a77ffd378a331e2edc7a957acc2e637e52cead51ccfaeda", "size": 2972, "status": "c"}, "assets/qa_raw/international_comparison/medium010_result.json|1777383640.6252823|2490": {"hash": "db36042723a5e86f53e64a6b836c8752d3c2507a906ad0913b51e71cd8ed5496", "size": 2490, "status": "c"}, "assets/qa_raw/international_comparison/medium011_result.json|1777383640.6252823|2405": {"hash": "633cd959d31d6f42d176f0feac3ecaf2e8f681555702533114a4301ef8288941", "size": 2405, "status": "c"}, "assets/qa_raw/international_comparison/medium012_result.json|1777383640.6252823|2484": {"hash": "f077174ef10b1d63282176c553f16350aa85b22d2c4b81413a8a1009ad38b221", "size": 2484, "status": "c"}, "assets/qa_raw/international_comparison/medium013_result.json|1777383640.6252823|2251": {"hash": "32837f60e922e331e7f5db54fc13a32c5dba288eef1a5bd6b690f52f73d95009", "size": 2251, "status": "c"}, "assets/qa_raw/international_comparison/medium014_result.json|1777383640.6409695|3127": {"hash": "8b99f53a7cbfae45fc9950659b48e5e788cd983a073c3ab38bb5f5f2bc54d07e", "size": 3127, "status": "c"}, "assets/qa_raw/international_comparison/medium015_result.json|1777383640.6409695|3321": {"hash": "40710c29958a729a19755b3da05f31db1c0a8c6fd5e9dba1758794faebb3d595", "size": 3321, "status": "c"}, "assets/qa_raw/international_comparison/medium016_result.json|1777383640.6409695|2221": {"hash": "842d90e811995ecc8a0f3ad66e685620c404dbf64d68cd0393c85aad723222c1", "size": 2221, "status": "c"}, "assets/qa_raw/international_comparison/medium017_result.json|1777383640.6409695|3374": {"hash": "c3c7f65937169feb289a0e73b240051c3fe2ccc370a9ffdafa1a8b7f0d9fd7e0", "size": 3374, "status": "c"}, "assets/qa_raw/international_comparison/medium018_result.json|1777383640.6409695|2392": {"hash": "48d124a3ed9b844ae85d6649dd2e06abcc494bb88168fe0acd1058bb2b3e90da", "size": 2392, "status": "c"}, "assets/qa_raw/international_comparison/medium019_result.json|1777383640.6409695|2304": {"hash": "45fde252f499c7a328a5d45d8fd3e8cc8f5a4a01abb96eb04765ca48b70e69b3", "size": 2304, "status": "c"}, "assets/qa_raw/international_comparison/medium020_result.json|1777383640.6409695|3012": {"hash": "aa78614640f785a6a985fc8b4d8fc5fcb1a22ec6e6b75e61fef45d31cfa6c08a", "size": 3012, "status": "c"}, "assets/qa_raw/international_comparison/medium021_result.json|1777383640.6409695|2697": {"hash": "ec4a83aa7f14c8537e8c6da8938d4226de3faf9d131292c8e6d4ea89aac74aa9", "size": 2697, "status": "c"}, "assets/qa_raw/international_comparison/medium022_result.json|1777383640.6409695|3088": {"hash": "681467fb569fca4cdfbf585a932e4e21a8b2f083372757512f17a81ed9dde449", "size": 3088, "status": "c"}, "assets/qa_raw/international_comparison/medium023_result.json|1777383640.6409695|2959": {"hash": "88c289e8b4f63ff3fe7477123d19e97a19dfbb9cbcf0ee7c23dbf56318294735", "size": 2959, "status": "c"}, "assets/qa_raw/international_comparison/medium024_result.json|1777383640.6409695|3402": {"hash": "b0724b82c8f538c3af5220941d75b94c57c415639e163ed286d0e507faedf0da", "size": 3402, "status": "c"}, "assets/qa_raw/international_comparison/medium025_result.json|1777383640.65667|2311": {"hash": "836968c170a8d0d82a2eaa4f24f9913947f6429463240a8ac4df58b43c1c8933", "size": 2311, "status": "c"}, "assets/qa_raw/risk_assessment/hard001_result.json|1777383640.65667|2941": {"hash": "a838ad554eeae2c957a191cc60f617fc729a55cebcc7f22fd31326ae1639867e", "size": 2941, "status": "c"}, "assets/qa_raw/risk_assessment/hard002_result.json|1777383640.65667|3055": {"hash": "76ca93701ce7c548b03e1e15bb25939c587cbd2b74b1ed1e9a354f5b1db3231e", "size": 3055, "status": "c"}, "assets/qa_raw/risk_assessment/hard003_result.json|1777383640.65667|4813": {"hash": "3412493ec710f9c4222f29111e18cde14c83154f27b918562cea0c5f11abb750", "size": 4813, "status": "c"}, "assets/qa_raw/risk_assessment/hard004_result.json|1777383640.65667|3105": {"hash": "bd788b179053c663655be2e4ac1f54812d0698c024091977c023ac72d2fe3ede", "size": 3105, "status": "c"}, "assets/qa_raw/risk_assessment/hard005_result.json|1777383640.65667|3178": {"hash": "b037fac0d28e28f45dac2f9a2aeb983301bd58adba50a2d30d5575ef0ca913a4", "size": 3178, "status": "c"}, "assets/qa_raw/risk_assessment/hard006_result.json|1777383640.65667|3802": {"hash": "4cf8a4c42c780d9f7588e9a6fc40a14bf6fa912bb78513ef91c9e75010ea0174", "size": 3802, "status": "c"}, "assets/qa_raw/risk_assessment/hard007_result.json|1777383640.65667|2621": {"hash": "13a9582af119a924bcf02fd75df51c78fbafde8b7dfcc2949dbb97282cab9715", "size": 2621, "status": "c"}, "assets/qa_raw/risk_assessment/hard008_result.json|1777383640.65667|2227": {"hash": "b6bc73301ada8601e78c753c0b9764a9ad8316ef810755ac81c9696e0e9a46cd", "size": 2227, "status": "c"}, "assets/qa_raw/risk_assessment/hard009_result.json|1777383640.65667|3291": {"hash": "ddb4dc135afdf3faa43dbb5efce5a8c8f34ccb5b25e60e27e1879e48f3e8ec3e", "size": 3291, "status": "c"}, "assets/qa_raw/risk_assessment/hard010_result.json|1777383640.65667|3777": {"hash": "47007968629c336449f7d61c7dd1a061884045bb8ec0df586490c216d6f53538", "size": 3777, "status": "c"}, "assets/qa_raw/risk_assessment/hard011_result.json|1777383640.6723459|4204": {"hash": "a704aac6d7b7ccdb0ee4090602df4fccf295f0743e622bf365763c2a7f47a374", "size": 4204, "status": "c"}, "assets/qa_raw/risk_assessment/hard012_result.json|1777383640.6723459|2802": {"hash": "83d0f74b8433ecfd66bc1eba867869de869bebe461059cd8f51ba57ed9b7cec4", "size": 2802, "status": "c"}, "assets/qa_raw/risk_assessment/hard013_result.json|1777383640.6723459|3466": {"hash": "4ce4ca88c1de758d6487768d0d10962b8540fe591fef0ced7a562c3138a0b863", "size": 3466, "status": "c"}, "assets/qa_raw/risk_assessment/medium001_result.json|1777383640.6723459|3366": {"hash": "aab7f513eed9d0f97bafac66ca4235505f64dc601d5d3ef0eebb27c245954e61", "size": 3366, "status": "c"}, "assets/qa_raw/risk_assessment/medium002_result.json|1777383640.6723459|3297": {"hash": "b9a25271e073e8d9f378071ef04ad7e7efc573dd7f56a279b85132ece7526ea7", "size": 3297, "status": "c"}, "assets/qa_raw/risk_assessment/medium003_result.json|1777383640.6723459|3548": {"hash": "57f0161f8281a281b24767bd331012b03616a85ece8d35ea151cccc794f07555", "size": 3548, "status": "c"}, "assets/qa_raw/risk_assessment/medium004_result.json|1777383640.6723459|3894": {"hash": "39143c7f7ca01a741e697baef3c9e838825ab4def53548d2ba1c74779fcebf32", "size": 3894, "status": "c"}, "assets/qa_raw/risk_assessment/medium005_result.json|1777383640.6723459|3831": {"hash": "c6a7017f2aef9058402b897eb3f5b6aae401d35c387a8c1d21595b8e2cd1c2ce", "size": 3831, "status": "c"}, "assets/qa_raw/risk_assessment/medium006_result.json|1777383640.6723459|3523": {"hash": "697feb76afb8125532be15a6420369c61d7fdf484ec657ed62ddc8092afc8847", "size": 3523, "status": "c"}, "assets/qa_raw/risk_assessment/medium007_result.json|1777383640.6723459|4129": {"hash": "7d47ff59af3cf9c2727e2e8dc8314a393c4ed314c7d68a3829e161c9fce23a2c", "size": 4129, "status": "c"}, "assets/qa_raw/risk_assessment/medium008_result.json|1777383640.688124|5133": {"hash": "3e7569bb7333b21521f239caa3e530020ed8557e6fbabc176bd242a034c1aa38", "size": 5133, "status": "c"}, "assets/qa_raw/risk_assessment/medium009_result.json|1777383640.688124|3553": {"hash": "de1d7b1a7b2f03a918017763942c7da22b855ba8adb68e230a0d1572a1d5df2b", "size": 3553, "status": "c"}, "assets/qa_raw/risk_assessment/medium010_result.json|1777383640.688124|3787": {"hash": "adc8d72cebe581bacd983497ef0a8e0f29b8f3e30ab9bab02a4f268a24bc1cbe", "size": 3787, "status": "c"}, "assets/qa_raw/risk_assessment/medium011_result.json|1777383640.688124|3921": {"hash": "88dfb33a8521862a758b9015a08a5dbca532eeb822edbf2a69205521efeb1c78", "size": 3921, "status": "c"}, "dataclaw/__init__.py|1777384240.8919768|0": {"hash": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", "size": 0, "status": "c"}, "dataclaw/__pycache__/__init__.cpython-312.pyc|1777385255.5657413|171": {"hash": "32bf712b9a0cc5d4b84a6d2205e4d441e63209aaadc68457bc449715ecfb3fd0", "size": 171, "status": "c"}, "dataclaw/__pycache__/lib_tasks.cpython-312.pyc|1777385255.5711854|7764": {"hash": "8c350110d5be60602e6bf4c43ab4e5abf7c6d9909a76b282959bd08d1c68ce8a", "size": 7764, "status": "c"}, "dataclaw/build_tasks.py|1777384240.892943|6765": {"hash": "5148e81a8aff1d0f4ae08681c9fc1ab87788011eff4da1799c2856f2763bac75", "size": 6765, "status": "c"}, "dataclaw/eval/__init__.py|1777384240.8939416|0": {"hash": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", "size": 0, "status": "c"}, "dataclaw/lib_tasks.py|1777384240.8959465|6283": {"hash": "d67ed5824d8c04226fe34492c9de255756a333036fa98d4c67b3eb829727447d", "size": 6283, "status": "c"}, "dataclaw/eval/run_batch.py|1777384240.8949378|38995": {"hash": "70ab176c9faed7090355df0be2216650dd87c0794f0bf74071d34defc86819fc", "size": 38995, "status": "c"}, "dataclaw/utils/__init__.py|1777384240.8969412|0": {"hash": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", "size": 0, "status": "c"}, "dataclaw/utils/__pycache__/__init__.cpython-312.pyc|1777385255.6985795|177": {"hash": "3f8481953e568cd1b64e5ce8d5fc2e378f4127656c2b29ea666c0a2537fbb3aa", "size": 177, "status": "c"}, "dataclaw/utils/__pycache__/docker_utils.cpython-312.pyc|1777385255.7148151|31479": {"hash": "cdbdc5372669b17b9270bdd62a3cbd01aa6e02137768313a51b8ab208f0c52c3", "size": 31479, "status": "c"}, "dataclaw/utils/__pycache__/grading.cpython-312.pyc|1777385255.7299564|19465": {"hash": "a41487e390656066d2c324ed63cf59cb65280bace51af164c2cd7c3ef0390ecb", "size": 19465, "status": "c"}, "dataclaw/utils/__pycache__/process_grading.cpython-312.pyc|1777385255.7542684|29460": {"hash": "92d62ae2bbd5ea98cd1856c572ac5cf774ee8b12b1ae73c9ff5310838c4f3bd1", "size": 29460, "status": "c"}, "dataclaw/utils/docker_utils.py|1777384240.8979416|25416": {"hash": "cb3afb4fb0cb29731fc7ba6c8cbe04d90ab11813f0704094e3906589ac707ac3", "size": 25416, "status": "c"}, "dataclaw/utils/grading.py|1777384240.8999395|18394": {"hash": "ae4200628281da4afd4b8e56ceb5d298dcdeb0b17e8410bc8b147ff835a072be", "size": 18394, "status": "c"}, "dataclaw/utils/process_grading.py|1777384240.9009378|28246": {"hash": "80c6e000c64270a26da0fd8541287c57d58b62c0d18c075c6c339f8fb4a7a793", "size": 28246, "status": "c"}, "docs/index.html|1777384240.903942|876446": {"hash": "4617fa67f981088e7fbf315850fb6566cded87a7ab595cab73ad828361934eb2", "size": 876446, "status": "c"}, "logo.png|1777384240.9674158|7072570": {"hash": "12f8390beafa1b94888cef1b7e78e329598894366be4529b0efd9344f689ec6c", "size": 7072570, "status": "c"}, "pyproject.toml|1777384240.9674158|874": {"hash": "ac4a9b6592404a1dd9a1580cff9edd0b61c6b16f68fc86d74252cd2a9ce0e21b", "size": 874, "status": "c"}, "script/docker_save_image.sh|1777384240.9674158|1407": {"hash": "383c711d16be15997df337b3e338a3557ca79b8f4773b6048a375f6b31e21687", "size": 1407, "status": "c"}, "script/run.sh|1777384240.9674158|926": {"hash": "ecf6558d560656f296b12f696bd38152b45965fac285947311fbbed233f2134d", "size": 926, "status": "c"}, "tasks/task_001_comprehensive_decision_easy_easy001.md|1777384240.9674158|3849": {"hash": "c60775330b985bc309f8b91155ea67386fdb2bcccb3dab1a66ba41cd1f8ec9a0", "size": 3849, "status": "c"}, "tasks/task_002_comprehensive_decision_easy_easy002.md|1777384240.9674158|4157": {"hash": "6593c843c39f18e398d5da3cba99bb61cb503f773ad558e8c9c6092c54a108f4", "size": 4157, "status": "c"}, "tasks/task_003_comprehensive_decision_easy_easy003.md|1777384240.9674158|3875": {"hash": "ea7d7e8520c33437ff2a346079d47ae89bb4cd584dfb73c10a18758e3f07893f", "size": 3875, "status": "c"}, "tasks/task_004_comprehensive_decision_easy_easy004.md|1777384240.9674158|4186": {"hash": "a20d9042985f76879ac0d937879aa5fe084e9743c768787fed60fc7569f287f9", "size": 4186, "status": "c"}, "tasks/task_005_comprehensive_decision_easy_easy005.md|1777384240.9674158|4133": {"hash": "2c67d94cbfebf8add084937e9c83303d06b227e23f74a6973696d8774da70ad3", "size": 4133, "status": "c"}, "tasks/task_006_comprehensive_decision_easy_easy006.md|1777384240.9674158|3946": {"hash": "c9ccadbfa7a54334e2534d396fd7ec67c4ff9ea2188b80d9c24639a20f79b68b", "size": 3946, "status": "c"}, "tasks/task_007_comprehensive_decision_hard_hard001.md|1777384240.9674158|4930": {"hash": "9cf587d24f0a6e2880791187868a484436d68cf7fd2794a64719fbac3f30f862", "size": 4930, "status": "c"}, "tasks/task_008_comprehensive_decision_hard_hard002.md|1777384240.9674158|4849": {"hash": "0d2800eab651e5f5f7c390a536f4835582f495e148ac01a37ef68817c9d0b2fa", "size": 4849, "status": "c"}, "tasks/task_009_comprehensive_decision_hard_hard003.md|1777384240.9674158|5145": {"hash": "c419bc2e5e43765c4ca61a1a482209dfb666710ea6d9165ab2dd6b79d42a6f2d", "size": 5145, "status": "c"}, "tasks/task_010_comprehensive_decision_hard_hard004.md|1777384240.9674158|4854": {"hash": "5e720afd6a709f384f6659829b1fd3001b89de1ebd1891efb74e92104320cf1f", "size": 4854, "status": "c"}, "tasks/task_011_comprehensive_decision_hard_hard005.md|1777384240.9830606|4861": {"hash": "a35722327034c4397d75893b65f3255f77d82ee5ee12478af3d7f7b741d147bd", "size": 4861, "status": "c"}, "tasks/task_012_comprehensive_decision_hard_hard006.md|1777384240.9830606|4891": {"hash": "21cc2680bde39e902cb368800e33cd9009da7f08f6c88184a7bb2f8dfc1081ad", "size": 4891, "status": "c"}, "tasks/task_013_comprehensive_decision_hard_hard007.md|1777384240.9830606|4707": {"hash": "597e3fcfb4dfa6e6a908ff3107d91ac6cd75734425c1f4f418dce09dd55f1bb3", "size": 4707, "status": "c"}, "tasks/task_014_comprehensive_decision_hard_hard008.md|1777384240.9830606|4788": {"hash": "9eb67806613ae218780f3e358076b20417915f11d9b7d77e9b45b2c3a13ac6b8", "size": 4788, "status": "c"}, "tasks/task_015_comprehensive_decision_hard_hard009.md|1777384240.9830606|4713": {"hash": "5276116a1081fc5b51e495d9447a901905390061f2f4c8f5151174b3e5081a74", "size": 4713, "status": "c"}, "tasks/task_016_comprehensive_decision_hard_hard010.md|1777384240.9830606|4738": {"hash": "0cae7c7fa7ddb552abe2fdef3c9ffe9e3acdf89a6a73d490db4d87b33559fc3a", "size": 4738, "status": "c"}, "tasks/task_017_comprehensive_decision_hard_hard011.md|1777384240.9830606|4870": {"hash": "c0b3e2de2bab930e786ec4044f2a64ab132593eab29e7b99e0d9dfa9e6244bf9", "size": 4870, "status": "c"}, "tasks/task_018_comprehensive_decision_hard_hard012.md|1777384240.9830606|4785": {"hash": "9d941abc2d7964435e02fa7d67ff519b4cafb0519cdf792c1a2c5906913e5490", "size": 4785, "status": "c"}, "tasks/task_019_comprehensive_decision_hard_hard013.md|1777384240.9830606|4369": {"hash": "a4046f8b64a59343fa98a564beed1b64c0efb029c881857162fa3cee0bd8436a", "size": 4369, "status": "c"}, "tasks/task_020_comprehensive_decision_hard_hard014.md|1777384240.9830606|4402": {"hash": "454314264861b4c5bd763528b2b68adf8336b8deec8af24c86b68f07ba756c5b", "size": 4402, "status": "c"}, "tasks/task_021_comprehensive_decision_hard_hard015.md|1777384240.9830606|4709": {"hash": "b30ec692692134a6baa2fc4e0dd329b11cf2b5a7008c8134356e23c2dbbc0b18", "size": 4709, "status": "c"}, "tasks/task_022_comprehensive_decision_hard_hard016.md|1777384240.9830606|4745": {"hash": "a37f0733ebfefe2062d745fc5d38c4e8a2a15dc338c9c927cfd8a3d8a130d821", "size": 4745, "status": "c"}, "tasks/task_023_comprehensive_decision_hard_hard017.md|1777384240.9830606|4524": {"hash": "c21ad437400e104d286a1f55de3fc842cbe21c44ab39ce4f25c02d94d27cf93e", "size": 4524, "status": "c"}, "tasks/task_024_comprehensive_decision_hard_hard018.md|1777384240.9830606|4371": {"hash": "90b9cdec67492ef8f838c59f0c4ab8925f8a71cb400c18d0d19e7ba2d2eb1b80", "size": 4371, "status": "c"}, "tasks/task_025_comprehensive_decision_hard_hard019.md|1777384240.9830606|4790": {"hash": "ddd3c8ae3c24ede81237dfd927402a45d474246e7b7261732c7c05bcf2333ebb", "size": 4790, "status": "c"}, "tasks/task_026_comprehensive_decision_medium_medium001.md|1777384240.9830606|4433": {"hash": "27814139c460510991c929e63fa56d7244ca79c3cbbe4a88b9f2f305ab6cb4fa", "size": 4433, "status": "c"}, "tasks/task_027_comprehensive_decision_medium_medium002.md|1777384240.9830606|4331": {"hash": "f281b17176afa32f3d1f08031ebcb92671bf3165e23a671c097dbce4d23861a8", "size": 4331, "status": "c"}, "tasks/task_028_comprehensive_decision_medium_medium003.md|1777384240.9830606|4357": {"hash": "510f451e12c2205d7a30a3427f076fb5de74c6d0a906f3441a16ceed6a6dbb44", "size": 4357, "status": "c"}, "tasks/task_029_comprehensive_decision_medium_medium004.md|1777384240.9830606|4489": {"hash": "b9baff097282129df0dc83d521f3cb878b7e78989cd0ad72c51499d323209cc9", "size": 4489, "status": "c"}, "tasks/task_030_comprehensive_decision_medium_medium005.md|1777384240.9830606|4068": {"hash": "6e9f950acaa1f0b673259e968f5cb9e310475fa76a0d16206537a170403bd41e", "size": 4068, "status": "c"}, "tasks/task_031_comprehensive_decision_medium_medium006.md|1777384240.9988937|4082": {"hash": "86ecfdb15f8fd143374b5d774eb679213cd4de157b2050707a93e65bc02a36b2", "size": 4082, "status": "c"}, "tasks/task_032_comprehensive_decision_medium_medium007.md|1777384240.9988937|4380": {"hash": "e117f0923f72d7be2b5b495df481db6b5e5678b0e68a02755dc10376160507d6", "size": 4380, "status": "c"}, "tasks/task_033_comprehensive_decision_medium_medium008.md|1777384240.9988937|4068": {"hash": "70b5efc390cfde2c4e0cfa5723c02c2480733e428da379b8ccf4fb76645d5d10", "size": 4068, "status": "c"}, "tasks/task_034_comprehensive_decision_medium_medium009.md|1777384240.9988937|4243": {"hash": "b21bdd8f568655d401bf9152965d8205c8d8a4468b3590f645750a67cec6bca2", "size": 4243, "status": "c"}, "tasks/task_035_comprehensive_decision_medium_medium010.md|1777384240.9988937|4329": {"hash": "3a683dfb12af4a438b3ec2d78964934669d4440a22fff104a01d16b90ccdf0e5", "size": 4329, "status": "c"}, "tasks/task_036_comprehensive_decision_medium_medium011.md|1777384240.9988937|3881": {"hash": "492254f45ef312f6716ea58c99f9dd039ea6ddb78a4c51331cd8f93ebad97283", "size": 3881, "status": "c"}, "tasks/task_037_comprehensive_decision_medium_medium012.md|1777384240.9988937|4289": {"hash": "b84d9743d8ddb24a85ae64535642a599d53eaeb1367ed451863849097b7f8067", "size": 4289, "status": "c"}, "tasks/task_038_comprehensive_decision_medium_medium013.md|1777384240.9988937|4720": {"hash": "adea540bd06aa5edaea4d7758cf32deba26cdd38966e46fc9727a86c04c03c23", "size": 4720, "status": "c"}, "tasks/task_039_comprehensive_decision_medium_medium014.md|1777384240.9988937|4654": {"hash": "907245ee7dd6aa2483e81087cba2a4d79dc522986a2918bf57f2206d430da6cc", "size": 4654, "status": "c"}, "tasks/task_040_comprehensive_decision_medium_medium015.md|1777384240.9988937|5140": {"hash": "4d09a8eb965104f5cb63cf10a95b60a2b6b6fe894df20b17f74c3a8be252f0c8", "size": 5140, "status": "c"}, "tasks/task_041_comprehensive_decision_medium_medium016.md|1777384240.9988937|3974": {"hash": "9fe40c91881a32b0b1c9d8eb1ede925ea022c63f7ff71cfde7428c63551cc195", "size": 3974, "status": "c"}, "tasks/task_042_comprehensive_decision_medium_medium017.md|1777384240.9988937|3926": {"hash": "95aec4fa30b9415b037d34d2175abe37e0cf3a67a486ddf25210fc173cc5e8e4", "size": 3926, "status": "c"}, "tasks/task_043_comprehensive_decision_medium_medium018.md|1777384240.9988937|3867": {"hash": "c2f6b9f64879a244a1e9f687613cd489695392faeca6cd4427ff524f535f9176", "size": 3867, "status": "c"}, "tasks/task_044_comprehensive_decision_medium_medium019.md|1777384240.9988937|3903": {"hash": "b03ad7c60d6544eb1fa15703690613b514de75419b9af23d55cb6d41bac12304", "size": 3903, "status": "c"}, "tasks/task_045_comprehensive_decision_medium_medium020.md|1777384240.9988937|3961": {"hash": "d3d3e6c08e84886649800545ad28008e2b1a8697af89325992c9f2dd8013163b", "size": 3961, "status": "c"}, "tasks/task_046_comprehensive_decision_medium_medium021.md|1777384240.9988937|3966": {"hash": "e9f28ebb431ef28c2a9262b294b67dd68f6ceb9d3078804f7ec01623f5cb31c1", "size": 3966, "status": "c"}, "tasks/task_047_comprehensive_decision_medium_medium022.md|1777384240.9988937|3853": {"hash": "a5a049bc5a0d2da5b0b64cc3d3b8e55d97e224388f3baed5067bacbbb4f88556", "size": 3853, "status": "c"}, "tasks/task_049_comprehensive_decision_medium_medium024.md|1777384240.9988937|3881": {"hash": "192c2405597af52de17617c87a9d3fe93064c4fb3126f0856ea94e30fa10410f", "size": 3881, "status": "c"}, "tasks/task_048_comprehensive_decision_medium_medium023.md|1777384240.9988937|3914": {"hash": "003fabddfe910a9eb237baced1a32728bd407ea096bd5633395d8f08e1bb573a", "size": 3914, "status": "c"}, "tasks/task_050_comprehensive_decision_medium_medium025.md|1777384240.9988937|3863": {"hash": "32422621fa571c96fd04bbfa62e3f054b612cc97913d65cd635e070b11009ba3", "size": 3863, "status": "c"}, "tasks/task_051_comprehensive_decision_medium_medium026.md|1777384240.9988937|3920": {"hash": "35b367816309193a08f3aed7674d7b579e24a70f7efa6a9bcc70a3536010e9ff", "size": 3920, "status": "c"}, "tasks/task_052_comprehensive_decision_medium_medium027.md|1777384241.0148966|4228": {"hash": "712645c2907cdcb15f7ded3d1dfd04b6a7893a76657fed67e7ba8c081ad358d8", "size": 4228, "status": "c"}, "tasks/task_053_comprehensive_decision_medium_medium028.md|1777384241.0148966|4747": {"hash": "45875185f36e657377b25674a4d7fa52967214fbf522bc295c857149c3f3c57f", "size": 4747, "status": "c"}, "tasks/task_055_comprehensive_decision_medium_medium030.md|1777384241.0148966|3902": {"hash": "af65645172cdaf1fd8238de06f75fa90e86cb18a137bbdc84713ae868f4325e6", "size": 3902, "status": "c"}, "tasks/task_054_comprehensive_decision_medium_medium029.md|1777384241.0148966|4006": {"hash": "be2aac02102c91f653c2c305c464c85854005cad815cfa4e01791ade18628b25", "size": 4006, "status": "c"}, "tasks/task_056_comprehensive_decision_medium_medium031.md|1777384241.0148966|3884": {"hash": "0dba5cb08bd8d734a11c6cdc0bf88245d4cadabff77d87a4b682487c75d17767", "size": 3884, "status": "c"}, "tasks/task_057_comprehensive_decision_medium_medium032.md|1777384241.0148966|3883": {"hash": "dab699cbe19644dec2d0eea21c70391db38e4a081e24ead35f749774d9c58153", "size": 3883, "status": "c"}, "tasks/task_058_comprehensive_decision_medium_medium033.md|1777384241.0148966|4100": {"hash": "b5cd58eca546d576a98fd1c50cff1349003fb98a6fc6698be86fb9d6fd9ff0f5", "size": 4100, "status": "c"}, "tasks/task_059_comprehensive_decision_medium_medium034.md|1777384241.0148966|4025": {"hash": "9919e6050445df98549b9effc4e3ef2e330a134febc7520968be8be0c4647b0f", "size": 4025, "status": "c"}, "tasks/task_060_comprehensive_decision_medium_medium035.md|1777384241.0148966|4068": {"hash": "d9ca322d6231d950c4b5a1953c2d8cfb8930733371a83f96c22e0133998fd2a6", "size": 4068, "status": "c"}, "tasks/task_061_comprehensive_decision_medium_medium036.md|1777384241.0148966|3970": {"hash": "12ae51942a4176c6bf4d323707a11ba19b872f4efbb1de45e51292913c7df4da", "size": 3970, "status": "c"}, "tasks/task_062_comprehensive_decision_medium_medium037.md|1777384241.0148966|3974": {"hash": "3029f6784e821cc54902a39c9129fbfe0f274f48ec34d28018f7cf2cfbebbff8", "size": 3974, "status": "c"}, "tasks/task_063_comprehensive_decision_medium_medium038.md|1777384241.0148966|3932": {"hash": "8d82835d10b21b7bc482b7467390e56827fe98e7b2254a6e56db6e314c7cc99e", "size": 3932, "status": "c"}, "tasks/task_064_comprehensive_decision_medium_medium039.md|1777384241.0148966|4904": {"hash": "6dd51203ec0e05c013baf809f1613b2eb3e6440f93bc29888cf81f6bb62dc57d", "size": 4904, "status": "c"}, "tasks/task_065_comprehensive_decision_medium_medium040.md|1777384241.0148966|4012": {"hash": "285e245f6550bd523899120cc974b842d103c0682624e3798b4792dbb809fb75", "size": 4012, "status": "c"}, "tasks/task_066_comprehensive_decision_medium_medium041.md|1777384241.0148966|3883": {"hash": "292597ce450ba511252cf2a1a8e858122494ef25ba7ebdfebe942524038ac458", "size": 3883, "status": "c"}, "tasks/task_067_comprehensive_decision_medium_medium042.md|1777384241.0148966|4090": {"hash": "28c3006b2d3c3ca5576e3c79b4e0eab8856924bf40a70278ca03b1563c8d4df4", "size": 4090, "status": "c"}, "tasks/task_068_comprehensive_decision_medium_medium043.md|1777384241.0148966|3887": {"hash": "10869183aca881886e3704dc4d3f73a4a6880e448967edf0bb807e7f81633429", "size": 3887, "status": "c"}, "tasks/task_069_comprehensive_decision_medium_medium044.md|1777384241.0148966|3992": {"hash": "bde30a7bdd41003a944b59eadba76b77aef73c0055d11f0f1a6aa3c18f23d4d6", "size": 3992, "status": "c"}, "tasks/task_070_comprehensive_decision_medium_medium045.md|1777384241.0148966|3912": {"hash": "d838b49d4c116a7dea02d857416d95674ce5b9dce5a9d0ac573e05149e32ca6d", "size": 3912, "status": "c"}, "tasks/task_071_enterprise_industry_analysis_easy_easy001.md|1777384241.0148966|3956": {"hash": "df5f2ac484a9e67f12920d45261c15ebe5a4095823428598cde401d9e3afdcb2", "size": 3956, "status": "c"}, "tasks/task_072_enterprise_industry_analysis_easy_easy002.md|1777384241.0148966|3958": {"hash": "8f71f96378cdf8b57f3a9d4842681e6655a124fd3946f45e0a4473be6b13b41f", "size": 3958, "status": "c"}, "tasks/task_073_enterprise_industry_analysis_easy_easy003.md|1777384241.030878|4003": {"hash": "05599d644d8927dffb0b86c41ee0c21fb8aabad8c9bf70b1c2d4c28dca5ebfbc", "size": 4003, "status": "c"}, "tasks/task_074_enterprise_industry_analysis_easy_easy004.md|1777384241.030878|3901": {"hash": "859cb98dc0d08b9de55fca5c5831d6954a095033caab6b9d283eac428f398639", "size": 3901, "status": "c"}, "tasks/task_075_enterprise_industry_analysis_easy_easy005.md|1777384241.030878|3885": {"hash": "c4bdb6c5a4615c1dc7d6cf3d0181472cfb9da68b93995c84200f124ac3c78711", "size": 3885, "status": "c"}, "tasks/task_076_enterprise_industry_analysis_easy_easy006.md|1777384241.030878|3919": {"hash": "45560d1807736f700fa58972e3d130f8231180a7c9202c459b219fa89874974d", "size": 3919, "status": "c"}, "tasks/task_077_enterprise_industry_analysis_easy_easy007.md|1777384241.030878|3919": {"hash": "c0ee49466653cfda904bf9e5448658efc525487560390fe864e9ebd4e102ac18", "size": 3919, "status": "c"}, "tasks/task_078_enterprise_industry_analysis_easy_easy008.md|1777384241.030878|3937": {"hash": "1f141c92469c13b853394bcc73cd507b2fed5ecf68d7058275ea24381d5245d9", "size": 3937, "status": "c"}, "tasks/task_079_enterprise_industry_analysis_easy_easy009.md|1777384241.030878|3892": {"hash": "7f2e293a5d1826a156da141991b4c5380645f0ae49b938481278ad8f8ab781ea", "size": 3892, "status": "c"}, "tasks/task_080_enterprise_industry_analysis_easy_easy010.md|1777384241.030878|3920": {"hash": "c9575706081349f0e468da05639dd0b089b3e4c81dc3c66640b64c88d1375776", "size": 3920, "status": "c"}, "tasks/task_081_enterprise_industry_analysis_easy_easy011.md|1777384241.030878|3889": {"hash": "a4c3ac305b037c4a45f7d4336e34139639f68a3355cee3e5d908f81bdc7849af", "size": 3889, "status": "c"}, "tasks/task_082_enterprise_industry_analysis_easy_easy012.md|1777384241.030878|3879": {"hash": "7d810bf2e36fc033fda121d03ae5ec6fa22130521998c2e7cc8c79e120deeb6b", "size": 3879, "status": "c"}, "tasks/task_083_enterprise_industry_analysis_easy_easy013.md|1777384241.030878|3879": {"hash": "4e3a2495435455992798f5068d2f7882bfe415d1b0c9950ed05501246a957bd2", "size": 3879, "status": "c"}, "tasks/task_084_enterprise_industry_analysis_easy_easy014.md|1777384241.030878|3873": {"hash": "6496d5d3f4d4b44159bcc1ef547ba6213c8001aa15eaa801a0b909a71a065a34", "size": 3873, "status": "c"}, "tasks/task_085_enterprise_industry_analysis_easy_easy015.md|1777384241.030878|3904": {"hash": "bc47a0771368acdf27ffb8b91de3266357217f73f0e632b58ddfa71269e35d0f", "size": 3904, "status": "c"}, "tasks/task_086_enterprise_industry_analysis_easy_easy016.md|1777384241.030878|3971": {"hash": "f1ff6b5efb6bd45b587352ef61b5af7182fa84246d26cf9c6890724d4d185d66", "size": 3971, "status": "c"}, "tasks/task_087_enterprise_industry_analysis_easy_easy017.md|1777384241.030878|3936": {"hash": "36fc4c17c1c1e25946390e36fcc90953a27b3153524ac636869016cfa1b3e257", "size": 3936, "status": "c"}, "tasks/task_088_enterprise_industry_analysis_easy_easy018.md|1777384241.030878|3922": {"hash": "5e02eece4553f57dd02675a66a138de257df0ff312f8c5724ae3f0510fd925be", "size": 3922, "status": "c"}, "tasks/task_089_enterprise_industry_analysis_easy_easy019.md|1777384241.030878|3901": {"hash": "f5db7aeeb2bb4400c762a34337f9444e7824115ebb7e3ff1d0ccbae528b91a1b", "size": 3901, "status": "c"}, "tasks/task_091_enterprise_industry_analysis_easy_easy021.md|1777384241.030878|3917": {"hash": "e4b169788b5c299b4c974474eb2882dd173ff2b162326a7d6932016b381c2134", "size": 3917, "status": "c"}, "tasks/task_090_enterprise_industry_analysis_easy_easy020.md|1777384241.030878|3901": {"hash": "e393cef8f41b587cc6e7c594e0669093af50b261404563937e15ab7e889f6508", "size": 3901, "status": "c"}, "tasks/task_092_enterprise_industry_analysis_easy_easy022.md|1777384241.030878|3930": {"hash": "0a2be9d6139431d65c160e1522cf928b04763e3960e4a9f1be26ce4b16d155a5", "size": 3930, "status": "c"}, "tasks/task_093_enterprise_industry_analysis_easy_easy023.md|1777384241.030878|4025": {"hash": "91987d80037e753e41bcf05910c42693263d7f8f10ef613e204617f4c9a1a8df", "size": 4025, "status": "c"}, "tasks/task_094_enterprise_industry_analysis_easy_easy024.md|1777384241.030878|3925": {"hash": "1215847c3e34e2840edbe135b467e46d0961453c7f112a534aec153924618d21", "size": 3925, "status": "c"}, "tasks/task_095_enterprise_industry_analysis_easy_easy025.md|1777384241.0466597|3898": {"hash": "95627c4aa0fd893f784c66d06f8d937645fc2f1887cf16f8c4b906c6ad365ffa", "size": 3898, "status": "c"}, "tasks/task_096_enterprise_industry_analysis_easy_easy026.md|1777384241.0466597|3897": {"hash": "7753c2858bbff9d755db7e320f01c887270e197c341a0bb108f91d976251b848", "size": 3897, "status": "c"}, "tasks/task_097_enterprise_industry_analysis_easy_easy027.md|1777384241.0466597|3939": {"hash": "06a4a26c469523940facdd93959635cf78211d5395849e375546aa22f36a2951", "size": 3939, "status": "c"}, "tasks/task_098_enterprise_industry_analysis_easy_easy028.md|1777384241.0466597|3899": {"hash": "d633624282e3b9231fa988977897783667c0d771823752e30dc79c479d54bdf9", "size": 3899, "status": "c"}, "tasks/task_099_enterprise_industry_analysis_easy_easy029.md|1777384241.0466597|3907": {"hash": "ff9e176879d3529e66e79ce2ce9c25f8d3354d81ab4a39dd1bde8b3f731ee8b0", "size": 3907, "status": "c"}, "tasks/task_100_enterprise_industry_analysis_easy_easy030.md|1777384241.0502036|3875": {"hash": "ff3d5439069ac16ebeda7d45bd4a27e9f4351e7df9e8ee62b5ba108f02452ce6", "size": 3875, "status": "c"}, "tasks/task_101_enterprise_industry_analysis_easy_easy031.md|1777384241.0502036|3966": {"hash": "1983e7e7da34b484e0e1d745995a0f7e0d6138ddca2897c35a9c593d5c5af7ec", "size": 3966, "status": "c"}, "tasks/task_102_enterprise_industry_analysis_easy_easy032.md|1777384241.0502036|3929": {"hash": "2160ca71b0bf0120c4d6c82ef4795362f34cb9240529bf44d79054b1346ba698", "size": 3929, "status": "c"}, "tasks/task_103_enterprise_industry_analysis_easy_easy033.md|1777384241.0502036|3946": {"hash": "11aa106f7150616d1545dfa27ddfe2310f1928ffb3ee419ec155b670ebd7c66c", "size": 3946, "status": "c"}, "tasks/task_104_enterprise_industry_analysis_easy_easy034.md|1777384241.0502036|3987": {"hash": "5a76d8927283f910e09da8c7a5c64553073ce8c630c615f8861cafeec75d15e3", "size": 3987, "status": "c"}, "tasks/task_105_enterprise_industry_analysis_easy_easy035.md|1777384241.0502036|3859": {"hash": "17e507a956ccb89c8508ca3bd9e6af7dda495792ca30682b3bebc63e3ba02a71", "size": 3859, "status": "c"}, "tasks/task_106_enterprise_industry_analysis_easy_easy036.md|1777384241.0502036|3933": {"hash": "771a35ac59e4fad440a6892c9ec643858af1d01058d1d975461c8a29c2151442", "size": 3933, "status": "c"}, "tasks/task_107_enterprise_industry_analysis_easy_easy037.md|1777384241.0502036|3871": {"hash": "f71cfdd20b63533a71f1e816974d2f410f549bf0999cae3a6fe2ddf74e3701aa", "size": 3871, "status": "c"}, "tasks/task_108_enterprise_industry_analysis_easy_easy038.md|1777384241.0502036|3883": {"hash": "6f89837bfb27998f38447ae60b3314770eedac98c1f1a1379ac5daaf5df7876e", "size": 3883, "status": "c"}, "tasks/task_109_enterprise_industry_analysis_easy_easy039.md|1777384241.0502036|3879": {"hash": "cb2cf1e2e5ca9eff99e75b3ba507afeb99a4afcde1e99ace889f7afb8c2529f2", "size": 3879, "status": "c"}, "tasks/task_110_enterprise_industry_analysis_easy_easy040.md|1777384241.0502036|3885": {"hash": "c60a280cf76358bc6e86d7bf2157185103d8ca871eaabdcbe38c87cc80bb3b2b", "size": 3885, "status": "c"}, "tasks/task_111_enterprise_industry_analysis_easy_easy041.md|1777384241.0502036|3985": {"hash": "bb424132a7650b2030b9e15c2b8de04a2aff8bf6b02c44b994adf6ef048019b7", "size": 3985, "status": "c"}, "tasks/task_112_enterprise_industry_analysis_easy_easy042.md|1777384241.0502036|3890": {"hash": "8760880bfdbe9f5f81fb861ac5248f88912cfa00c5de03d55b2e023bd7f468bc", "size": 3890, "status": "c"}, "tasks/task_113_enterprise_industry_analysis_easy_easy043.md|1777384241.0502036|3985": {"hash": "eeac118bd74c4e11f2ff5c4cb233f761eeeb61ddac804d6022f51d9391229c2b", "size": 3985, "status": "c"}, "tasks/task_114_enterprise_industry_analysis_easy_easy044.md|1777384241.0502036|3895": {"hash": "ab695b8679c897182e30fd9e188981528d7b3c299ed3c755cc32ecd3e9ee1e5b", "size": 3895, "status": "c"}, "tasks/task_115_enterprise_industry_analysis_easy_easy045.md|1777384241.0502036|3893": {"hash": "e5c35820c31c3edf803817ca211b59ee80ca49136ed57a5c05b72c1e14da7fbc", "size": 3893, "status": "c"}, "tasks/task_116_enterprise_industry_analysis_easy_easy046.md|1777384241.0623188|3894": {"hash": "7403198ca37ec240c83e1aa5544b974305bd860dff87b6955add50a74e61014b", "size": 3894, "status": "c"}, "tasks/task_117_enterprise_industry_analysis_easy_easy047.md|1777384241.0623188|3881": {"hash": "4fe2d058dea5281c30d39114cd09d65543e9fb1434725050dca2b1ec8e5b55a6", "size": 3881, "status": "c"}, "tasks/task_118_enterprise_industry_analysis_easy_easy048.md|1777384241.0623188|3888": {"hash": "1b278daff743fb0ff7fb3cf509535b5a25025df0ca740c4cdcc0516ff14ff6ff", "size": 3888, "status": "c"}, "tasks/task_119_enterprise_industry_analysis_easy_easy049.md|1777384241.0623188|3871": {"hash": "e9278fdee1f475dfeb7dc67932705a5a36e42886a1d9f70d41e82d9a0504ab46", "size": 3871, "status": "c"}, "tasks/task_120_enterprise_industry_analysis_easy_easy050.md|1777384241.0623188|3872": {"hash": "87701c664277464b6bb1382dfbb2cb7990abc1c6610bb035f02a8bc18cc1d3bc", "size": 3872, "status": "c"}, "tasks/task_121_enterprise_industry_analysis_easy_easy051.md|1777384241.0623188|3882": {"hash": "5fed1e45d561064d879a8e104220f5c0ef8871f435239193e88b7894d5c43cc0", "size": 3882, "status": "c"}, "tasks/task_122_enterprise_industry_analysis_easy_easy052.md|1777384241.0623188|3880": {"hash": "410fe73ffcce706908ea403a182f2cbcf86e1ab67272a87ecc50026958664d49", "size": 3880, "status": "c"}, "tasks/task_123_enterprise_industry_analysis_easy_easy053.md|1777384241.0623188|3916": {"hash": "81c1e9382b1167afaf147cc93a267fda4a626943efd5b5363aa9f9f5ea7c4a57", "size": 3916, "status": "c"}, "tasks/task_124_enterprise_industry_analysis_easy_easy054.md|1777384241.0623188|3881": {"hash": "0adee26b800bd171e8b9117ff619e7dc2b40f4b0f82bac8958b9f9cddcde8e0c", "size": 3881, "status": "c"}, "tasks/task_125_enterprise_industry_analysis_easy_easy055.md|1777384241.0623188|3966": {"hash": "3ca95e504fecbb58ca84956b0460eb925e3da72149c55b8d3aaef377dd58c5d0", "size": 3966, "status": "c"}, "tasks/task_126_enterprise_industry_analysis_easy_easy056.md|1777384241.0623188|3974": {"hash": "70caf4ccd6ab67ad972c189bc2b7c65718956d90da9ecc039bf3d60f7853b096", "size": 3974, "status": "c"}, "tasks/task_127_enterprise_industry_analysis_easy_easy057.md|1777384241.0623188|3901": {"hash": "8543de68ac612ae841b45c9f8b113716e2fdc20a68e27969115bc4bd2a1a5898", "size": 3901, "status": "c"}, "tasks/task_128_enterprise_industry_analysis_easy_easy058.md|1777384241.0623188|3910": {"hash": "1e0d37b2c55408dbb35c48aad4a081776274b46f0f2302c3360e3108295c5748", "size": 3910, "status": "c"}, "tasks/task_129_enterprise_industry_analysis_easy_easy059.md|1777384241.0623188|3917": {"hash": "27b3a3e0717dda9ed941b95205691f920368ac915999249e581d79e7bab1b9fd", "size": 3917, "status": "c"}, "tasks/task_130_enterprise_industry_analysis_easy_easy060.md|1777384241.0623188|4078": {"hash": "4366347cb02482c667a618031ad2939d39dab751a6f2795fa5687313f27ba277", "size": 4078, "status": "c"}, "tasks/task_131_enterprise_industry_analysis_easy_easy061.md|1777384241.0623188|4044": {"hash": "c812bcdc4b700be9c6efca6f21a11721d0165f4ccf0feae506369e0d9f1bab7e", "size": 4044, "status": "c"}, "tasks/task_132_enterprise_industry_analysis_easy_easy062.md|1777384241.0623188|4020": {"hash": "18b803ef6e66009322169a4015b6192b8d492efe6f7c81807a1b1c5a56ba842c", "size": 4020, "status": "c"}, "tasks/task_133_enterprise_industry_analysis_easy_easy063.md|1777384241.0623188|3979": {"hash": "8aca6989a055de7fc0d4fa737941273c3244876216852cd70fe55cc3076bd85d", "size": 3979, "status": "c"}, "tasks/task_134_enterprise_industry_analysis_easy_easy064.md|1777384241.0623188|3995": {"hash": "d3775ada051923ccebf06c6005a5418d90c13f57305fb996b1ed4e8addf3c8fa", "size": 3995, "status": "c"}, "tasks/task_135_enterprise_industry_analysis_easy_easy065.md|1777384241.0623188|4076": {"hash": "82d70d0f878feb888773b8ae8a1658be64f43840fd7079e1eff315204707d856", "size": 4076, "status": "c"}, "tasks/task_136_enterprise_industry_analysis_easy_easy066.md|1777384241.0623188|4001": {"hash": "f0061006f34f473842781aadbec4b0e457e14029a799dcc1f645b301bec06a1d", "size": 4001, "status": "c"}, "tasks/task_137_enterprise_industry_analysis_easy_easy067.md|1777384241.0781844|4070": {"hash": "ae4df74f8cfcb22c25c16eeaeaf1c2995e4565c079462ac690c48758dddfa5c3", "size": 4070, "status": "c"}, "tasks/task_138_enterprise_industry_analysis_easy_easy068.md|1777384241.0781844|3912": {"hash": "3dbb286a3833cf404366d121fdc5fccc6e8b9ef12740f71c31bf8820873fb4f4", "size": 3912, "status": "c"}, "tasks/task_139_enterprise_industry_analysis_easy_easy069.md|1777384241.0781844|3887": {"hash": "f2ad48aa0cc7134553dc61683cdb30d63ab60b1445211285c290fa89db4e7ea9", "size": 3887, "status": "c"}, "tasks/task_140_enterprise_industry_analysis_easy_easy070.md|1777384241.0781844|4016": {"hash": "898a9d0aed63f536ba8f0553a0fa98372bd66deca745c2cc582a2abdd2ac47a1", "size": 4016, "status": "c"}, "tasks/task_141_enterprise_industry_analysis_easy_easy071.md|1777384241.0781844|3892": {"hash": "8eb7dda61222508284231745bdb9492b8d45dba6752a311244ae6961419e0e41", "size": 3892, "status": "c"}, "tasks/task_142_enterprise_industry_analysis_easy_easy072.md|1777384241.0781844|4024": {"hash": "f3bf4698b9bef4af3e7598f1ccd6993c2d8ecf0b82a2a10af041a16ef6a8c00e", "size": 4024, "status": "c"}, "tasks/task_143_enterprise_industry_analysis_easy_easy073.md|1777384241.0781844|4026": {"hash": "8ad0b541e668af4c3c627bd40f1cb1b89e47c3f5f7a81c60e8c10e426572358e", "size": 4026, "status": "c"}, "tasks/task_144_enterprise_industry_analysis_easy_easy074.md|1777384241.0781844|3900": {"hash": "df3bb71373b02852369957874f527ce4b0100830aad01725775b1fcc37e2f2cb", "size": 3900, "status": "c"}, "tasks/task_145_enterprise_industry_analysis_easy_easy075.md|1777384241.0781844|3984": {"hash": "e5497f5088f2f71aafae0fd51053f3b9c02a77c8f0fd626ccafb4d6675364644", "size": 3984, "status": "c"}, "tasks/task_147_enterprise_industry_analysis_easy_easy077.md|1777384241.0781844|4001": {"hash": "1af27d3c80f8982bcf10d57f67705b34c6e77e6ed574f87fe4d381f95e693bf0", "size": 4001, "status": "c"}, "tasks/task_146_enterprise_industry_analysis_easy_easy076.md|1777384241.0781844|3999": {"hash": "574f2c38837fac3c4f9da8a328ee9ee9befebae91b53e6f1456979a6f50b633b", "size": 3999, "status": "c"}, "tasks/task_148_enterprise_industry_analysis_easy_easy078.md|1777384241.0781844|4051": {"hash": "54a0395d860f833c2fd941a89c20f1fa9cb97f7d768b3290726829e7757a5359", "size": 4051, "status": "c"}, "tasks/task_149_enterprise_industry_analysis_easy_easy079.md|1777384241.0781844|3936": {"hash": "20daf2a8d17b094896cbb0c5985f6c391a77d174274026c589f6a4bee1e3ea6c", "size": 3936, "status": "c"}, "tasks/task_150_enterprise_industry_analysis_easy_easy080.md|1777384241.0781844|4041": {"hash": "2d1c72a8bde76c7e944c280849770b5cb7a755ace2a83261f85c49ac3da2d45b", "size": 4041, "status": "c"}, "tasks/task_151_enterprise_industry_analysis_easy_easy081.md|1777384241.0781844|3975": {"hash": "aee6061e88f8367e66740d79c73509032101d86710b56a549a0a7f1784b565d4", "size": 3975, "status": "c"}, "tasks/task_152_enterprise_industry_analysis_easy_easy082.md|1777384241.0781844|4027": {"hash": "a5ce0c798a30116edb634317a20e96da1e05a5980883af1e8b24dee7f1270202", "size": 4027, "status": "c"}, "tasks/task_153_enterprise_industry_analysis_easy_easy083.md|1777384241.0781844|3960": {"hash": "ea405186b88f9f9a29bc71fee68967690940248eb0fae0c574c8e4f4d6823ec0", "size": 3960, "status": "c"}, "tasks/task_154_enterprise_industry_analysis_easy_easy084.md|1777384241.0781844|4057": {"hash": "97b0373895a6756c3b1162061586a8c226c1372d783148d07a0b033fbf4749c4", "size": 4057, "status": "c"}, "tasks/task_155_enterprise_industry_analysis_easy_easy085.md|1777384241.0938413|3951": {"hash": "01440b56d9f86ce886cea505e6c541808e2d2e00da9453d636234267a77240ba", "size": 3951, "status": "c"}, "tasks/task_156_enterprise_industry_analysis_easy_easy086.md|1777384241.0938413|3978": {"hash": "9d6f099173555bab6ce70b49ec75fb124aa881789c679e17f9d76afaca4a3f69", "size": 3978, "status": "c"}, "tasks/task_157_enterprise_industry_analysis_easy_easy088.md|1777384241.0938413|4096": {"hash": "ae0d6b47142d2f6c58335daef9c7e107039d8f387a4908cdcdee08079fd71c96", "size": 4096, "status": "c"}, "tasks/task_158_enterprise_industry_analysis_easy_easy090.md|1777384241.0938413|4027": {"hash": "2a3900f60fd2d81d66c5b3ed5150d91407e03f6941e5cfe9969a877eb3e27165", "size": 4027, "status": "c"}, "tasks/task_159_enterprise_industry_analysis_easy_easy091.md|1777384241.0938413|3859": {"hash": "0e3d3811c32beae25985dd47441805e2b43fad096649b0f1d90c690c4531a7be", "size": 3859, "status": "c"}, "tasks/task_160_enterprise_industry_analysis_easy_easy092.md|1777384241.0938413|3862": {"hash": "0e6d4da680cd851b516a69bf827ddfbababdaa79562f3c067d5e2881a7607f42", "size": 3862, "status": "c"}, "tasks/task_161_enterprise_industry_analysis_easy_easy093.md|1777384241.0938413|3871": {"hash": "ba61bc21f04800dad14ed6ebb97228f41c626d6f336dec2158e108f6e0483a63", "size": 3871, "status": "c"}, "tasks/task_162_enterprise_industry_analysis_easy_easy094.md|1777384241.0938413|3870": {"hash": "26c6b05cd08321432fc8079b26742f2826790ccb4410f3ad6222c65967719118", "size": 3870, "status": "c"}, "tasks/task_163_enterprise_industry_analysis_easy_easy095.md|1777384241.0938413|3842": {"hash": "4b97d07e784bbe38718caddd6183a61aca9e9cff0269b85a258044bdbe657d7d", "size": 3842, "status": "c"}, "tasks/task_164_enterprise_industry_analysis_easy_easy096.md|1777384241.0938413|3868": {"hash": "81f89726c117db1cd08f0ee40fe5b6fea1a8c91dec184a309aa1c482720bc63f", "size": 3868, "status": "c"}, "tasks/task_165_enterprise_industry_analysis_easy_easy097.md|1777384241.0938413|3880": {"hash": "181195825354eb3d02a4afe2ec820435d8f8ef69dd4ece4533f5697507ad6e25", "size": 3880, "status": "c"}, "tasks/task_166_enterprise_industry_analysis_easy_easy098.md|1777384241.0938413|3868": {"hash": "04c27bafae23b1ab72fe8896f2ba26f55c76fddfa5c763926f15c69b1a5f29f4", "size": 3868, "status": "c"}, "tasks/task_167_enterprise_industry_analysis_easy_easy099.md|1777384241.0938413|3903": {"hash": "2f114cf19f179d72926edc3d8d1b0c91e98a70eae043d4eca99ef4e3f2007cc2", "size": 3903, "status": "c"}, "tasks/task_168_enterprise_industry_analysis_easy_easy100.md|1777384241.1089647|3923": {"hash": "4b93e50bad0b81e4d1fe1b0d51e3b59ac442782cc8147761cda3a773108d66c7", "size": 3923, "status": "c"}, "tasks/task_169_enterprise_industry_analysis_easy_easy101.md|1777384241.1099725|3857": {"hash": "0d42d8f4ed95f846d257e0c184a39169b74047aed2f5292ebdd347cd7ac4d425", "size": 3857, "status": "c"}, "tasks/task_170_enterprise_industry_analysis_easy_easy102.md|1777384241.1099725|3863": {"hash": "469e1190625c9fa3b2ac4d6b5d1d3a08fa55822626e56e2feb82bd27449f66d7", "size": 3863, "status": "c"}, "tasks/task_171_enterprise_industry_analysis_easy_easy103.md|1777384241.1099725|3853": {"hash": "c4ed43d4760babe94762dd0c73ab93f0caab1ff84cac6ed919d67d8bfeca6796", "size": 3853, "status": "c"}, "tasks/task_172_enterprise_industry_analysis_easy_easy104.md|1777384241.1099725|3851": {"hash": "6c41fd709556f853b93d14737b5bb089c5cb59b885bb1f8657b040b1d667ff15", "size": 3851, "status": "c"}, "tasks/task_173_enterprise_industry_analysis_easy_easy105.md|1777384241.1099725|3830": {"hash": "c5b13af081bd80ff14ce8c11cd59aa52e07dd9bf9042086783d04337d45a5c1c", "size": 3830, "status": "c"}, "tasks/task_174_enterprise_industry_analysis_easy_easy106.md|1777384241.11395|3932": {"hash": "52b464f6a43e165be469eb4cf9f92052cb808ea32ca80b5095618ec4bdffc93b", "size": 3932, "status": "c"}, "tasks/task_175_enterprise_industry_analysis_easy_easy107.md|1777384241.11395|3852": {"hash": "8ba65f92742f86b0fc507d2f6b86e5ccbc93ddbf41e12432eaff297bd9034e91", "size": 3852, "status": "c"}, "tasks/task_176_enterprise_industry_analysis_easy_easy108.md|1777384241.1159601|3832": {"hash": "04f05da92b577218e6f234bc138fc07c00ed4c4d78667d9282499b72546dcf7b", "size": 3832, "status": "c"}, "tasks/task_177_enterprise_industry_analysis_easy_easy109.md|1777384241.1159601|3856": {"hash": "64d9a9566b4a9c3e3d39ce56b26df919b8d26c1f984a5b10bc3035abe73161d9", "size": 3856, "status": "c"}, "tasks/task_178_enterprise_industry_analysis_easy_easy110.md|1777384241.1159601|3836": {"hash": "b6c0ad57d751dfb5bd4610f4ea25085b9c34590d5d1613ffe3ef6bc5832613bf", "size": 3836, "status": "c"}, "tasks/task_179_enterprise_industry_analysis_easy_easy111.md|1777384241.1185093|3829": {"hash": "9d7147e4e698aa785112ca64c18bc026a8164207243034e8c56014d1077e0b83", "size": 3829, "status": "c"}, "tasks/task_180_enterprise_industry_analysis_easy_easy112.md|1777384241.1185093|3867": {"hash": "47aa986b13810405268d849439a07994aec8eb3a77891cbe9c95ffe45a2bbd19", "size": 3867, "status": "c"}, "tasks/task_181_enterprise_industry_analysis_easy_easy113.md|1777384241.1185093|3854": {"hash": "d4df3c888c24ed4c612036bf77a05a4c907bcff637b626452dbe9f8719c68768", "size": 3854, "status": "c"}, "tasks/task_182_enterprise_industry_analysis_easy_easy114.md|1777384241.1185093|3861": {"hash": "f0f35c196431649a284aaafb671d4ff08699e7ecbde1dddfb414b5f748c4e898", "size": 3861, "status": "c"}, "tasks/task_183_enterprise_industry_analysis_easy_easy115.md|1777384241.1223948|3831": {"hash": "6567d97cf535268f8d9616730028579cb755ddc2241b27a331ebd8d1322b8c5e", "size": 3831, "status": "c"}, "tasks/task_184_enterprise_industry_analysis_easy_easy116.md|1777384241.1223948|3847": {"hash": "af965d972588fd1a681d3e39e702150b51a397f3d15e21bb14c91eb2180438b2", "size": 3847, "status": "c"}, "tasks/task_185_enterprise_industry_analysis_easy_easy117.md|1777384241.1223948|3817": {"hash": "69306bb9c50afbacba05882cc0f241a42b7aa3a6a0e5a9aa4f5e9e8d8e700fd5", "size": 3817, "status": "c"}, "tasks/task_186_enterprise_industry_analysis_medium_medium001.md|1777384241.125702|3960": {"hash": "612e0d6c77a54ad8caf823493d1e3afbc83812de93826e345a65ee2a8828b5ab", "size": 3960, "status": "c"}, "tasks/task_187_enterprise_industry_analysis_medium_medium002.md|1777384241.125702|3971": {"hash": "b688072761f9001f3c632e6cf1b77f18c13d59f8874c920115b94a49016bb826", "size": 3971, "status": "c"}, "tasks/task_188_enterprise_industry_analysis_medium_medium003.md|1777384241.125702|3931": {"hash": "234cbb847ecdfea9d6b1a3016b699e3b5a6ea17a57d4984fb2abc27b59b8a304", "size": 3931, "status": "c"}, "tasks/task_189_enterprise_industry_analysis_medium_medium004.md|1777384241.125702|3974": {"hash": "7ee81e14bc643254b312d5f7408ba50a2b8e32c9587217e186b75a2be65b9287", "size": 3974, "status": "c"}, "tasks/task_190_enterprise_industry_analysis_medium_medium005.md|1777384241.125702|3994": {"hash": "1c53ccdec53418e4b5603f83f0466647b2fc686ced373c3e049d9810d144a371", "size": 3994, "status": "c"}, "tasks/task_191_enterprise_industry_analysis_medium_medium006.md|1777384241.125702|3912": {"hash": "5afe4afa7032833f8407c0282bc410c0677a7861e688b97b71ee03493830a3ff", "size": 3912, "status": "c"}, "tasks/task_192_enterprise_industry_analysis_medium_medium007.md|1777384241.125702|3988": {"hash": "85da2b4775e975942e096bd3046f47b5d56ea622a22af83c6831bb5ecd4d0a09", "size": 3988, "status": "c"}, "tasks/task_193_enterprise_industry_analysis_medium_medium008.md|1777384241.1332972|4022": {"hash": "f8ad662a3a5db975ec1aa23813ebbe260262c1bdef318b7899e292ceccf59ed5", "size": 4022, "status": "c"}, "tasks/task_194_enterprise_industry_analysis_medium_medium009.md|1777384241.1332972|3986": {"hash": "883c9ac7674ea27cfa5deea13a5908624808d984fe15b712d875601a24bb7ef0", "size": 3986, "status": "c"}, "tasks/task_195_enterprise_industry_analysis_medium_medium010.md|1777384241.1332972|3932": {"hash": "1cb6f1a384d3a5b59d9f76b5a5ef77b263577c737e7cab6f0e86bb832d799f12", "size": 3932, "status": "c"}, "tasks/task_196_enterprise_industry_analysis_medium_medium011.md|1777384241.135305|4020": {"hash": "c83d58b4cf7267a67b637f19804ced597e6705bf73f28ca72a19605412125f40", "size": 4020, "status": "c"}, "tasks/task_197_enterprise_industry_analysis_medium_medium012.md|1777384241.135305|3926": {"hash": "833ba1519f3101c9f22795313c51baafafe98bb458281a2195d0d9bd0f0e14d5", "size": 3926, "status": "c"}, "tasks/task_198_enterprise_industry_analysis_medium_medium013.md|1777384241.135305|3916": {"hash": "230b9c3d3a48ab4b0e3effc1ef5bb5f2f9ecaff23ffb50e76586c00f1ef38954", "size": 3916, "status": "c"}, "tasks/task_199_enterprise_industry_analysis_medium_medium014.md|1777384241.1373253|4062": {"hash": "051b4a17a75fee24c49d6e259c1cc66549da87321a5eeccb3040acbd3a77a8fc", "size": 4062, "status": "c"}, "tasks/task_200_enterprise_industry_analysis_medium_medium015.md|1777384241.1373253|3958": {"hash": "8a54b7cf72d5f16f7342aea2482a9d765ab0288f0a84684aad5833699ade9a52", "size": 3958, "status": "c"}, "tasks/task_201_enterprise_industry_analysis_medium_medium016.md|1777384241.1393342|4061": {"hash": "3cc81e7c9b70479d78acf13a3c896d417190038b3b8cd5af2602e6a83a72d2e7", "size": 4061, "status": "c"}, "tasks/task_202_enterprise_industry_analysis_medium_medium017.md|1777384241.1393342|4028": {"hash": "2f420b3c07a7d7d7b46d66bd09f57798f6184d26a30094b542bb43918e2b3e00", "size": 4028, "status": "c"}, "tasks/task_203_enterprise_industry_analysis_medium_medium018.md|1777384241.1413405|4116": {"hash": "a1d57473d5c46baacb6dcf8063319b7deaeb15fb3a55f98b24ed2c0d3b961854", "size": 4116, "status": "c"}, "tasks/task_204_enterprise_industry_analysis_medium_medium019.md|1777384241.1413405|4099": {"hash": "a20b84129cc2c19f20b56e6befc7b1f339a295da473b73d49aa008e4fb977584", "size": 4099, "status": "c"}, "tasks/task_205_enterprise_industry_analysis_medium_medium020.md|1777384241.1413405|4083": {"hash": "c8ad8f5d1b0d66ca166f596b256e8552d40ba0d2d201df6c80c115ac7b4aaf5f", "size": 4083, "status": "c"}, "tasks/task_206_enterprise_industry_analysis_medium_medium021.md|1777384241.1413405|4005": {"hash": "3607047b29e755be78936ac7dbe7c3604522a8310c2e2ba4efe1f33994533ff2", "size": 4005, "status": "c"}, "tasks/task_207_enterprise_industry_analysis_medium_medium022.md|1777384241.1413405|4068": {"hash": "1ac0d6fe0c4189ea7cb6c1172c91effcbf8d70f5008dc85048af7bfe7af20428", "size": 4068, "status": "c"}, "tasks/task_208_enterprise_industry_analysis_medium_medium023.md|1777384241.1413405|4078": {"hash": "4f11d62a846e5012efa047992e45794c0562722c3012143b20eba0cf9518cb52", "size": 4078, "status": "c"}, "tasks/task_209_enterprise_industry_analysis_medium_medium024.md|1777384241.1474755|4016": {"hash": "ec5ed2647ca78f599560c78266ca18f69411fa4817a0dbd5b942ccb54f282ee3", "size": 4016, "status": "c"}, "tasks/task_210_enterprise_industry_analysis_medium_medium025.md|1777384241.1474755|3963": {"hash": "9f6e21d5bcbb27af32411d20df1605ee12d927e160b6c8df633dc45bf7a4e777", "size": 3963, "status": "c"}, "tasks/task_211_enterprise_industry_analysis_medium_medium026.md|1777384241.1474755|3952": {"hash": "9344c1f2d03536a1cb362e754a0b9bf4f5cc9eda82f8be931045278a20ca6086", "size": 3952, "status": "c"}, "tasks/task_212_enterprise_industry_analysis_medium_medium027.md|1777384241.1494846|3978": {"hash": "37b480a70b02165e7999bdbf708d964229a4091f441bb4515f096e14c6cccc3c", "size": 3978, "status": "c"}, "tasks/task_213_enterprise_industry_analysis_medium_medium028.md|1777384241.1494846|3959": {"hash": "62910ac96049ff1fcf8923893b43f12014ad9d6868da0db07ab1108c6eecf467", "size": 3959, "status": "c"}, "tasks/task_214_enterprise_industry_analysis_medium_medium029.md|1777384241.1514907|4048": {"hash": "abdca9074cae1e37b71df5bad7355431d4526814b3fbd048798aabee62ff72c1", "size": 4048, "status": "c"}, "tasks/task_215_enterprise_industry_analysis_medium_medium030.md|1777384241.1514907|4005": {"hash": "ae90fab51ec09b23e88d3b075cf0757ee974ac504c510efe9b768522a3ce7238", "size": 4005, "status": "c"}, "tasks/task_216_enterprise_industry_analysis_medium_medium031.md|1777384241.153497|4068": {"hash": "1417459590544da7df63a8f1b7600e8ec24873c0ed038f23147a8f563915f103", "size": 4068, "status": "c"}, "tasks/task_217_enterprise_industry_analysis_medium_medium032.md|1777384241.153497|4068": {"hash": "5f107dcb833e97964764c34c685db8cf332fab2ab8f659f8b94379e8064b0733", "size": 4068, "status": "c"}, "tasks/task_218_enterprise_industry_analysis_medium_medium033.md|1777384241.153497|4096": {"hash": "af50991172ae60ebad8617599c17dbeeb709a28341bbd1e5c456a4cd46166316", "size": 4096, "status": "c"}, "tasks/task_219_enterprise_industry_analysis_medium_medium034.md|1777384241.1555068|3967": {"hash": "0ef849f67864f8fce93f6eaf1f52ed566cdfee80d01feda46782c8688869ce1d", "size": 3967, "status": "c"}, "tasks/task_220_enterprise_industry_analysis_medium_medium035.md|1777384241.1555068|4119": {"hash": "5cc70c282ea24ef113db368d4c973cc30a6027020c8563447e9f25eda156d7bf", "size": 4119, "status": "c"}, "tasks/task_221_enterprise_industry_analysis_medium_medium036.md|1777384241.157257|3995": {"hash": "c57aba6e8618b3cdb5854774cd8809a284c0da0bd8f4e5460c820e49e6ad3e7c", "size": 3995, "status": "c"}, "tasks/task_222_enterprise_industry_analysis_medium_medium037.md|1777384241.157257|4248": {"hash": "9e64fd75ceb0420bc3e92a2b5585b13e110673d6972f75f2ade109bf82a10ba6", "size": 4248, "status": "c"}, "tasks/task_223_enterprise_industry_analysis_medium_medium038.md|1777384241.157257|4042": {"hash": "5c079b15970c1247ad667e567fd878de3a4a5e6ed61de64ff5175f99ba2ef01b", "size": 4042, "status": "c"}, "tasks/task_224_enterprise_industry_analysis_medium_medium039.md|1777384241.157257|4071": {"hash": "a22933688b29624fa67b66d7a3d9ba442092445eca4fc4b1fc031a69c5bd6c7d", "size": 4071, "status": "c"}, "tasks/task_225_enterprise_industry_analysis_medium_medium040.md|1777384241.157257|4079": {"hash": "46965c23dc767a4d2ca01ab97c3b7d9fe22e7db2a8596432cfdf3e12a3566e1a", "size": 4079, "status": "c"}, "tasks/task_226_enterprise_industry_analysis_medium_medium041.md|1777384241.157257|4073": {"hash": "996b7724b89cf06a50bad7a21af0b6681a65d22c4df9866069511ebb3012f04d", "size": 4073, "status": "c"}, "tasks/task_227_enterprise_industry_analysis_medium_medium042.md|1777384241.157257|4256": {"hash": "84809f7ac2089e6a5afaa9225deeeefc4f41f8008d6b3422de0344d62470859e", "size": 4256, "status": "c"}, "tasks/task_228_enterprise_industry_analysis_medium_medium043.md|1777384241.157257|4059": {"hash": "6f2369321895f48f74d35deed08cd0e9661b6b1fe03001d6c6046a2d3bcba766", "size": 4059, "status": "c"}, "tasks/task_229_enterprise_industry_analysis_medium_medium044.md|1777384241.157257|4115": {"hash": "5d05bc444e5803762bddff177ac422fef9d92e0386ad1646c729a199109f01d7", "size": 4115, "status": "c"}, "tasks/task_230_enterprise_industry_analysis_medium_medium045.md|1777384241.157257|4053": {"hash": "cdb3d94bb9da6f8833590ab8edf755dbdecdf6ff522d082e0c07cc5749bd2e5b", "size": 4053, "status": "c"}, "tasks/task_231_enterprise_industry_analysis_medium_medium046.md|1777384241.157257|4057": {"hash": "0436027294d1df1300a52c0c46534318db9b34ca66f2096198593ec66adb0500", "size": 4057, "status": "c"}, "tasks/task_232_enterprise_industry_analysis_medium_medium047.md|1777384241.157257|4081": {"hash": "4b70431306f69e10cbc7ea98def3eed13e63acef94b220165ed307d638767c80", "size": 4081, "status": "c"}, "tasks/task_233_enterprise_industry_analysis_medium_medium048.md|1777384241.157257|4132": {"hash": "c3347a8d3bcf6207d3cb3cf8488d593e51d6245eb39e2b00e3316c05e6f4ba84", "size": 4132, "status": "c"}, "tasks/task_234_enterprise_industry_analysis_medium_medium049.md|1777384241.157257|4071": {"hash": "f8b0d39718b22679dffc69f1d8b25e2ae5e29a452c2f58a3623327fae71cac81", "size": 4071, "status": "c"}, "tasks/task_235_enterprise_industry_analysis_medium_medium050.md|1777384241.157257|4083": {"hash": "9fa5a6ea2f782231612f31b0e72ee0cadd677808c5342bc1074f6f8015b28954", "size": 4083, "status": "c"}, "tasks/task_236_enterprise_industry_analysis_medium_medium051.md|1777384241.157257|4054": {"hash": "16533fd1bdf734aa458110504b789f99e4866a7834466a0195d629982c4849f7", "size": 4054, "status": "c"}, "tasks/task_237_enterprise_industry_analysis_medium_medium052.md|1777384241.1732392|4095": {"hash": "5af052d6218939d2a905497235ad1fb29e56ef7241f00569795e7fcfd97f83f5", "size": 4095, "status": "c"}, "tasks/task_238_enterprise_industry_analysis_medium_medium053.md|1777384241.1732392|4129": {"hash": "0e73da81e75d281a4b0ded6aa87992399e9d26b3c94e87d9fbdfb7786474657d", "size": 4129, "status": "c"}, "tasks/task_239_enterprise_industry_analysis_medium_medium054.md|1777384241.1732392|4125": {"hash": "3c320ac0685ecbeca73cd1be84560fa6754c03c7eb4583e51171a4dfb2db3921", "size": 4125, "status": "c"}, "tasks/task_240_enterprise_industry_analysis_medium_medium055.md|1777384241.1732392|4123": {"hash": "3489be47890033d9f3fdf931c48f2c9e4032655f2efb0a857e2a56947306e536", "size": 4123, "status": "c"}, "tasks/task_241_enterprise_industry_analysis_medium_medium056.md|1777384241.1732392|4055": {"hash": "37f907af58b0b7d60afef78874d019017dc17544ceac5a58679a5cfcf8d2711c", "size": 4055, "status": "c"}, "tasks/task_242_enterprise_industry_analysis_medium_medium057.md|1777384241.1732392|4020": {"hash": "0c7b08406c9925d6a93e34fa387cb81aee675deedaf26d590aec548ba4acb2cb", "size": 4020, "status": "c"}, "tasks/task_243_enterprise_industry_analysis_medium_medium058.md|1777384241.1732392|4078": {"hash": "2a02e51cbbb0e60b7897d8cfa2ece759fd8baaacd5c8b358476f9cc4d4bba338", "size": 4078, "status": "c"}, "tasks/task_244_enterprise_industry_analysis_medium_medium059.md|1777384241.180854|4134": {"hash": "b5a37ee6f6b88eb072f255bae35d1b1c2e6b75ee32a622279a5941ef3e33b3ef", "size": 4134, "status": "c"}, "tasks/task_245_enterprise_industry_analysis_medium_medium060.md|1777384241.180854|4025": {"hash": "7e391c2455f7878ac34fd250278e52f77b99bf7cdf1dfabbca31e1b6d60f9bc5", "size": 4025, "status": "c"}, "tasks/task_246_enterprise_industry_analysis_medium_medium061.md|1777384241.1828637|4019": {"hash": "636da7236ff18e308193860ca4083156ac2d137435960084ecc486e2d4e2d5e8", "size": 4019, "status": "c"}, "tasks/task_247_enterprise_industry_analysis_medium_medium062.md|1777384241.1828637|3941": {"hash": "9f5e93f8c7d208fe81336848e4748bb065a2d7167e17d7d5db00c114547869e6", "size": 3941, "status": "c"}, "tasks/task_248_enterprise_industry_analysis_medium_medium063.md|1777384241.184874|4148": {"hash": "eaf9c5847c069e24051230d624ff22f65348a6c5a5cf7a888305373862559ac5", "size": 4148, "status": "c"}, "tasks/task_249_enterprise_industry_analysis_medium_medium064.md|1777384241.184874|4039": {"hash": "9e0ed549bbee3048426c9fbc14fba8664096f99a6c078921ff11f1decbc3999d", "size": 4039, "status": "c"}, "tasks/task_250_enterprise_industry_analysis_medium_medium065.md|1777384241.1868834|4443": {"hash": "0bedee61e7c27b2290bce81ef891c8b2a4a8f94d72dc0b881adf0f5722f40098", "size": 4443, "status": "c"}, "tasks/task_251_enterprise_industry_analysis_medium_medium066.md|1777384241.1868834|4231": {"hash": "a411327b716d50d9c1ec199105e737d8c08524103c52a0582741c5abe1cbdb94", "size": 4231, "status": "c"}, "tasks/task_252_enterprise_industry_analysis_medium_medium067.md|1777384241.1888921|4428": {"hash": "52de607169c5a936ddba69066606464bc6dea73fb644f604e26b2f1b6762753c", "size": 4428, "status": "c"}, "tasks/task_253_enterprise_industry_analysis_medium_medium068.md|1777384241.1888921|4098": {"hash": "a0adce5764fc849115d425701e01f634b2718c8fa3bbced09259a3703d205719", "size": 4098, "status": "c"}, "tasks/task_254_enterprise_industry_analysis_medium_medium069.md|1777384241.1909003|4077": {"hash": "add642ac2812f97934b2377732d85689f961ba55c4369f9ca5a4d9f1f64226c2", "size": 4077, "status": "c"}, "tasks/task_255_enterprise_industry_analysis_medium_medium070.md|1777384241.1909003|4123": {"hash": "586828870c39350d3b2b1996aff19573a3e682a3a6d4552290fef0da200b63b5", "size": 4123, "status": "c"}, "tasks/task_256_enterprise_industry_analysis_medium_medium071.md|1777384241.1929078|4368": {"hash": "4eae6051d037d9f2bfd6be6f5721403fcbc307ec21239ddafa1b2f2f8b363c33", "size": 4368, "status": "c"}, "tasks/task_257_enterprise_industry_analysis_medium_medium072.md|1777384241.1929078|4388": {"hash": "6d64849934d832edd8da5b0648986f7e80761b5b0b0f11e0a7e7e2c938b1789f", "size": 4388, "status": "c"}, "tasks/task_258_enterprise_industry_analysis_medium_medium073.md|1777384241.1949158|4031": {"hash": "2f6c5ffd10e4432657db9d24c9857a58a629bea2b1d6937b88da4afa5fe2ecd0", "size": 4031, "status": "c"}, "tasks/task_260_enterprise_industry_analysis_medium_medium075.md|1777384241.1949158|4030": {"hash": "bc27b1db6bfcbc58439fb75aed2a637639d7cf5cd2c2f5e4f950afff92cee01a", "size": 4030, "status": "c"}, "tasks/task_259_enterprise_industry_analysis_medium_medium074.md|1777384241.1949158|4042": {"hash": "a7ab439795cd4c86c672aada6e7e554fb14a24614665d6ee320614fa3eb0569f", "size": 4042, "status": "c"}, "tasks/task_261_enterprise_industry_analysis_medium_medium076.md|1777384241.196923|4014": {"hash": "2c0a3db1b0aa59776ae98914f04592d44096d2dc5753abf46a86c15f2ee4f265", "size": 4014, "status": "c"}, "tasks/task_262_enterprise_industry_analysis_medium_medium077.md|1777384241.196923|4238": {"hash": "921e851f62bd33d2797223f95cac459bd990093794d44c73b60bbb067ed304da", "size": 4238, "status": "c"}, "tasks/task_263_enterprise_industry_analysis_medium_medium078.md|1777384241.1989324|4053": {"hash": "9731cdd6076d237aa1c000178e80c872e1780442a8fe8dad8426d2b22128bc62", "size": 4053, "status": "c"}, "tasks/task_264_enterprise_industry_analysis_medium_medium079.md|1777384241.1989324|4094": {"hash": "646046d96ceb8da1fdf16cc578fb61ca06da58b08e44607d2773008042f9e41d", "size": 4094, "status": "c"}, "tasks/task_265_enterprise_industry_analysis_medium_medium080.md|1777384241.1989324|3999": {"hash": "97bbca0b64093d4fb6938ad8b8fffe4eab838283cf758bda8399f48f4d325887", "size": 3999, "status": "c"}, "tasks/task_266_enterprise_industry_analysis_medium_medium081.md|1777384241.2009397|4055": {"hash": "64f0c7d559002aa655912bac432d1764d21418f95fd2b21fde787588cb60d69c", "size": 4055, "status": "c"}, "tasks/task_267_enterprise_industry_analysis_medium_medium082.md|1777384241.2009397|4034": {"hash": "e2f59bece29742fd9eac8d22cc12dafe555f33dfe822d864e786390cb63c07b7", "size": 4034, "status": "c"}, "tasks/task_268_enterprise_industry_analysis_medium_medium083.md|1777384241.2029467|4055": {"hash": "fe79e7392992e5b56eddaaa7ab43410b291ee8c75937a68fe6a6a535b658f41a", "size": 4055, "status": "c"}, "tasks/task_269_enterprise_industry_analysis_medium_medium084.md|1777384241.2029467|4305": {"hash": "ad90bba938071a4f62f2816fd0243526eeda9b6bfc65e13ac4234995ebe0efbc", "size": 4305, "status": "c"}, "tasks/task_270_enterprise_industry_analysis_medium_medium085.md|1777384241.2046947|4202": {"hash": "65c15e40c9659d03947394f88fbcae0a498ca43d0e148faad6a0a79c50754815", "size": 4202, "status": "c"}, "tasks/task_271_enterprise_industry_analysis_medium_medium086.md|1777384241.2046947|4351": {"hash": "d5261b23e2688602dad2e7580bde416648f41c7442d4a051d4df81381641b6dc", "size": 4351, "status": "c"}, "tasks/task_272_enterprise_industry_analysis_medium_medium087.md|1777384241.206704|4054": {"hash": "280fe99f7def68ae5c3b6f3fb14b30eb374fab04b26764d6dbd671e4fecfd36f", "size": 4054, "status": "c"}, "tasks/task_273_enterprise_industry_analysis_medium_medium088.md|1777384241.206704|4156": {"hash": "0a24c880a9e76625d237ff72b35dae7e7117735f13ca63e2addfd9570fd9e266", "size": 4156, "status": "c"}, "tasks/task_274_enterprise_industry_analysis_medium_medium089.md|1777384241.208712|4039": {"hash": "83a9f3b00dd176a91f6cdbbdd7a53e4e3eee6590a8f235f693d7ccc4e3d50213", "size": 4039, "status": "c"}, "tasks/task_275_enterprise_industry_analysis_medium_medium090.md|1777384241.208712|4086": {"hash": "9bc0467854cb881d48c29add940414947f494855ae8cb5fedeae257ac2ad4c41", "size": 4086, "status": "c"}, "tasks/task_276_enterprise_industry_analysis_medium_medium091.md|1777384241.208712|4055": {"hash": "62ca7eca26c1b87973205fe0aaed0fec6c79d6d036a0eacabbc7658006d3afc0", "size": 4055, "status": "c"}, "tasks/task_277_enterprise_industry_analysis_medium_medium092.md|1777384241.2107193|4008": {"hash": "fdf44862e853649c750627c99f1d27d3f8e5cc26357a4694b0062cbfa7695455", "size": 4008, "status": "c"}, "tasks/task_278_enterprise_industry_analysis_medium_medium093.md|1777384241.2107193|4020": {"hash": "5038497c3d82a3002261c78b8141599faad7139a694295f5ec507d6c8b8c3602", "size": 4020, "status": "c"}, "tasks/task_279_enterprise_industry_analysis_medium_medium094.md|1777384241.2107193|4021": {"hash": "7bd52d38034fd5348beada3f20e68cf4ba117ff291a7918901d95d54bc32fb2e", "size": 4021, "status": "c"}, "tasks/task_280_enterprise_industry_analysis_medium_medium095.md|1777384241.2107193|4058": {"hash": "c01c1d681dc2cb4d46cb29d372eca9f4633b9c54bd808b0553486b5c2d0cbd4f", "size": 4058, "status": "c"}, "tasks/task_281_enterprise_industry_analysis_medium_medium096.md|1777384241.2107193|4101": {"hash": "6b3acdc497354818cb32cd27e3d0db9f568bf037884489044d54c761f3e5384c", "size": 4101, "status": "c"}, "tasks/task_282_enterprise_industry_analysis_medium_medium097.md|1777384241.2107193|4216": {"hash": "9823bf3031a2d3dd57b1ac0229d63bcf72d61ad78245438eaa019c147f37d92c", "size": 4216, "status": "c"}, "tasks/task_283_enterprise_industry_analysis_medium_medium098.md|1777384241.2107193|4022": {"hash": "9e56352193c7bc69aaa086d2d8f41fe496fde84774f2f36744387d8ac3041f53", "size": 4022, "status": "c"}, "tasks/task_284_enterprise_industry_analysis_medium_medium099.md|1777384241.2107193|3968": {"hash": "01b11a7f49cd9d618b2f62977ed710fc8cde5ba067e285e35609aceae52d2db7", "size": 3968, "status": "c"}, "tasks/task_285_enterprise_industry_analysis_medium_medium100.md|1777384241.2107193|4019": {"hash": "fb781007ec22fc560c08c6610eb7bc654e7631a26b4bc1941cfdceeb47123c46", "size": 4019, "status": "c"}, "tasks/task_286_enterprise_industry_analysis_medium_medium101.md|1777384241.2107193|4084": {"hash": "225061667f22006eed08a38cebedc4aacd48f1fa4564289687b697eca9bc387b", "size": 4084, "status": "c"}, "tasks/task_287_enterprise_industry_analysis_medium_medium102.md|1777384241.2107193|4369": {"hash": "de1681f749ec4a0e2ab761ccafdf39d5399e4c967abbdb9cd96ad587a784f52c", "size": 4369, "status": "c"}, "tasks/task_288_enterprise_industry_analysis_medium_medium103.md|1777384241.2107193|4348": {"hash": "359891e5e9bdbde749133b650db91dc4575b5751998aee70da90a38c0e23d6c0", "size": 4348, "status": "c"}, "tasks/task_289_enterprise_industry_analysis_medium_medium104.md|1777384241.220356|4310": {"hash": "25934c8e755aa3ddb03561f7aa35abff930eb57066f9ba87295362772374a780", "size": 4310, "status": "c"}, "tasks/task_290_enterprise_industry_analysis_medium_medium105.md|1777384241.220356|4090": {"hash": "43eeca9bb6a8834f86d585f48792f172f077472ca85371d44c7e15d19289c385", "size": 4090, "status": "c"}, "tasks/task_291_enterprise_industry_analysis_medium_medium106.md|1777384241.220356|4085": {"hash": "33d6059bc8b915fcb7719ce4b6bf486409fc782ea8a30a780f6a92805046f6a5", "size": 4085, "status": "c"}, "tasks/task_292_enterprise_industry_analysis_medium_medium107.md|1777384241.220356|4074": {"hash": "e7828f934a9f93196a069f02bf1a718595397557d24f2d6febeb49a8c2a749c0", "size": 4074, "status": "c"}, "tasks/task_293_enterprise_industry_analysis_medium_medium108.md|1777384241.220356|4335": {"hash": "4e7c273640f6220530225a606ca1c4e9aee176d191fca4f402a42cdf5ad50cff", "size": 4335, "status": "c"}, "tasks/task_294_enterprise_industry_analysis_medium_medium109.md|1777384241.220356|4471": {"hash": "10cf5bee512fca4b321666d89e75d92577fdadf29c7117f155dad28d1a4f28dd", "size": 4471, "status": "c"}, "tasks/task_295_enterprise_industry_analysis_medium_medium110.md|1777384241.220356|4014": {"hash": "57b3b5088767d2c208a90023bf932aca97556d49c7696726cf57cd045466a731", "size": 4014, "status": "c"}, "tasks/task_296_enterprise_industry_analysis_medium_medium111.md|1777384241.220356|4297": {"hash": "45a2f8d5e755c26edd4533b910844c098ebb8fb4304ca6f12999fbe435cca245", "size": 4297, "status": "c"}, "tasks/task_297_enterprise_industry_policy_analysis_easy_easy067.md|1777384241.220356|3939": {"hash": "e707ad397c2a861bbc34db3934cdb6131cac0596cf1df43cc6b16ae6d9506439", "size": 3939, "status": "c"}, "tasks/task_298_enterprise_industry_policy_analysis_easy_easy068.md|1777384241.220356|3909": {"hash": "0a791d824112b0585f722f5489eea9ebc0733832c1d1ccded628796184a4c8f2", "size": 3909, "status": "c"}, "tasks/task_299_enterprise_industry_policy_analysis_easy_easy069.md|1777384241.220356|4005": {"hash": "0193948589da878d1e781b18bb9f81f802becedffedae05c6d723acfb8e1671f", "size": 4005, "status": "c"}, "tasks/task_300_enterprise_industry_policy_analysis_easy_easy070.md|1777384241.220356|3941": {"hash": "8414947fad983df6b1d3e3bfbe6a8944b21d883b4e2d230fee2c3d32288722ca", "size": 3941, "status": "c"}, "tasks/task_301_enterprise_industry_policy_analysis_easy_easy071.md|1777384241.220356|3937": {"hash": "4305219a5091b29a057fa6536e309d236fc56985d0892535f698d61e7e495eb2", "size": 3937, "status": "c"}, "tasks/task_302_enterprise_industry_policy_analysis_easy_easy072.md|1777384241.220356|3945": {"hash": "858342b51e6d55ff3fe0d0c55b23a7ea3ef084a79ea89750d261f9220ca142d5", "size": 3945, "status": "c"}, "tasks/task_303_enterprise_industry_policy_analysis_easy_easy073.md|1777384241.220356|3917": {"hash": "2116ddaf54cdd4fdec9df48955f4b9a0dd788715c40b75e9b8cac1b80f27907c", "size": 3917, "status": "c"}, "tasks/task_304_enterprise_industry_policy_analysis_easy_easy074.md|1777384241.220356|3960": {"hash": "825f2a4658466e647611cef6e344aea0a777e41d19adfccd421e8024224b8d06", "size": 3960, "status": "c"}, "tasks/task_305_enterprise_industry_policy_analysis_easy_easy075.md|1777384241.220356|4062": {"hash": "3851b23914e1b8292503c83350d24e1133654102b450ebbdef0631a72babf7f5", "size": 4062, "status": "c"}, "tasks/task_306_enterprise_industry_policy_analysis_easy_easy076.md|1777384241.220356|3948": {"hash": "91934665d0416c7c29288f3b42925c7fec055375d907cd96f1e0dd24129d7ce3", "size": 3948, "status": "c"}, "tasks/task_307_enterprise_industry_policy_analysis_medium_medium001.md|1777384241.220356|4181": {"hash": "3adc4e29ac6fc8293a293d61d32666270d485a7333d04a43f7889e55c302309e", "size": 4181, "status": "c"}, "tasks/task_308_enterprise_industry_policy_analysis_medium_medium002.md|1777384241.220356|4180": {"hash": "7aca8bd66081ad8201229b56cc18863d62797fe7af84bf8851722b0d33d7b4a2", "size": 4180, "status": "c"}, "tasks/task_309_enterprise_industry_policy_analysis_medium_medium003.md|1777384241.220356|4184": {"hash": "bbb7a36bcc3ffd20ad736e3b5fc954bab194820fca1f05fdd74672c9e128bd7c", "size": 4184, "status": "c"}, "tasks/task_310_enterprise_industry_policy_analysis_medium_medium004.md|1777384241.220356|4211": {"hash": "e5a4e8817e316f83e5222bd54d6689a17743d28a4d2dafd998183f61a35b241e", "size": 4211, "status": "c"}, "tasks/task_311_enterprise_industry_policy_analysis_medium_medium005.md|1777384241.2364237|4156": {"hash": "b72ac4f4821879352005e8342c978b7313f2778489c7a48ee2a1fb3d9b6dc9f4", "size": 4156, "status": "c"}, "tasks/task_312_enterprise_industry_policy_analysis_medium_medium006.md|1777384241.2364237|4196": {"hash": "15e994a9686ef16a51ce0d50b176daba18ec10a795c28ffb7e11359a4f180d25", "size": 4196, "status": "c"}, "tasks/task_313_enterprise_industry_policy_analysis_medium_medium007.md|1777384241.2364237|4170": {"hash": "2acde666d6f92b99ab574349e0896e7bf0e7ccd38bc2dff83b320958c2217558", "size": 4170, "status": "c"}, "tasks/task_314_enterprise_industry_policy_analysis_medium_medium008.md|1777384241.2364237|4115": {"hash": "b54f3e5d7e4441b8e2a0fdea689a6473a2aac4f0903048f155de12a2ff08273d", "size": 4115, "status": "c"}, "tasks/task_315_enterprise_industry_policy_analysis_medium_medium009.md|1777384241.2364237|4548": {"hash": "fae9a345492e3dc102b1c550484b7614e3a1126450d55c1006f13fc5586a5f00", "size": 4548, "status": "c"}, "tasks/task_316_enterprise_industry_policy_analysis_medium_medium010.md|1777384241.2364237|4119": {"hash": "184b2dcabb1533b45e28f5c788b07a56297d87164c6df60247c7d4c8baa0ac1b", "size": 4119, "status": "c"}, "tasks/task_317_enterprise_industry_policy_analysis_medium_medium011.md|1777384241.2364237|4245": {"hash": "bc700c42d7bc87be14d3c47b87475bb66ab51823482be6dbf3b6ed421e1b73d5", "size": 4245, "status": "c"}, "tasks/task_318_enterprise_industry_policy_analysis_medium_medium012.md|1777384241.2364237|4220": {"hash": "e9aa6f03ce960bb8fb2761f5e108f1631bbd4832ff23e5578f64f95420cae7a9", "size": 4220, "status": "c"}, "tasks/task_319_enterprise_industry_policy_analysis_medium_medium013.md|1777384241.2364237|4244": {"hash": "e5e257675da8ac7cac526ce6684b2b9d0fc13c33a17ec3b37727d95b89044c50", "size": 4244, "status": "c"}, "tasks/task_320_enterprise_industry_policy_analysis_medium_medium014.md|1777384241.2364237|4259": {"hash": "91f681ebff3f442e1c86c14fdc480391fbd56dbb4c67e7dbf9a3fa72fec9a9ba", "size": 4259, "status": "c"}, "tasks/task_321_enterprise_industry_policy_analysis_medium_medium015.md|1777384241.2364237|4203": {"hash": "3aafab345ee5bf831c9d46fddffc200cdfd90c2c5d9c1213470e4cdf799a8f52", "size": 4203, "status": "c"}, "tasks/task_322_enterprise_industry_policy_analysis_medium_medium016.md|1777384241.2364237|4155": {"hash": "5527256fd73556217cfe7aee0d410fb06965ee35ea4ec13b70c56e26700ba6ca", "size": 4155, "status": "c"}, "tasks/task_323_enterprise_industry_policy_analysis_medium_medium017.md|1777384241.2364237|4242": {"hash": "bf9b0ac525d8c7938432346d9a9ecb5e3c34ea4f515779d6a0911477423e4858", "size": 4242, "status": "c"}, "tasks/task_324_enterprise_industry_policy_analysis_medium_medium018.md|1777384241.2364237|4237": {"hash": "3307f0dc66994295b18477c0d2d9251d9ab3063f99c00cd116c31823c1c5b74f", "size": 4237, "status": "c"}, "tasks/task_325_enterprise_industry_policy_analysis_medium_medium019.md|1777384241.2364237|4159": {"hash": "b498cae1ef23464c152399c6e907f1c5d5f9619fc63893266d23a2319b55fe1d", "size": 4159, "status": "c"}, "tasks/task_326_enterprise_industry_policy_analysis_medium_medium020.md|1777384241.2364237|4177": {"hash": "4618019a9edeed54148e51c262373c8d5669313da1872bc331687144d9df1ccd", "size": 4177, "status": "c"}, "tasks/task_327_enterprise_industry_policy_analysis_medium_medium021.md|1777384241.2364237|4073": {"hash": "74ec358897518d1d0ae1fdf012ee5e12beee584dd40abfe72f9e6a3293875cc1", "size": 4073, "status": "c"}, "tasks/task_328_enterprise_industry_policy_analysis_medium_medium022.md|1777384241.2364237|3957": {"hash": "bf19d8e8fdb3a3d1f2e86906ba3584626be1ea6d001705d97102f0a3fd9f8bb1", "size": 3957, "status": "c"}, "tasks/task_329_enterprise_industry_policy_analysis_medium_medium023.md|1777384241.2364237|4197": {"hash": "3a66ed12d974099b734cfb85c257db83ad0e0cd21f73a0c5f804f9446d45cbbe", "size": 4197, "status": "c"}, "tasks/task_330_enterprise_industry_policy_analysis_medium_medium024.md|1777384241.2364237|4250": {"hash": "1f376feb89c68e1dcaa376a730794de1c86d3134ad1c39992b28b69be302a33d", "size": 4250, "status": "c"}, "tasks/task_331_enterprise_industry_policy_analysis_medium_medium025.md|1777384241.2520964|4180": {"hash": "60f264c0dd3cbc3096b47ae1137a090f2ddd144ff006219e9c1d1725c53f7d63", "size": 4180, "status": "c"}, "tasks/task_332_enterprise_industry_policy_analysis_medium_medium026.md|1777384241.2520964|4257": {"hash": "e8506c12b6a0afcae27c7db168cd5ee5d3f41133e5164322073d80057c9c721f", "size": 4257, "status": "c"}, "tasks/task_333_enterprise_industry_policy_analysis_medium_medium027.md|1777384241.2520964|4237": {"hash": "cc140606bc1b51493f0428b5e69c2a0dbdd80d40648a7e2c19a69f1e46e730f9", "size": 4237, "status": "c"}, "tasks/task_334_enterprise_industry_policy_analysis_medium_medium028.md|1777384241.2520964|4222": {"hash": "0475c9f5208701c44466dc5bc89ed944211140c54df6d967d4c44869407a1aba", "size": 4222, "status": "c"}, "tasks/task_335_enterprise_industry_policy_analysis_medium_medium029.md|1777384241.2520964|4192": {"hash": "f946f7755a123f67cd3481fd392a767999919975a4e1c8bc4082bb4c604ba890", "size": 4192, "status": "c"}, "tasks/task_336_enterprise_industry_policy_analysis_medium_medium030.md|1777384241.2520964|4209": {"hash": "bc90078f951fab53b87487be4b72738706a3387da8e88b23435f0314a1f37174", "size": 4209, "status": "c"}, "tasks/task_337_enterprise_industry_policy_analysis_medium_medium031.md|1777384241.2520964|4248": {"hash": "4a0e3be6f62077ae7358f8920baf14503b8f234b4fafa04cc169b118ead7263a", "size": 4248, "status": "c"}, "tasks/task_338_enterprise_industry_policy_analysis_medium_medium032.md|1777384241.2520964|4094": {"hash": "1a4702a588f7b28fd712d784c48db85a3c4d48008f9d79a77bfbee68a7c079f2", "size": 4094, "status": "c"}, "tasks/task_339_enterprise_industry_policy_analysis_medium_medium033.md|1777384241.2520964|4283": {"hash": "a00fd26f9df9062333a56b52464d5466b4007732911cf1a1dc415f1006846cf2", "size": 4283, "status": "c"}, "tasks/task_340_enterprise_industry_policy_analysis_medium_medium034.md|1777384241.2520964|4551": {"hash": "8e0287035d2dc3615aa48b0842cc75cbe909172221d01e28832414e32dd1b79f", "size": 4551, "status": "c"}, "tasks/task_341_enterprise_industry_policy_analysis_medium_medium035.md|1777384241.2520964|4229": {"hash": "8aa12f398cdba4c4337a59db021028479f01401db27bc2cd0c3b781bba2ab26b", "size": 4229, "status": "c"}, "tasks/task_342_enterprise_industry_policy_analysis_medium_medium036.md|1777384241.2520964|4285": {"hash": "a755e269437a9647c7197dc0bcbf804f97b3f7c01105ddf46d94c41e4f03c117", "size": 4285, "status": "c"}, "tasks/task_343_enterprise_industry_policy_analysis_medium_medium037.md|1777384241.2520964|4542": {"hash": "cd3600a0943bbc04380de920fb3c4004c45fb5d69c988999449e0ad4a95095da", "size": 4542, "status": "c"}, "tasks/task_344_enterprise_industry_policy_analysis_medium_medium038.md|1777384241.2520964|4224": {"hash": "d8b09103c6852ea8db67392354d72edd3fad60966fe908bcd5366ebff144f8df", "size": 4224, "status": "c"}, "tasks/task_345_enterprise_industry_policy_analysis_medium_medium039.md|1777384241.2520964|4507": {"hash": "a0b65bea79191a2342522b7695d3184569ecb905c1be4c46bf3b405c556b361e", "size": 4507, "status": "c"}, "tasks/task_346_enterprise_industry_policy_analysis_medium_medium040.md|1777384241.2520964|4235": {"hash": "ff19153c6cd96d9edd79ad3a232dabe910c1d71eb27eefc911924aa6eab052bb", "size": 4235, "status": "c"}, "tasks/task_347_enterprise_industry_policy_analysis_medium_medium041.md|1777384241.2520964|4181": {"hash": "e2088000156a641bac3cb5f8d4e9511179b3f5b993432b7af72d88035a5a8145", "size": 4181, "status": "c"}, "tasks/task_348_enterprise_industry_policy_analysis_medium_medium042.md|1777384241.2520964|4222": {"hash": "2181a1d3f9a3a3573529c2f1ae2ef2e2ce17f8a373f52599e3772728952f7ae6", "size": 4222, "status": "c"}, "tasks/task_349_enterprise_industry_policy_analysis_medium_medium043.md|1777384241.2520964|4177": {"hash": "0e3045dced05dbb702bb561f3be888c46098eebece281527fe333d150f006fb1", "size": 4177, "status": "c"}, "tasks/task_350_enterprise_industry_policy_analysis_medium_medium044.md|1777384241.2520964|4291": {"hash": "ededa0996b7242a8538b9b0bb5b3eb4fd322ec136bc6296eb4583053007289d2", "size": 4291, "status": "c"}, "tasks/task_351_enterprise_industry_policy_analysis_medium_medium045.md|1777384241.2520964|4116": {"hash": "e7d6b7d37af7cc5fab02a34e73526cb408ee80f9382c0bbae266f841741d563c", "size": 4116, "status": "c"}, "tasks/task_352_enterprise_industry_policy_analysis_medium_medium046.md|1777384241.2677789|3999": {"hash": "7d7aa2f4201ca9847af15d287049ef52ee513d69647fafcf30bd4a45527fd4bb", "size": 3999, "status": "c"}, "tasks/task_353_enterprise_industry_policy_analysis_medium_medium047.md|1777384241.2677789|4227": {"hash": "c0802e7dd965dda58db89a135052b3655deb7b6dd458f70ce1daf24556cf1353", "size": 4227, "status": "c"}, "tasks/task_354_enterprise_industry_policy_analysis_medium_medium048.md|1777384241.2677789|4122": {"hash": "6e361ad2adc70d53d3467743a5b95a8c09b0f715290db6e358ecc70efe72c35f", "size": 4122, "status": "c"}, "tasks/task_355_enterprise_industry_policy_analysis_medium_medium049.md|1777384241.2677789|4176": {"hash": "17485af1a4f101cce6cfb84ec47c2c965ffaf128fd4f4aa06b494fcb88fd3000", "size": 4176, "status": "c"}, "tasks/task_356_enterprise_industry_policy_analysis_medium_medium050.md|1777384241.2677789|3939": {"hash": "c08910d4c841d0baca6364765a64f0d3ae6b7e5395911477432b6555169ac688", "size": 3939, "status": "c"}, "tasks/task_357_enterprise_industry_policy_analysis_medium_medium051.md|1777384241.2677789|3958": {"hash": "baa3716e4be0f380516d731ba2a8c82b070545796476f348dce542aa631fe897", "size": 3958, "status": "c"}, "tasks/task_358_enterprise_industry_policy_analysis_medium_medium052.md|1777384241.2677789|4109": {"hash": "be8fca4a814ec9907c660346d1c187d3d97ae30a3d93b9135f623278010454c8", "size": 4109, "status": "c"}, "tasks/task_359_enterprise_industry_policy_analysis_medium_medium053.md|1777384241.2677789|4285": {"hash": "4693806653379e9c33b4b4e50df4a80f7e77d0db69b8faa66c29633b2f09693f", "size": 4285, "status": "c"}, "tasks/task_360_enterprise_industry_policy_analysis_medium_medium054.md|1777384241.273728|3992": {"hash": "1e0fa73e5739eafa649a704c304cfa065e4466332e392cb675ef65afb392f783", "size": 3992, "status": "c"}, "tasks/task_361_enterprise_industry_policy_analysis_medium_medium055.md|1777384241.273728|4192": {"hash": "dd2b5ba9a229dfefa41708c9b7cc03f5afe6a0f58ae527e11a649d857f7558c2", "size": 4192, "status": "c"}, "tasks/task_362_enterprise_industry_policy_analysis_medium_medium056.md|1777384241.273728|4024": {"hash": "fca001b64bf5fee98affc5c855420d7ab53b8661d1a3f1498272557ef154a62f", "size": 4024, "status": "c"}, "tasks/task_363_enterprise_industry_policy_analysis_medium_medium057.md|1777384241.2757344|3931": {"hash": "0adc79f4c6e0de7b712dab2e12d1ef991c525fe7911da08f62128095166becb0", "size": 3931, "status": "c"}, "tasks/task_364_enterprise_industry_policy_analysis_medium_medium058.md|1777384241.2757344|3968": {"hash": "4232757655e97c03f05ed019c4a8d01d6a582fe7f1c2681da929b2d3bef634d6", "size": 3968, "status": "c"}, "tasks/task_365_enterprise_industry_policy_analysis_medium_medium059.md|1777384241.2757344|4214": {"hash": "a0b776ab38851317553b6d96b23e896a0990385cc0edef7b88261d9d58a2fac9", "size": 4214, "status": "c"}, "tasks/task_366_enterprise_industry_policy_analysis_medium_medium060.md|1777384241.277742|3987": {"hash": "1bce0c0d08137b1f164305700989b0710c9661c70421e3cf4a74fef6772a0711", "size": 3987, "status": "c"}, "tasks/task_367_enterprise_industry_policy_analysis_medium_medium061.md|1777384241.277742|4020": {"hash": "4395658cc58d2d820fdf28ba57be48344ecc944e288ad9ec1dae57a33dd2be70", "size": 4020, "status": "c"}, "tasks/task_368_enterprise_industry_policy_analysis_medium_medium062.md|1777384241.2797482|4207": {"hash": "9d006d9272fa289115b084b4c98844f38caf6ca66e3e51c03367b041e1825d73", "size": 4207, "status": "c"}, "tasks/task_369_enterprise_industry_policy_analysis_medium_medium063.md|1777384241.2797482|3937": {"hash": "fae4f70e658128d96cb90519e918f7b0e2832cfbbd4038d51e7d37d03e46a525", "size": 3937, "status": "c"}, "tasks/task_370_enterprise_industry_policy_analysis_medium_medium064.md|1777384241.2797482|4096": {"hash": "612f25108a11c98e97cb90815e5ad2d002dff465ef6aaf53bbc86c4afb4e08c0", "size": 4096, "status": "c"}, "tasks/task_371_enterprise_industry_policy_analysis_medium_medium065.md|1777384241.281755|3926": {"hash": "2a73cf06658e5718f47db8cce2a1488280658af867618c1ef991981df6392ba9", "size": 3926, "status": "c"}, "tasks/task_372_enterprise_industry_policy_analysis_medium_medium066.md|1777384241.281755|3993": {"hash": "12c7a711765c1cbc5e48ffc1b58a89a6aabc059c8edb7dd99df54887590dd257", "size": 3993, "status": "c"}, "tasks/task_373_hypothesis_verification_hard_hard001.md|1777384241.283503|4600": {"hash": "15b60941a1dfd8e32671774ed6d3c5069da1d8d19f11a64725f5b6cd87163c99", "size": 4600, "status": "c"}, "tasks/task_374_hypothesis_verification_hard_hard002.md|1777384241.283503|4592": {"hash": "8c1477839f43bb639d68a08b34ae02930fa55261d3ca8b4bcfc5ce211d60b6ce", "size": 4592, "status": "c"}, "tasks/task_375_hypothesis_verification_hard_hard003.md|1777384241.283503|4490": {"hash": "f80e90f2247730927c5ffe4f9b2550fd9700d799d3f94680c3df4136819211df", "size": 4490, "status": "c"}, "tasks/task_376_hypothesis_verification_hard_hard004.md|1777384241.283503|4459": {"hash": "889ded893eeaf0c9eab56589b6a8294733bbe2b5a8e7cdf95b5f4251350bf11b", "size": 4459, "status": "c"}, "tasks/task_377_hypothesis_verification_hard_hard005.md|1777384241.283503|4652": {"hash": "a381adf78d5efc605c6a45d9ee20951c19d879d8cda0cc117db31103a3d6bd8f", "size": 4652, "status": "c"}, "tasks/task_378_hypothesis_verification_hard_hard006.md|1777384241.283503|4497": {"hash": "1764cd8e5205d17e8872e56f013c69c1d680723e27352eeaf87ede79ccc603d0", "size": 4497, "status": "c"}, "tasks/task_379_hypothesis_verification_hard_hard007.md|1777384241.283503|4630": {"hash": "0f82abfdcd29495b4ae3afb5a2c649cbf3f881fbb066444ce7b53afbc492e954", "size": 4630, "status": "c"}, "tasks/task_380_hypothesis_verification_hard_hard008.md|1777384241.283503|4909": {"hash": "0020a6ecf511b9d1ee8073aa9fa175ff795ccf0beb8d8e2a649c69492b02f309", "size": 4909, "status": "c"}, "tasks/task_381_hypothesis_verification_hard_hard009.md|1777384241.283503|4626": {"hash": "79974248ade401620b501d3a726ade11868bc6b437c176ebf12388607664635a", "size": 4626, "status": "c"}, "tasks/task_382_hypothesis_verification_hard_hard010.md|1777384241.283503|4597": {"hash": "dc2c53124da30545d421d62fda128d8b16753928d51eb391ecc8e6921068b6ee", "size": 4597, "status": "c"}, "tasks/task_383_hypothesis_verification_hard_hard011.md|1777384241.283503|4408": {"hash": "aac12eab9c280f9ac790b0cb7cc431c367212e21f36adc520c864d071fe3c726", "size": 4408, "status": "c"}, "tasks/task_384_hypothesis_verification_hard_hard012.md|1777384241.283503|4743": {"hash": "83bac7aab338ea41fb6c9987ae2e1c1e567f1464e76c3f2ed24223021f0ea84a", "size": 4743, "status": "c"}, "tasks/task_385_hypothesis_verification_hard_hard013.md|1777384241.283503|4935": {"hash": "47b99d1b75a1b41e22cd70936796d33f94876954ccd4e655aa4e59dd95827aec", "size": 4935, "status": "c"}, "tasks/task_386_hypothesis_verification_hard_hard014.md|1777384241.283503|4854": {"hash": "4dc360f0515e110480b50af8115d8c306dc52904921571911cc423111a033e20", "size": 4854, "status": "c"}, "tasks/task_387_hypothesis_verification_hard_hard015.md|1777384241.283503|4972": {"hash": "3e63251045d10494a798e510a7bb1c528bc2de6e1c0c5f1cb4298ad642c95b84", "size": 4972, "status": "c"}, "tasks/task_388_hypothesis_verification_medium_medium001.md|1777384241.283503|4397": {"hash": "4df61855ea434809e6f78a7012184cf453f83f13f9f0ed340e1cf3fdd952fc70", "size": 4397, "status": "c"}, "tasks/task_389_hypothesis_verification_medium_medium002.md|1777384241.283503|4240": {"hash": "25c34381250e2b7c450726c5e54cc68c7719d59d3fbbe9eb1227e7d97d870aff", "size": 4240, "status": "c"}, "tasks/task_390_hypothesis_verification_medium_medium003.md|1777384241.283503|4335": {"hash": "a709fff404658562296b1b8da914e7a7a55c3ebe536025f11eb5bc2b5a8903e6", "size": 4335, "status": "c"}, "tasks/task_391_hypothesis_verification_medium_medium004.md|1777384241.283503|4200": {"hash": "fd776943fb47fa456746f2952285b613d2f0572c0878248737e9a57cfa505ca3", "size": 4200, "status": "c"}, "tasks/task_392_hypothesis_verification_medium_medium005.md|1777384241.2992375|4293": {"hash": "425663a46c27e7a35751f8689e57940a0ffe550adb9444eb391d4e147aae508b", "size": 4293, "status": "c"}, "tasks/task_393_hypothesis_verification_medium_medium006.md|1777384241.2992375|4126": {"hash": "18f1691673c065ded9555f5ba9b99fd4963808914ed2d164eaa9e5d2d5ab9f4e", "size": 4126, "status": "c"}, "tasks/task_394_hypothesis_verification_medium_medium007.md|1777384241.2992375|4621": {"hash": "de9c2fe19d8af040cf7d7df5b11d07dbe1e3f86dd3de8e3cb5a615025a9fde62", "size": 4621, "status": "c"}, "tasks/task_395_hypothesis_verification_medium_medium008.md|1777384241.2992375|4670": {"hash": "0e01514db4753069cf6e2e2c85764e320c0f5237520b43cde3fcb3de58a41d47", "size": 4670, "status": "c"}, "tasks/task_396_hypothesis_verification_medium_medium009.md|1777384241.2992375|4371": {"hash": "02b31c1619ed54833c7651b0116e221b5bf07a669f498622b7744e594ce8deea", "size": 4371, "status": "c"}, "tasks/task_397_hypothesis_verification_medium_medium010.md|1777384241.2992375|5033": {"hash": "50e257865bbe26821b51fe166f8240185d27e0f6ad3bf96e5925dadb4fad9015", "size": 5033, "status": "c"}, "tasks/task_398_hypothesis_verification_medium_medium011.md|1777384241.2992375|4495": {"hash": "9c180aff6b6ae5f8bbfcf9f60155de8ba587bcf824c940b40e0670457096819b", "size": 4495, "status": "c"}, "tasks/task_399_hypothesis_verification_medium_medium012.md|1777384241.2992375|4466": {"hash": "829954414bfefb7b1e5d4df1b55370f9ec81f29af87e3b8a20f26ccd55676693", "size": 4466, "status": "c"}, "tasks/task_400_hypothesis_verification_medium_medium013.md|1777384241.2992375|4619": {"hash": "07a8c4d0a5c99f76a8782930bfaa9da8462022248c1fc52c25eb8fd4cc59c1a3", "size": 4619, "status": "c"}, "tasks/task_401_hypothesis_verification_medium_medium014.md|1777384241.2992375|4913": {"hash": "3e96d28b54c539d6ae69294cc26b04e81860fa92c98bb92fdafce35858ad22b5", "size": 4913, "status": "c"}, "tasks/task_402_industry_planning_hard_hard001.md|1777384241.2992375|4550": {"hash": "68200de38b6e8ee96b187f36c56aaf9307b5f788943c5c1db499b614d870236b", "size": 4550, "status": "c"}, "tasks/task_403_industry_planning_hard_hard002.md|1777384241.2992375|4690": {"hash": "1ed33afcd0a83023d152884a40825063e4934fac5d800f645cf12aba25ebd8c5", "size": 4690, "status": "c"}, "tasks/task_404_industry_planning_hard_hard003.md|1777384241.2992375|4769": {"hash": "0a5160b7930c22d6c5cdd18e4a786b5593ea3f221541b047e1e4c62e1f252c10", "size": 4769, "status": "c"}, "tasks/task_405_industry_planning_hard_hard004.md|1777384241.2992375|4852": {"hash": "09e13d2ce65ab99d36473c7322ff07047720e96d3dc3079abdbff643b3e00731", "size": 4852, "status": "c"}, "tasks/task_406_industry_planning_hard_hard005.md|1777384241.2992375|4744": {"hash": "043669a63bc42f6eb534e7c644b7bc56b41b3555c7a5ee8caf68e6df07ec838b", "size": 4744, "status": "c"}, "tasks/task_407_industry_planning_hard_hard006.md|1777384241.2992375|4798": {"hash": "d05669065ebfae31df6fef167ed3452f42fb9da8de2849027864728c313e658b", "size": 4798, "status": "c"}, "tasks/task_408_industry_planning_hard_hard007.md|1777384241.2992375|4731": {"hash": "3860d6a1ea7d31be3234d3fbd42e6f60ce0de1d15f0ac2b8caee986bf8384eac", "size": 4731, "status": "c"}, "tasks/task_409_industry_planning_hard_hard008.md|1777384241.2992375|4985": {"hash": "d52966ce06c2fe993bc0580f35a84b14ad4c4d69dae7ad5b0c25c653979d4ef0", "size": 4985, "status": "c"}, "tasks/task_410_industry_planning_hard_hard009.md|1777384241.2992375|4945": {"hash": "1de192555f48e16d3268a27cda571bb853767cf1a38e7889f20bfcaa4db44041", "size": 4945, "status": "c"}, "tasks/task_411_industry_planning_hard_hard010.md|1777384241.2992375|4914": {"hash": "de0f96bb27d1ff84cb4afbf84d099a212be36b77df1a406ceb2f5fb0cee317c3", "size": 4914, "status": "c"}, "tasks/task_412_industry_planning_hard_hard011.md|1777384241.3151557|4764": {"hash": "38419147b85cea362da6291016073e8cd24819bf71bb29aa1d6dae8869e97523", "size": 4764, "status": "c"}, "tasks/task_413_industry_planning_hard_hard012.md|1777384241.3151557|4510": {"hash": "0a4fb7bec52c0993a4432e5fee724bcdf3008162bf72440404fa3f603c61755d", "size": 4510, "status": "c"}, "tasks/task_414_industry_planning_hard_hard013.md|1777384241.3151557|4504": {"hash": "b195802142b24614f512b15d8e94e94dafc035defb5f0686124ef926bbda5a40", "size": 4504, "status": "c"}, "tasks/task_415_industry_planning_hard_hard014.md|1777384241.3151557|4567": {"hash": "16c97089b17478eac79608c4ed45b550aee86a9f247d3fca888370218b14cf01", "size": 4567, "status": "c"}, "tasks/task_416_industry_planning_medium_medium001.md|1777384241.3151557|4619": {"hash": "9e55fdf0522f787e9b82dc96c6482e1bb8fb69c5d7a454768bfa85a760d092c6", "size": 4619, "status": "c"}, "tasks/task_417_industry_planning_medium_medium002.md|1777384241.3151557|4262": {"hash": "0dcf29d76e34898fff054232cd834b1ed6cc33e5d94932e7ac9a21f129d243a5", "size": 4262, "status": "c"}, "tasks/task_418_industry_planning_medium_medium003.md|1777384241.3151557|4655": {"hash": "0209c72d0580c47a3d833a8f06af70acdbc2424e859917934638fe18f24394d9", "size": 4655, "status": "c"}, "tasks/task_419_industry_planning_medium_medium004.md|1777384241.3151557|4150": {"hash": "9e9319207863dff6342a463cbd9b4c51ec332d1cfbd41af10e4ace378c555f3e", "size": 4150, "status": "c"}, "tasks/task_420_industry_planning_medium_medium005.md|1777384241.3151557|4879": {"hash": "b78f37e3a5fe28579e7d5ca7202466dd1b586c97bdc1b0276c3ef7b809af26dd", "size": 4879, "status": "c"}, "tasks/task_421_industry_planning_medium_medium006.md|1777384241.3151557|5042": {"hash": "8910c70edde096057264dae5581ed64928b6ae3fde855e546850033cdbb3ec33", "size": 5042, "status": "c"}, "tasks/task_422_industry_planning_medium_medium007.md|1777384241.3151557|4817": {"hash": "5d4ec7da778028bd1a4a917a9d134c6d30b121f75f5806d037a13f105dad506e", "size": 4817, "status": "c"}, "tasks/task_423_industry_planning_medium_medium008.md|1777384241.3151557|4286": {"hash": "ccbb5d39f30769a91f4cd88b1481244e43130d97064eb85ea417fe7e8c0682fb", "size": 4286, "status": "c"}, "tasks/task_424_industry_planning_medium_medium009.md|1777384241.3151557|4868": {"hash": "43240d6f56a0b59d64dfea893d6c8909ff2af4c5cfc8ada4b869b4559bf29195", "size": 4868, "status": "c"}, "tasks/task_425_industry_planning_medium_medium010.md|1777384241.3151557|4565": {"hash": "47c70c28cb3d20068b05dc72400685cf50f90a4a55820de3f8b16f58009993a8", "size": 4565, "status": "c"}, "tasks/task_426_industry_planning_medium_medium011.md|1777384241.3151557|4548": {"hash": "9c583e1d2f17c8efc36242c1463a1d40fe963d6a53cc39a145595d599a555f44", "size": 4548, "status": "c"}, "tasks/task_427_industry_planning_medium_medium012.md|1777384241.3151557|4275": {"hash": "5473b71f5d88c31777acf57901154bee8bb80a736ae56ab94b89ddbc4094fe53", "size": 4275, "status": "c"}, "tasks/task_428_industry_planning_medium_medium013.md|1777384241.3151557|4891": {"hash": "bbb6e99f3e806c1c63107f9d54e63d72e07b803c14bb9133e304a28a65ca4168", "size": 4891, "status": "c"}, "tasks/task_429_industry_planning_medium_medium014.md|1777384241.3151557|4773": {"hash": "dc74623957e88238b25e76a0fecb74be0f6381e7c4fe9e42275ffc8b26d73c81", "size": 4773, "status": "c"}, "tasks/task_430_international_comparison_hard_hard001.md|1777384241.3151557|4893": {"hash": "27b401f04ae8c49581da865e1014a4cda9c082907dcb423b3415554924fa2834", "size": 4893, "status": "c"}, "tasks/task_431_international_comparison_hard_hard002.md|1777384241.3151557|4644": {"hash": "60bf54b6de6c4c7406d2748f0ba472b6ddb985540939a0d9ec2afa04d9af5551", "size": 4644, "status": "c"}, "tasks/task_432_international_comparison_hard_hard003.md|1777384241.3312073|4436": {"hash": "29e1def5dfcd4de6ce38bc306c1fd6f009243d2d60a4afaf0dedeaec451e5601", "size": 4436, "status": "c"}, "tasks/task_433_international_comparison_hard_hard004.md|1777384241.3312073|4574": {"hash": "ecd63a68fda3ba5dfcc0e372d1e84d70d7cf676e69b5536a46978aae6edc488c", "size": 4574, "status": "c"}, "tasks/task_434_international_comparison_hard_hard005.md|1777384241.3312073|4666": {"hash": "d3f3e3cfbefe7fb049b1d7b84fbaf82aa8aaea1aadcf69cd591a3e1b61920514", "size": 4666, "status": "c"}, "tasks/task_435_international_comparison_hard_hard006.md|1777384241.3312073|5111": {"hash": "b439e7f03a554ed96d2844750767b8cae3fcf97ac35b0db315fb03e8aa982938", "size": 5111, "status": "c"}, "tasks/task_436_international_comparison_hard_hard007.md|1777384241.3312073|5155": {"hash": "9271aaadd539c26190ae76df5773cd81adf9e99dda4f3ad86656ef8787d68331", "size": 5155, "status": "c"}, "tasks/task_437_international_comparison_hard_hard008.md|1777384241.3312073|4938": {"hash": "58aaab0eea5f7b6a8d87113a21a70a82e8ab0f310fe5a8fdbe4778bc9c9ca20a", "size": 4938, "status": "c"}, "tasks/task_438_international_comparison_hard_hard009.md|1777384241.3312073|4944": {"hash": "d2024314145ad3941e3d92635464ccf5a2570ac04baf5e6aa1368396c00f7c5d", "size": 4944, "status": "c"}, "tasks/task_439_international_comparison_hard_hard010.md|1777384241.3312073|5061": {"hash": "27cc51c4a02807e6e4ba1146db29cfb1610bc51feb960da59a95b73731ce0d65", "size": 5061, "status": "c"}, "tasks/task_440_international_comparison_hard_hard011.md|1777384241.3312073|4819": {"hash": "dc076ed11feaad87bbb9ef9763f780af1e73ca8c59b77dcb58baad45669ca165", "size": 4819, "status": "c"}, "tasks/task_441_international_comparison_hard_hard012.md|1777384241.3312073|5048": {"hash": "b4bcd8140273fa59fd86ffaba950c084c269fafe11ea90e88a5010f34274a2e2", "size": 5048, "status": "c"}, "tasks/task_442_international_comparison_hard_hard013.md|1777384241.3312073|5048": {"hash": "a289af5eba0558bcded34d42ca6ffb97b93b42799e497cad1caa81c2d077cea9", "size": 5048, "status": "c"}, "tasks/task_443_international_comparison_hard_hard014.md|1777384241.3312073|4693": {"hash": "1d7b1619f09b8a24ad139617b6503c472271054096dc69ebaac646a97d5b9db9", "size": 4693, "status": "c"}, "tasks/task_444_international_comparison_medium_medium001.md|1777384241.3312073|4314": {"hash": "bcef07a225fd3ea3ccb74e60d3e80e7e6028f7b237665b9ba55a7cd724603b84", "size": 4314, "status": "c"}, "tasks/task_445_international_comparison_medium_medium002.md|1777384241.3312073|4378": {"hash": "f3b97b8b46989f5ae3c0618219f7eee6887989cf687876bad9f1b435bf0cece7", "size": 4378, "status": "c"}, "tasks/task_446_international_comparison_medium_medium003.md|1777384241.3312073|4348": {"hash": "19244602fdee10ec73f8e8f43201e2bbe8729785bdfa32ca225d157af8d0b5af", "size": 4348, "status": "c"}, "tasks/task_447_international_comparison_medium_medium004.md|1777384241.3312073|4307": {"hash": "21439b4ce60148e3c5ad55ba30411c10089642206fd2b23798a7e43f40d29e49", "size": 4307, "status": "c"}, "tasks/task_448_international_comparison_medium_medium005.md|1777384241.3312073|4384": {"hash": "83ad22ea3f21ca5a23e5a8b486fc8386034797e727f5ac3f42e80a3221a62914", "size": 4384, "status": "c"}, "tasks/task_449_international_comparison_medium_medium006.md|1777384241.3312073|4260": {"hash": "84d4aab3421eb2d876f58a8d55fb0bc79914e9d2103fe3359b1e6234746def57", "size": 4260, "status": "c"}, "tasks/task_450_international_comparison_medium_medium007.md|1777384241.3312073|4357": {"hash": "fe488e0316d252a3e2d8b751ecb9e37a1b21e141984ddfb9440e61950c04139c", "size": 4357, "status": "c"}, "tasks/task_451_international_comparison_medium_medium008.md|1777384241.3312073|4353": {"hash": "7dd5092eeaadf17b16941b7592010689647fcde855160a3de55213ad7c1fab42", "size": 4353, "status": "c"}, "tasks/task_452_international_comparison_medium_medium009.md|1777384241.3312073|4477": {"hash": "8724ea925801a3fef05fb49310fe8f1301af19003ae8ff417c0b3fba6cc84b30", "size": 4477, "status": "c"}, "tasks/task_453_international_comparison_medium_medium010.md|1777384241.347264|4415": {"hash": "60c98758e50275ca3d844e8420b49354b48395bcbf4397ea40e0c811880f0dda", "size": 4415, "status": "c"}, "tasks/task_454_international_comparison_medium_medium011.md|1777384241.347264|4349": {"hash": "f5f712bdc23267d75b40454de65b400d262a12396a23a6bcae411535f546d47d", "size": 4349, "status": "c"}, "tasks/task_455_international_comparison_medium_medium012.md|1777384241.347264|4334": {"hash": "143ec1a4ac82641b8856133defa8c85473cea2c6a1ee199baa5823f6f7fa255e", "size": 4334, "status": "c"}, "tasks/task_456_international_comparison_medium_medium013.md|1777384241.347264|4402": {"hash": "ce0d789848a88bb221cc77ba61dd1754f292ebb3f0f36862e187e5e22dfec32f", "size": 4402, "status": "c"}, "tasks/task_457_international_comparison_medium_medium014.md|1777384241.347264|4648": {"hash": "06128300738b70c9fa6995dc67be37c0dfbeeadd8c2c759f96f24830f3d94262", "size": 4648, "status": "c"}, "tasks/task_458_international_comparison_medium_medium015.md|1777384241.347264|4734": {"hash": "16e0106cf21ec33591d6982ee176e617903ec72d01358770ea6162096879a80f", "size": 4734, "status": "c"}, "tasks/task_459_international_comparison_medium_medium016.md|1777384241.347264|4311": {"hash": "3ec22fbc81557a002c69911c6769e7317c9962e4e23deced3053f2566310a5b8", "size": 4311, "status": "c"}, "tasks/task_460_international_comparison_medium_medium017.md|1777384241.347264|4707": {"hash": "2330e0de22293f7066fbb0674c46383f259e73aba2b5f319c394ec6bb088a3b3", "size": 4707, "status": "c"}, "tasks/task_461_international_comparison_medium_medium018.md|1777384241.347264|4372": {"hash": "d0fcbcde59b970660d067cad8c0c685d9724f792c9fa4c67ee451e412aa69bc4", "size": 4372, "status": "c"}, "tasks/task_462_international_comparison_medium_medium019.md|1777384241.347264|4336": {"hash": "e26306b877ee508b236d7324b825e1085fa74219a7ba9a459c9b85aa19188f49", "size": 4336, "status": "c"}, "tasks/task_463_international_comparison_medium_medium020.md|1777384241.347264|4584": {"hash": "100a749d9e8f13243c198b1d7215f746714b0c49eb6f8f550bb8962da9723a8d", "size": 4584, "status": "c"}, "tasks/task_464_international_comparison_medium_medium021.md|1777384241.347264|4416": {"hash": "924d18859d0d2bc6321825f00eb69741943cf01568c7898203139cb52f798689", "size": 4416, "status": "c"}, "tasks/task_465_international_comparison_medium_medium022.md|1777384241.347264|4516": {"hash": "7e1b5bfe8ce724239bee7e4553c80948f337ceff578ecd589f2d41c3e0a717e4", "size": 4516, "status": "c"}, "tasks/task_466_international_comparison_medium_medium023.md|1777384241.347264|4720": {"hash": "7590414c0dd479a93aaa2c89109906e1231a2bbb776937ae23fbedfcc7375b63", "size": 4720, "status": "c"}, "tasks/task_467_international_comparison_medium_medium024.md|1777384241.347264|4720": {"hash": "dcc57c4dd7176e7543b6c71fb50f705d37b3304889026592baea6c54594d36ca", "size": 4720, "status": "c"}, "tasks/task_468_international_comparison_medium_medium025.md|1777384241.347264|4297": {"hash": "5d16aa918e3a0ad3585bb764e1b56ff031cb0fd845c1a892eeaed983b95c6a2c", "size": 4297, "status": "c"}, "tasks/task_469_risk_assessment_hard_hard001.md|1777384241.347264|4036": {"hash": "8f297afe05fba5571ceb3235882080e74f2398f14ef7ed3d57fa9e9653559f5f", "size": 4036, "status": "c"}, "tasks/task_470_risk_assessment_hard_hard002.md|1777384241.347264|4148": {"hash": "aa2a728877e147027dcd8fb22e1019d57ddc6fb47d9d7531f2f792dca402fee1", "size": 4148, "status": "c"}, "tasks/task_471_risk_assessment_hard_hard003.md|1777384241.347264|4712": {"hash": "cd4cdc971bb986565c91ba11ab3548ffa7e597ead48a80f8a74d5f28dafc1f16", "size": 4712, "status": "c"}, "tasks/task_472_risk_assessment_hard_hard004.md|1777384241.347264|4445": {"hash": "e4d0af4a1ac506da8dc0b777b5a675ea364fb08df39a73e4101f26f567df218f", "size": 4445, "status": "c"}, "tasks/task_473_risk_assessment_hard_hard005.md|1777384241.347264|4652": {"hash": "54117042135b9a013851dc8c5533ea7d67768a487191d5b388d423a2ef4a7450", "size": 4652, "status": "c"}, "tasks/task_474_risk_assessment_hard_hard006.md|1777384241.3630605|4652": {"hash": "cc4d1b592a8381910c54cee4b9e71bda12ab6485aebf577d9b9c7c14d96fc057", "size": 4652, "status": "c"}, "tasks/task_475_risk_assessment_hard_hard007.md|1777384241.3630605|4388": {"hash": "d758e244566dd15d2358971d657867a474b2fbe1f16a9b69b0b5892356ea8a07", "size": 4388, "status": "c"}, "tasks/task_476_risk_assessment_hard_hard008.md|1777384241.3630605|4113": {"hash": "467d3aaf296c181b42e67272c7fcc1e111b823aa5fd188e5c6a15b45ee7e8bad", "size": 4113, "status": "c"}, "tasks/task_477_risk_assessment_hard_hard009.md|1777384241.3630605|4737": {"hash": "f080df0562ec54567ce98001e0145faa411922f60f6e91a53793a9b7043d9b09", "size": 4737, "status": "c"}, "tasks/task_478_risk_assessment_hard_hard010.md|1777384241.3630605|4822": {"hash": "213ad46798d11629834b0f06e52f437ccf79ca13b745bf4f275ec7570b4ed40c", "size": 4822, "status": "c"}, "tasks/task_479_risk_assessment_hard_hard011.md|1777384241.3630605|4710": {"hash": "b677f384f72e1c2433f5d460d9dd37f63eaf754f1eadb55fe112d947e0a90a7f", "size": 4710, "status": "c"}, "tasks/task_480_risk_assessment_hard_hard012.md|1777384241.3630605|4389": {"hash": "633d3751bdd8fb8aa04c347b416a6ee3c7041768d12be90679e588bd02860201", "size": 4389, "status": "c"}, "tasks/task_481_risk_assessment_hard_hard013.md|1777384241.3630605|4478": {"hash": "a51e9cd3177ae79dad25a8be5d8e37c9a68fb105b7065030ea45499941fe875b", "size": 4478, "status": "c"}, "tasks/task_482_risk_assessment_medium_medium001.md|1777384241.3630605|4571": {"hash": "e2b6f2fecaa470c70d05b3fc3312402984c0642f9399123407f353bee8cbb200", "size": 4571, "status": "c"}, "tasks/task_483_risk_assessment_medium_medium002.md|1777384241.3630605|4115": {"hash": "ee9267e531d84252db51758ff69d409457d81d2c88cc994284a5a727753971cf", "size": 4115, "status": "c"}, "tasks/task_484_risk_assessment_medium_medium003.md|1777384241.3630605|4358": {"hash": "590f60b66a4c8a8fe306a25ec6680f036302b7a2987fda88ccec141f5f576914", "size": 4358, "status": "c"}, "tasks/task_485_risk_assessment_medium_medium004.md|1777384241.3630605|4604": {"hash": "a87e78768de05ab2b476bf70374df423fa009db74f497033e1cc3f3ec8c05978", "size": 4604, "status": "c"}, "tasks/task_486_risk_assessment_medium_medium005.md|1777384241.3630605|4879": {"hash": "1ae521971acb108048c8c4e37363dc3eebd0db8fe0a59af5a018ffbab049cda7", "size": 4879, "status": "c"}, "tasks/task_487_risk_assessment_medium_medium006.md|1777384241.3630605|4640": {"hash": "65169367dc566918faa6f3513a7c3fff72eeb658f5f1f472346abbf447bf1baa", "size": 4640, "status": "c"}, "tasks/task_488_risk_assessment_medium_medium007.md|1777384241.3630605|4682": {"hash": "3a54b9b7a420cd265b87a4db4189e695429cf48bbd1405c0ce4a4aca3e38a3c1", "size": 4682, "status": "c"}, "tasks/task_489_risk_assessment_medium_medium008.md|1777384241.3630605|4645": {"hash": "94b28ae7dece2907c98ca9165f2683cdfef3e3edc13610556c1d43f07afb415f", "size": 4645, "status": "c"}, "tasks/task_490_risk_assessment_medium_medium009.md|1777384241.3630605|4581": {"hash": "fbf15472e65f40fb9afe7db6b49a681e7c17b6cc8b729a96d35ceecce1e4b1db", "size": 4581, "status": "c"}, "tasks/task_491_risk_assessment_medium_medium010.md|1777384241.3630605|4657": {"hash": "671b95e92c5a6b829a5889ed7116a113dca042c94e778fff75436b08a2e26e1b", "size": 4657, "status": "c"}, "tasks/task_492_risk_assessment_medium_medium011.md|1777384241.3630605|4579": {"hash": "dd90c01abe9b3881a2c387769a333927fd36c8553f58ed1fda738d9082ee112d", "size": 4579, "status": "c"}, "logo.png|1777437658.4414198|428558": {"hash": "88c366ccbd7006a31c86d9936d3d4523831d5c03dc921032b3a46c9c8a96732c", "size": 428558, "status": "c"}}}
\ No newline at end of file
diff --git a/Dockerfile b/Dockerfile
new file mode 100644
index 0000000000000000000000000000000000000000..078efd617659496892e595a8a74f1a1056383ac6
--- /dev/null
+++ b/Dockerfile
@@ -0,0 +1,26 @@
+FROM ubuntu:22.04
+
+ENV DEBIAN_FRONTEND=noninteractive \
+    OPENCLAW_VERSION=2026.3.24
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    ca-certificates \
+    curl \
+    git \
+    python3 \
+    python3-pip \
+    bash \
+    tini \
+    && rm -rf /var/lib/apt/lists/*
+
+RUN curl -fsSL https://deb.nodesource.com/setup_22.x | bash - \
+    && apt-get update \
+    && apt-get install -y --no-install-recommends nodejs \
+    && rm -rf /var/lib/apt/lists/*
+
+RUN npm install -g "openclaw@${OPENCLAW_VERSION}"
+
+RUN mkdir -p /tmp_workspace /root/.openclaw/workspace
+
+ENTRYPOINT ["tini", "--"]
+CMD ["tail", "-f", "/dev/null"]
diff --git a/Images/.gitkeep b/Images/.gitkeep
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/LICENSE b/LICENSE
new file mode 100644
index 0000000000000000000000000000000000000000..b328f21241a9cfedd868087187101511e1e4112e
--- /dev/null
+++ b/LICENSE
@@ -0,0 +1,22 @@
+MIT License
+
+Copyright (c) 2026 The DataClaw Authors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+
diff --git a/README.md b/README.md
index 7be5fc7f47d5db027d120b8024982df93db95b74..5ed4c389e4801def5cde6626e9ddeb1ec293c80c 100644
--- a/README.md
+++ b/README.md
@@ -1,3 +1,373 @@
 ---
 license: mit
+task: data-analysis
+task_categories:
+  - question-answering
+  - tabular-question-answering
 ---
+
+<div align="center">
+
+<h1>DataClaw</h1>
+
+<img src="logo.png" alt="DataClaw Logo" width="220"/>
+
+<br/>
+<br/>
+
+[![🏆 Leaderboard](https://img.shields.io/badge/🏆_Leaderboard-DataClaw-red)](https://gtmllab.github.io/DataClaw/)
+![Tasks](https://img.shields.io/badge/Tasks-492-blue)
+![Categories](https://img.shields.io/badge/Categories-7-green)
+
+</div>
+
+> A data-analysis benchmark for OpenClaw-style end-to-end agents. Every task is grounded in real-world data and has a single objective gold answer.
+
+
+
+[简体中文](README.zh-CN.md)
+
+## 🌊 Data Analysis Tasks Are Changing in the OpenClaw Era
+
+With the emergence of end-to-end agents like OpenClaw, data analysis is no longer equivalent to static QA — "read a passage, output one answer." Real-world data analysis tasks often require agents to locate evidence across heterogeneous files, filter and join entities across tables, perform statistical and normalization calculations, verify intermediate results, and strictly follow output constraints.
+
+This means the core difficulty of a benchmark has shifted from answer generation alone to full agent-driven execution. A truly valuable data-analysis benchmark must test not only whether the final answer is correct, but also whether the agent can reliably complete a series of steps — retrieval, filtering, computation, verification, and constraint compliance — in complex data environments.
+
+DataClaw is designed for exactly this shift. It evaluates not abstract capability divorced from execution, but how OpenClaw-style end-to-end agents actually perform on data analysis tasks under real data conditions, explicit task constraints, and a reproducible execution protocol.
+
+## 🔍 What Is DataClaw?
+
+DataClaw is a process-oriented data-analysis benchmark for realistic, complex data environments. Its core goal is not merely to measure agents' end-task performance, but to serve as a high-fidelity testbed that also evaluates, at fine granularity, how agents evolve when facing real-world complexity and multi-step reasoning.
+
+DataClaw simulates at scale the noisy, weakly-semantic, cross-domain data environments found in the real world. Complex data-analysis questions are authored by domain experts in finance and computer science, and each task's process annotations and unique objective answers are cross-verified by human experts with AI assistance. Process annotations include task milestones, human-corrected reference trajectories, and evidence data sources. DataClaw adopts OpenClaw as its unified agent framework.
+
+
+## 🎯 Why DataClaw?
+
+- **From idealized data environments to imperfect real-world data environments.** DataClaw contains a mix of structured and unstructured data, covering enterprise profiles, business operating status, regional industry statistics, national industry statistics, and policy texts. All data is collected from the real world and comes with friction such as missing indicators, inconsistent definitions, and inconsistent naming. Tasks face realistic data environments, not over-cleaned single-table lookups.
+- **From single-shot static queries to multi-step dynamic reasoning.** DataClaw tasks typically require agents to complete a multi-stage chain of operations rather than producing a one-shot answer. The challenge for agents comes not only from retrieval but also from cross-source integration, metric construction, aggregation computation, and format constraint compliance.
+- **From outcome-oriented evaluation to process-oriented evaluation.** DataClaw goes beyond simple outcome-accuracy evaluation and dissects how the agent's execution unfolds at fine granularity. Outcome-oriented evaluation paradigms focus only on final accuracy. This black-box approach ignores intermediate reasoning and provides little actionable signal for guiding optimization.
+
+## 🏗️ Repository Layout
+
+
+Key directories and scripts:
+
+- `assets/database/`: Benchmark data files, injected wholesale into the container workspace at run time. The root contains `internal_metrics.csv` (internal business-logic knowledge base); `enterprise/`, `industry/`, and `policy/` hold the three theme-domain datasets.
+- `assets/qa_raw/`: Raw task source files.
+- `assets/qa_gold/`: Minimized gold files derived from `qa_raw`.
+- `tasks/`: Generated OpenClaw task spec files.
+- `dataclaw/build_tasks.py`: Builder that produces `qa_gold` and `tasks/` from `qa_raw`.
+- `dataclaw/eval/run_batch.py`: Host-side evaluation orchestrator; one isolated container per task.
+- `dataclaw/utils/docker_utils.py`: Container lifecycle management, OpenClaw onboarding, and model configuration.
+- `dataclaw/utils/grading.py`: Outcome scoring (LLM-judged Acc).
+- `dataclaw/utils/process_grading.py`: Process scoring (EE on correct tasks; GPR / TPE on incorrect tasks).
+- `script/docker_save_image.sh`: Image build and export script.
+
+### ⚙️ Evaluation Lifecycle
+
+Each evaluation task runs in its own Docker container. The host orchestrator manages the full lifecycle:
+
+```text
+Host (dataclaw/eval/run_batch.py)
+  |
+  +-- For each task (parallel via --parallel N):
+      1. docker run   -> start isolated container
+      2. docker cp    -> inject workspace files
+      3. docker exec  -> OpenClaw onboard
+      4. docker exec  -> start gateway (background)
+      5. docker exec  -> set model and run agent
+      6. docker exec  -> run llm_judge scoring
+      7. docker cp    -> collect logs and results
+      8. docker rm    -> remove container
+```
+
+
+## 🚀 User Quick Start
+
+### 1. Obtain the Pre-built Image
+
+Download the pre-built image archive from Releases and load it:
+
+```bash
+docker load -i <dataclaw-image-archive>.tar
+```
+
+After loading, confirm the local image tag matches `DOCKER_IMAGE` in `.env`.
+
+### 2. Clone This Repository
+
+```bash
+git clone <repository-url>
+cd <repository-dir>
+```
+
+Use the actual repository URL shown on the GitHub page.
+
+### 3. Install Python Dependencies
+
+```bash
+pip install pyyaml python-dotenv
+```
+
+> `pyproject.toml` requires Python `>=3.10`. For a fuller local dev setup, install additional dev dependencies as you prefer.
+
+### 4. Configure Environment
+
+Copy the template:
+
+```bash
+cp .env.example .env
+```
+
+Edit `.env` and pay attention to at least the following fields:
+
+| Variable | Required | Description |
+| --- | --- | --- |
+| `DEFAULT_MODEL` | Yes | Model under test, e.g. `openrouter/anthropic/claude-sonnet-4.6` |
+| `OPENROUTER_API_KEY` | One of two | Used when the main model or judge is called via OpenRouter |
+| `OPENCLAW_CUSTOM_BASE_URL` + `OPENCLAW_CUSTOM_API_KEY` | One of two | Custom OpenAI-compatible API |
+| `OPENCLAW_CUSTOM_MODEL_ID` | No | Explicit model id at the custom provider for the main model |
+| `JUDGE_MODEL` | No | Judge model; default in `.env.example` |
+| `JUDGE_CUSTOM_BASE_URL` + `JUDGE_CUSTOM_API_KEY` | No | Separate custom endpoint for the judge |
+| `JUDGE_CUSTOM_MODEL_ID` | No | Explicit model id for the judge custom endpoint |
+| `DOCKER_IMAGE` | No | Local image tag; must match the loaded image |
+
+#### 🔌 Custom OpenAI-compatible API
+
+If you do not use OpenRouter, set in `.env`:
+
+```bash
+OPENCLAW_CUSTOM_BASE_URL=https://your-api-url/v1
+OPENCLAW_CUSTOM_API_KEY=your_api_key
+OPENCLAW_CUSTOM_MODEL_ID=your-provider/your-model
+DEFAULT_MODEL=your-provider/your-model
+```
+
+If the API runs on the host:
+
+```bash
+OPENCLAW_CUSTOM_BASE_URL=http://host.docker.internal:8000/v1
+```
+
+When the judge uses a separate endpoint:
+
+```bash
+JUDGE_CUSTOM_BASE_URL=https://your-judge-api-url/v1
+JUDGE_CUSTOM_API_KEY=your_judge_api_key
+JUDGE_CUSTOM_MODEL_ID=your-provider/your-judge-model
+```
+
+##### Common Auth Setups (Main Model vs Judge)
+
+| Scenario | Main model | Judge | Required config |
+| --- | --- | --- | --- |
+| A | Custom API | OpenRouter | `OPENCLAW_CUSTOM_*` + `OPENROUTER_API_KEY` |
+| B | OpenRouter | OpenRouter | `OPENROUTER_API_KEY` |
+| C | Custom API | Custom API (separate endpoint) | `OPENCLAW_CUSTOM_*` + `JUDGE_CUSTOM_*` |
+
+### 5. Run Evaluation
+
+```bash
+# Run all tasks
+python dataclaw/eval/run_batch.py --model openrouter/anthropic/claude-sonnet-4.6
+
+# Run selected tasks
+python dataclaw/eval/run_batch.py --model ... --suite task_001,task_002
+
+# Run in parallel
+python dataclaw/eval/run_batch.py --model ... --parallel 4
+
+# Run a single task file
+python dataclaw/eval/run_batch.py --task tasks/task_001_xxx.md
+
+# Or use the convenience script (reads DEFAULT_MODEL from .env)
+bash script/run.sh
+```
+
+### CLI Options
+
+| Flag | Default | Description |
+| --- | --- | --- |
+| `--model` / `-m` | `DEFAULT_MODEL` in `.env` | Model under test |
+| `--judge` | `JUDGE_MODEL` in `.env` | Judge model |
+| `--suite` / `-s` | `all` | `"all"` or comma-separated task IDs |
+| `--task` / `-t` | — | Path to a single `task.md` |
+| `--parallel` / `-p` | `1` | Parallel container count |
+| `--timeout-multiplier` | `1.0` | Scale all task timeouts |
+| `--runs` | `1` | Repeat runs per task |
+| `--resume` | — | Resume from last interrupted run |
+| `--verbose` / `-v` | — | Enable verbose logging |
+
+### 6. Results
+
+After a run completes, results are saved under `output/<task_id>/<model_timestamp_runid>/`:
+
+```text
+output/<task_id>/<suffix>/
+├── score.json               # outcome score (Acc)
+├── process_score.json       # process scores (EE / GPR / TPE)
+├── usage.json               # token usage, cost, elapsed time
+├── agent.log                # agent execution log
+├── gateway.log              # gateway log
+├── chat.jsonl               # full conversation record
+├── judge_chat.jsonl         # outcome-judge conversation
+└── judge_process_chat.jsonl # process-judge conversation
+```
+
+A global summary is written to:
+
+```text
+output/summary_<model>.json
+```
+
+### 7. Grading Rules
+
+DataClaw scores each run along **four metrics**.
+
+| Metric | Definition | Scope | Direction |
+| --- | --- | --- | --- |
+| **Acc** | LLM-judge semantic match between predicted answer â and gold answer a; multi-subquestion tasks return the normalized mean over the L sub-answers. | All tasks | ↑ |
+| **EE** | Execution Efficiency = N / T, where N is the gold reference step count and T is the agent's actual step count. EE > 1 means the agent solved it in fewer steps than the gold trajectory. | Correct tasks only | ↑ |
+| **GPR** | Goal Progress Rate = (1/M) Σⱼ 𝕀(mⱼ achieved); fraction of M annotated milestones the agent reaches anywhere in its trajectory. Captures partial process credit when the final answer is wrong. | Incorrect tasks only | ↑ |
+| **TPE** | Temporal Progress Efficiency = (Σⱼ 𝕀(mⱼ) · γ^max(tⱼ − N, 0)) / Σⱼ 𝕀(mⱼ), γ = 0.9; averages an exponential temporal-decay factor over the milestones the agent did achieve. Milestones reached by step N contribute 1; later ones decay. TPE = 1 means all achieved milestones were on time, lower values indicate later concentration. Range [0, 1]. | Incorrect tasks only | ↑ |
+
+### 8. Resume After Interruption
+
+Long batch evaluations may be interrupted unexpectedly. The evaluation framework automatically saves a progress file after each task completes:
+
+```text
+output/progress_<model>.json
+```
+
+Simply append `--resume` to the original command to resume:
+
+```bash
+# Original run (interrupted midway)
+python dataclaw/eval/run_batch.py --model openrouter/anthropic/claude-sonnet-4.6 --suite all
+
+# Resume (keep other arguments the same)
+python dataclaw/eval/run_batch.py --model openrouter/anthropic/claude-sonnet-4.6 --suite all --resume
+```
+
+On resume, the framework automatically verifies that `--model`, `--suite`, and `--runs` match the previous run; if they don't, it exits with an error. Once all tasks are completed, the progress file is removed automatically.
+
+To discard previous progress and start fresh, simply delete the progress file:
+
+```bash
+rm output/progress_<model>.json
+```
+
+### 9. Cleanup
+
+Interrupted runs may leave behind uncleaned containers. Clean them up as follows:
+
+```bash
+IMAGE=<your-docker-image-tag>
+docker ps -a --filter "ancestor=${IMAGE}" -q | xargs -r docker rm -f
+```
+
+Preview containers that would be removed:
+
+```bash
+IMAGE=<your-docker-image-tag>
+docker ps -a --filter "ancestor=${IMAGE}" --format "{{.Names}}\t{{.Status}}"
+```
+
+## 📊 Dataset Statistics
+
+DataClaw's data does not come from synthetic samples or teaching examples; it is built on the publishing team's long-term, front-line data accumulation and industry insights from research on Chinese enterprises, industries, and policies. The current version is mainly based on data from 2022. After necessary de-identification, tasks are constructed to avoid model knowledge leakage as much as possible while preserving the information noise and data friction found in real business settings. Task authoring and annotation are conducted by a professional team from Lingnan College, Sun Yat-sen University, balancing academic rigor and practical usability.
+
+### 🗂️ Data Environment Statistics
+
+Under a theme-domain view and business-oriented taxonomy, the current data environment is organized into **3 theme domains**: **Enterprise**, **Industry**, and **Policy**; subcategories cover enterprise profiles and regional profiles, enterprise core competitiveness, business operating status, regional/national industry statistics, policy releases, and full policy texts — closely aligned with real research and consulting workflows. The 3 theme domains contain **17 independent data sources** (each data file counts as one source, all mounted under `assets/database/`): **17** are placed under theme subdirectories `enterprise/`, `industry/`, `policy/`; in addition, **1** root-level file, `internal_metrics.csv`, serves as an **internal business-logic knowledge base** and does not belong to any theme domain. Details below.
+
+| Dimension | Value | Notes |
+| --- | --- | --- |
+| Theme domains | 3 | Enterprise, Industry, Policy |
+| Secondary themes | 7 | Enterprise ×3 (profiles, core competitiveness, business status), Industry ×2 (regional industry, national industry), Policy ×2 (release status, full text) |
+| Total data sources | 17 | Injected into the container workspace at run time |
+| Format | CSV | Primarily CSV; includes both structured fields and unstructured long-text content |
+| Time span | Mainly concentrated in 2022 | Statistical periods vary across sources |
+
+**The 17 data sources by theme domain and secondary theme**
+
+<table>
+<thead>
+<tr><th>Theme domain</th><th>Secondary theme</th><th align="right">Sources</th><th>Files</th></tr>
+</thead>
+<tbody>
+<tr>
+<td style="white-space:nowrap">Enterprise</td>
+<td style="white-space:nowrap">Enterprise profiles</td>
+<td align="right">5</td>
+<td><code>enterprise/company_profile.csv</code><br><code>enterprise/company_profile_as.csv</code><br><code>enterprise/company_profile_eu.csv</code><br><code>enterprise/company_profile_na.csv</code><br><code>enterprise/company_profile_oc.csv</code></td>
+</tr>
+<tr>
+<td style="white-space:nowrap">Enterprise</td>
+<td style="white-space:nowrap">Core competitiveness</td>
+<td align="right">1</td>
+<td><code>enterprise/company_core.csv</code></td>
+</tr>
+<tr>
+<td style="white-space:nowrap">Enterprise</td>
+<td style="white-space:nowrap">Business status</td>
+<td align="right">3</td>
+<td><code>enterprise/company_operation_status.csv</code><br><code>enterprise/company_operation_status_detail.csv</code><br><code>enterprise/company_operation_yearly_status.csv</code></td>
+</tr>
+<tr>
+<td style="white-space:nowrap">Industry</td>
+<td style="white-space:nowrap">Regional industry</td>
+<td align="right">3</td>
+<td><code>industry/regional_industry_status.csv</code><br><code>industry/regional_industry_status_detail.csv</code><br><code>industry/regional_industry_yearly_status.csv</code></td>
+</tr>
+<tr>
+<td style="white-space:nowrap">Industry</td>
+<td style="white-space:nowrap">National industry</td>
+<td align="right">3</td>
+<td><code>industry/national_industry_status.csv</code><br><code>industry/national_industry_status_detail.csv</code><br><code>industry/national_industry_yearly_status.csv</code></td>
+</tr>
+<tr>
+<td style="white-space:nowrap">Policy</td>
+<td style="white-space:nowrap">Policy release status</td>
+<td align="right">1</td>
+<td><code>policy/policy_release_status.csv</code></td>
+</tr>
+<tr>
+<td style="white-space:nowrap">Policy</td>
+<td style="white-space:nowrap">Full policy text</td>
+<td align="right">1</td>
+<td><code>policy/policy_resource.csv</code></td>
+</tr>
+</tbody>
+</table>
+
+> At execution time, agents typically need to align entities across files, join across tables, normalize definitions, and perform aggregation calculations, rather than simply looking up values in a single file; when needed, they must also consult business conventions in `internal_metrics.csv`. This is the core value of DataClaw for evaluating real-scenario data understanding and reasoning capabilities.
+
+### 📋 Task Statistics
+
+The current version contains **492** tasks across **7** categories, with an overall difficulty distribution of **131 easy / 286 medium / 75 hard**.
+
+| Category code | Meaning | Count | Difficulty split |
+| --- | --- | --- | --- |
+| `enterprise_industry_analysis` | Enterprise–industry analysis | 226 | easy 115 / medium 111 |
+| `enterprise_industry_policy_analysis` | Enterprise–industry–policy linkage analysis | 76 | easy 10 / medium 66 |
+| `comprehensive_decision` | Comprehensive decision | 70 | easy 6 / medium 45 / hard 19 |
+| `international_comparison` | International comparison | 39 | medium 25 / hard 14 |
+| `hypothesis_verification` | Hypothesis verification | 29 | medium 14 / hard 15 |
+| `industry_planning` | Industry planning | 28 | medium 14 / hard 14 |
+| `risk_assessment` | Risk assessment | 24 | medium 11 / hard 13 |
+
+> Except for the 39 `international_comparison` tasks, all others are explicitly restricted in the current task spec to use only `./database/`, with no web search.
+
+## 🙏 Acknowledgements
+
+DataClaw is jointly released by Prof. Chuan Chen's team at the School of Computer Science, Sun Yat-sen University, and the Southern Weekly Sci-Tech Power Research Center. We sincerely thank the Southern Weekly Sci-Tech Power Research Center for providing invaluable data and tremendous support.
+
+This project also builds on excellent open-source agent ecosystems. We gratefully acknowledge:
+
+ [WildClawBench](https://github.com/InternLM/WildClawBench)
+
+ [Claw-Eval](https://github.com/claw-eval/claw-eval)
+
+ [PinchBench](https://github.com/pinchbench/skill)
diff --git a/README.zh-CN.md b/README.zh-CN.md
new file mode 100644
index 0000000000000000000000000000000000000000..0184855a080b65457a78bf78db3a698b72620245
--- /dev/null
+++ b/README.zh-CN.md
@@ -0,0 +1,368 @@
+<div align="center">
+
+<h1>DataClaw</h1>
+
+<img src="logo.png" alt="DataClaw Logo" width="220"/>
+
+<br/>
+<br/>
+
+[![🏆 Leaderboard](https://img.shields.io/badge/🏆_Leaderboard-DataClaw-red)](https://gtmllab.github.io/DataClaw/)
+[![🤗 HuggingFace](https://img.shields.io/badge/🤗_HuggingFace-Dataset-yellow)](https://huggingface.co/datasets/GTMLLab/DataClaw)
+![Tasks](https://img.shields.io/badge/Tasks-492-blue)
+![Categories](https://img.shields.io/badge/Categories-7-green)
+
+</div>
+
+> 面向 OpenClaw 类端到端智能体的数据分析评测集，所有任务来自真实数据环境，且具备唯一客观答案。
+
+
+
+[English README](README.md)
+
+## 🌊 OpenClaw 时代的数据分析任务正在变化
+
+随着 OpenClaw 一类端到端智能体的出现，数据分析任务已经不再等价于“读一段文本，再输出一个答案”的静态问答问题。真实世界中的数据分析任务往往要求智能体在异构文件中定位证据、跨表筛选与关联实体、执行统计与归一化计算、校验中间结果，并严格遵循输出约束。
+
+这意味着评测基准的核心难点，已经从单一的答案生成，转向完整的智能体驱动执行。一个真正有价值的数据分析评测基准，不仅要考察最终答案是否正确，还要考察智能体是否能够在复杂数据环境中稳定完成检索、筛选、计算、验证与约束遵循等一系列步骤。
+
+DataClaw 正是面向这一变化而设计。它评测的不是脱离执行环境的抽象能力，而是在真实数据条件、明确任务约束和可复现执行协议下，OpenClaw 类端到端智能体完成数据分析任务的实际表现。
+
+## 🔍 DataClaw 是什么
+
+DataClaw 是一个面向真实复杂数据环境的过程导向数据分析任务评测基准。核心目标并非仅仅是衡量智能体在数据分析任务上的最终表现，而是旨在将其作为一个现实高保真的测试平台，同时细粒度地评估智能体在面临真实复杂环境和多步推理时的演化过程。
+
+DataClaw大规模地模拟了高噪声、弱语义、跨领域的真实世界数据环境。通过金融领域与计算机领域的人类专家撰写复杂数据分析任务问题，并由人类专家和AI辅助交叉核验提供每个任务的过程性标注和唯一客观答案。其中，过程性标注包括任务里程碑、人工订正参考轨迹和证据数据来源。DataClaw 采用 OpenClaw 作为统一 agent 框架。
+
+
+## 🎯 为什么选择 DataClaw
+
+- **从理想化的假设数据环境到非理想化的真实数据环境。** DataClaw 包含混合结构和非结构化数据，覆盖企业画像、企业经营状态、区域产业统计、全国行业统计与政策文本等资源。所有信息均来自真实世界采集，并面临指标缺失、口径不一致、命名不一致等摩擦；任务面对的是现实风格的数据环境，而非过度清洗后的单表查值。
+- **从单点静态查询到多步动态推理。** DataClaw任务通常要求智能体完成多阶段操作链，而不是一次性命中答案。对 agent 的挑战不仅来自检索，更来自跨源整合、指标构造、聚合计算与格式约束执行。
+- **从结果导向评估到过程导向评估。** DataClaw超越单纯的结果准确率评估，对智能体的运行演化过程进行解剖式评测。结果导向的评估范式仅关注最终准确率。这种黑盒方法忽视了中间推理过程，极大弱化了评测的优化指导意义。
+
+## 🏗️ 本仓库代码架构
+
+
+关键目录与脚本职责如下：
+
+- `assets/database/`：评测数据文件，运行时会整体注入到容器工作区。根目录含 `internal_metrics.csv`（业务逻辑内部知识库）；`enterprise/`、`industry/`、`policy/` 为三大主题域数据。
+- `assets/qa_raw/`：原始任务源文件。
+- `assets/qa_gold/`：由 `qa_raw` 归约得到的精简 gold 文件。
+- `tasks/`：生成后的 OpenClaw task 规范文件。
+- `dataclaw/build_tasks.py`：从 `qa_raw` 生成 `qa_gold` 与 `tasks/` 的构建器。
+- `dataclaw/eval/run_batch.py`：宿主机侧评测编排器，负责每任务独立容器执行。
+- `dataclaw/utils/docker_utils.py`：容器生命周期管理、OpenClaw onboarding 与模型配置。
+- `dataclaw/utils/grading.py`：结果评分（LLM-judge 计算 Acc）。
+- `dataclaw/utils/process_grading.py`：过程评分（correct 任务计 EE；incorrect 任务计 GPR / TPE）。
+- `script/docker_save_image.sh`：镜像构建与导出脚本。
+
+### ⚙️ 评测执行生命周期
+
+每个评测任务都在独立 Docker 容器中运行。宿主机编排器管理完整生命周期：
+
+```text
+宿主机 (dataclaw/eval/run_batch.py)
+  |
+  +-- 对每个任务（通过 --parallel N 并行）：
+      1. docker run   -> 启动隔离容器
+      2. docker cp    -> 注入工作区文件
+      3. docker exec  -> OpenClaw onboard
+      4. docker exec  -> 启动 gateway（后台）
+      5. docker exec  -> 设置模型并运行 agent
+      6. docker exec  -> 调用 llm_judge 评分
+      7. docker cp    -> 收集日志与结果
+      8. docker rm    -> 清理容器
+```
+
+
+## 🚀 使用者快速开始
+
+### 1. 获取预构建镜像
+
+从 Releases 下载预构建镜像压缩包并加载：
+
+```bash
+docker load -i <dataclaw-image-archive>.tar
+```
+
+加载后，请确认本地镜像 tag 与 `.env` 中的 `DOCKER_IMAGE` 一致。
+
+### 2. 克隆本仓库
+
+```bash
+git clone <repository-url>
+cd <repository-dir>
+```
+
+实际仓库地址请以当前 GitHub 页面为准。
+
+### 3. 安装 Python 依赖
+
+```bash
+pip install pyyaml python-dotenv
+```
+
+> `pyproject.toml` 要求 Python `>=3.10`。如需更完整的本地开发环境，也可以按你的习惯额外安装开发依赖。
+
+### 4. 配置环境
+
+复制模板：
+
+```bash
+cp .env.example .env
+```
+
+编辑 `.env`，至少关注以下字段：
+
+| 变量 | 必填 | 说明 |
+| --- | --- | --- |
+| `DEFAULT_MODEL` | 是 | 待评测主模型，例如 `openrouter/anthropic/claude-sonnet-4.6` |
+| `OPENROUTER_API_KEY` | 二选一 | 当主模型或 Judge 通过 OpenRouter 调用时使用 |
+| `OPENCLAW_CUSTOM_BASE_URL` + `OPENCLAW_CUSTOM_API_KEY` | 二选一 | 自定义 OpenAI-compatible API |
+| `OPENCLAW_CUSTOM_MODEL_ID` | 否 | 自定义主模型在 provider 中的显式 model id |
+| `JUDGE_MODEL` | 否 | Judge 模型，默认值见 `.env.example` |
+| `JUDGE_CUSTOM_BASE_URL` + `JUDGE_CUSTOM_API_KEY` | 否 | Judge 使用独立自定义 endpoint 时填写 |
+| `JUDGE_CUSTOM_MODEL_ID` | 否 | Judge 自定义 endpoint 的显式 model id |
+| `DOCKER_IMAGE` | 否 | 本地镜像 tag，需与实际加载的镜像一致 |
+
+#### 🔌 自定义 OpenAI-compatible API
+
+如果不使用 OpenRouter，可以在 `.env` 中设置：
+
+```bash
+OPENCLAW_CUSTOM_BASE_URL=https://your-api-url/v1
+OPENCLAW_CUSTOM_API_KEY=your_api_key
+OPENCLAW_CUSTOM_MODEL_ID=your-provider/your-model
+DEFAULT_MODEL=your-provider/your-model
+```
+
+如果 API 运行在宿主机上：
+
+```bash
+OPENCLAW_CUSTOM_BASE_URL=http://host.docker.internal:8000/v1
+```
+
+Judge 使用独立 endpoint 时：
+
+```bash
+JUDGE_CUSTOM_BASE_URL=https://your-judge-api-url/v1
+JUDGE_CUSTOM_API_KEY=your_judge_api_key
+JUDGE_CUSTOM_MODEL_ID=your-provider/your-judge-model
+```
+
+##### 主模型与 Judge 的常见认证组合
+
+| 场景 | 主模型 | Judge | 所需配置 |
+| --- | --- | --- | --- |
+| A | 自定义 API | OpenRouter | `OPENCLAW_CUSTOM_*` + `OPENROUTER_API_KEY` |
+| B | OpenRouter | OpenRouter | `OPENROUTER_API_KEY` |
+| C | 自定义 API | 自定义 API（独立 endpoint） | `OPENCLAW_CUSTOM_*` + `JUDGE_CUSTOM_*` |
+
+### 5. 运行评测
+
+```bash
+# 运行全部任务
+python dataclaw/eval/run_batch.py --model openrouter/anthropic/claude-sonnet-4.6
+
+# 运行指定任务
+python dataclaw/eval/run_batch.py --model ... --suite task_001,task_002
+
+# 并行运行
+python dataclaw/eval/run_batch.py --model ... --parallel 4
+
+# 运行单个任务文件
+python dataclaw/eval/run_batch.py --task tasks/task_001_xxx.md
+
+# 或使用便捷脚本（从 .env 读取 DEFAULT_MODEL）
+bash script/run.sh
+```
+
+### CLI 选项
+
+| 参数 | 默认值 | 说明 |
+| --- | --- | --- |
+| `--model` / `-m` | `.env` 中的 `DEFAULT_MODEL` | 待评测模型 |
+| `--judge` | `.env` 中的 `JUDGE_MODEL` | Judge 模型 |
+| `--suite` / `-s` | `all` | `"all"` 或逗号分隔的任务 ID |
+| `--task` / `-t` | — | 单个 `task.md` 文件路径 |
+| `--parallel` / `-p` | `1` | 并行容器数 |
+| `--timeout-multiplier` | `1.0` | 缩放所有任务超时时间 |
+| `--runs` | `1` | 每个任务的重复运行次数 |
+| `--resume` | — | 从上次中断处恢复运行 |
+| `--verbose` / `-v` | — | 启用详细日志 |
+
+### 6. 结果
+
+运行完成后，结果保存在 `output/<task_id>/<model_timestamp_runid>/` 下：
+
+```text
+output/<task_id>/<suffix>/
+├── score.json               # 结果评分（Acc）
+├── process_score.json       # 过程评分（EE / GPR / TPE）
+├── usage.json               # token 统计、费用、耗时
+├── agent.log                # agent 执行日志
+├── gateway.log              # gateway 日志
+├── chat.jsonl               # 完整对话记录
+├── judge_chat.jsonl         # outcome-judge 对话
+└── judge_process_chat.jsonl # process-judge 对话
+```
+
+全局汇总会写入：
+
+```text
+output/summary_<model>.json
+```
+
+### 7. 评分规则
+
+DataClaw 对每次运行从 **四个指标** 打分。
+
+| 指标 | 含义 | 计算范围 | 方向 |
+| --- | --- | --- | --- |
+| **Acc** | LLM-judge 评判预测答案 â 与标准答案 a 的语义一致性，含 L 个子问题时取归一化平均 | 全部任务 | ↑ |
+| **EE** | 执行效率 = N / T（N=gold 参考步数，T=agent 实际步数）。EE>1 表示比 gold 更省步 | 仅 correct 任务 | ↑ |
+| **GPR** | 目标进展率 = (1/M) Σⱼ 𝕀(mⱼ 达成)，agent 在轨迹中达成的里程碑占比，用于在答错时捕捉过程得分 | 仅 incorrect 任务 | ↑ |
+| **TPE** | 时序进展效率 = (Σⱼ 𝕀(mⱼ) · γ^max(tⱼ−N, 0)) / Σⱼ 𝕀(mⱼ)，γ=0.9；对 agent 已达成的里程碑做时序衰减平均，N 步内达成贡献 1，之后按 γ 指数衰减。所有已达成里程碑均在第 N 步内达成时 TPE=1，越晚 TPE 越低。取值范围 [0, 1]。 | 仅 incorrect 任务 | ↑ |
+
+### 8. 中断恢复
+
+长时间的批量评测可能因意外而中断。评测框架会在每个任务完成后自动保存进度文件：
+
+```text
+output/progress_<model>.json
+```
+
+只需在原命令后加上 `--resume` 即可恢复：
+
+```bash
+# 原始运行（中途中断）
+python dataclaw/eval/run_batch.py --model openrouter/anthropic/claude-sonnet-4.6 --suite all
+
+# 恢复运行（其余参数保持一致）
+python dataclaw/eval/run_batch.py --model openrouter/anthropic/claude-sonnet-4.6 --suite all --resume
+```
+
+恢复时，框架会自动校验 `--model`、`--suite`、`--runs` 是否与上次运行一致；若不一致，会直接报错退出。全部任务完成后，进度文件会被自动删除。
+
+如需放弃之前的进度并重新开始，删除对应进度文件即可：
+
+```bash
+rm output/progress_<model>.json
+```
+
+### 9. 清理
+
+如果运行被中断，可能会留下尚未清理的容器。可以按如下方式清理：
+
+```bash
+IMAGE=<your-docker-image-tag>
+docker ps -a --filter "ancestor=${IMAGE}" -q | xargs -r docker rm -f
+```
+
+预览会被移除的容器：
+
+```bash
+IMAGE=<your-docker-image-tag>
+docker ps -a --filter "ancestor=${IMAGE}" --format "{{.Names}}\t{{.Status}}"
+```
+
+## 📊 数据集统计信息
+
+DataClaw数据并非来源于合成样本或教学示例，而是基于发布团队在中国企业、产业与政策研究中的长期一线数据积累与行业洞察。当前版本以 2022 年相关数据为主，经过必要脱敏处理后构建任务，尽可能避免模型知识泄露且保留真实业务中的信息噪声与数据摩擦。任务撰写与标注由中山大学岭南学院专业团队完成，兼顾学术规范与业务可用性。
+
+### 🗂️ 数据环境统计信息
+
+当前版本数据环境在主题域内，按业务口径归纳为 **3 个主题域**：**企业**、**产业**、**政策**；细分类目覆盖企业画像与分区画像、企业核心竞争力、经营状态、区域/全国产业统计、政策发布与政策原文等，贴近真实研究与咨询分析链路。3个主题域包含**17 个独立数据源**（每个数据文件计为一个数据源，统一挂载到 `assets/database/`）：其中 **17** 个按主题域归入子目录 `enterprise/`、`industry/`、`policy/`；另有 **1** 个根目录文件 `internal_metrics.csv`，作为**业务逻辑内部知识库**，不归属任一主题域。详细说明见下。
+
+| 维度 | 统计值 | 说明 |
+| --- | --- | --- |
+| 主题域 | 3 | 企业、产业、政策 |
+| 二级主题 | 7 | 企业 3（画像、核心竞争力、经营状态），产业 2（区域产业、全国行业），政策 2（发布情况、原文） |
+| 数据源总数 | 17 | 运行时整体注入容器工作区 |
+| 文件格式 | CSV | 以 CSV 为主；同时包含结构化字段与非结构化长文本型内容 |
+| 时间跨度 | 主要集中于2022年 | 不同数据源统计周期不一致 |
+
+**各主题域及二级主题下的 17 个数据源**
+
+<table>
+<thead>
+<tr><th>主题域</th><th>二级主题</th><th align="right">数据源数量</th><th>数据文件</th></tr>
+</thead>
+<tbody>
+<tr>
+<td style="white-space:nowrap">企业</td>
+<td style="white-space:nowrap">企业画像</td>
+<td align="right">5</td>
+<td><code>enterprise/company_profile.csv</code><br><code>enterprise/company_profile_as.csv</code><br><code>enterprise/company_profile_eu.csv</code><br><code>enterprise/company_profile_na.csv</code><br><code>enterprise/company_profile_oc.csv</code></td>
+</tr>
+<tr>
+<td style="white-space:nowrap">企业</td>
+<td style="white-space:nowrap">企业核心竞争力</td>
+<td align="right">1</td>
+<td><code>enterprise/company_core.csv</code></td>
+</tr>
+<tr>
+<td style="white-space:nowrap">企业</td>
+<td style="white-space:nowrap">企业经营状态</td>
+<td align="right">3</td>
+<td><code>enterprise/company_operation_status.csv</code><br><code>enterprise/company_operation_status_detail.csv</code><br><code>enterprise/company_operation_yearly_status.csv</code></td>
+</tr>
+<tr>
+<td style="white-space:nowrap">产业</td>
+<td style="white-space:nowrap">区域产业</td>
+<td align="right">3</td>
+<td><code>industry/regional_industry_status.csv</code><br><code>industry/regional_industry_status_detail.csv</code><br><code>industry/regional_industry_yearly_status.csv</code></td>
+</tr>
+<tr>
+<td style="white-space:nowrap">产业</td>
+<td style="white-space:nowrap">全国行业</td>
+<td align="right">3</td>
+<td><code>industry/national_industry_status.csv</code><br><code>industry/national_industry_status_detail.csv</code><br><code>industry/national_industry_yearly_status.csv</code></td>
+</tr>
+<tr>
+<td style="white-space:nowrap">政策</td>
+<td style="white-space:nowrap">政策发布情况</td>
+<td align="right">1</td>
+<td><code>policy/policy_release_status.csv</code></td>
+</tr>
+<tr>
+<td style="white-space:nowrap">政策</td>
+<td style="white-space:nowrap">政策原文</td>
+<td align="right">1</td>
+<td><code>policy/policy_resource.csv</code></td>
+</tr>
+</tbody>
+</table>
+
+> 在任务执行层面，智能体通常需要在多文件间完成实体对齐、跨表关联、口径归一与聚合计算，而非单文件查值；必要时还需结合 `internal_metrics.csv` 中的业务约定。这也是 DataClaw 用于评估真实场景数据理解与推理能力的核心价值。
+
+### 📋 任务统计信息
+
+当前版本共 492 个任务，覆盖 7 个类别；整体难度分布为 131 easy / 286 medium / 75 hard。
+
+| 类别代码 | 含义 | 任务数 | 难度分布 |
+| --- | --- | --- | --- |
+| `enterprise_industry_analysis` | 企业-行业分析 | 226 | easy 115 / medium 111 |
+| `enterprise_industry_policy_analysis` | 企业-产业-政策联动分析 | 76 | easy 10 / medium 66 |
+| `comprehensive_decision` | 综合决策 | 70 | easy 6 / medium 45 / hard 19 |
+| `international_comparison` | 国际比较 | 39 | medium 25 / hard 14 |
+| `hypothesis_verification` | 假设验证 | 29 | medium 14 / hard 15 |
+| `industry_planning` | 产业规划 | 28 | medium 14 / hard 14 |
+| `risk_assessment` | 风险评估 | 24 | medium 11 / hard 13 |
+
+> 除 `international_comparison` 的 39 个任务外，其余任务在当前 task 规范中都显式限制为仅使用 `./database/`，不依赖 web search。
+
+## 🙏 致谢
+
+DataClaw 由中山大学计算机学院陈川团队与南方周末科创力研究中心联合发布，真诚感谢南方周末科创力研究中心提供的宝贵数据和巨大支持。
+
+同时，本项目构建在优秀的开源智能体生态系统之上。我们衷心感谢以下项目：
+
+ [WildClawBench](https://github.com/InternLM/WildClawBench)
+
+ [Claw-Eval](https://github.com/claw-eval/claw-eval)
+
+ [PinchBench](https://github.com/pinchbench/skill)
+
+
diff --git a/assets/database/bilingual_translation_english_chinese.json b/assets/database/bilingual_translation_english_chinese.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d91bfb626053797109ed840327ad541bb7974b0
--- /dev/null
+++ b/assets/database/bilingual_translation_english_chinese.json
@@ -0,0 +1,340 @@
+{
+  "company_items": {
+    "海山昌工设备公司": [
+      "Haishan Chang Industrial Equipment Company",
+      "Haishan Changgong Equipment Company"
+    ],
+    "驰宏冶冶有色金属公司": "Chihong Yeye Non-ferrous Metals Co., Ltd",
+    "众白锦贸连锁公司": "Zhongbai Jinmao Chain Company",
+    "三三达腾重工公司": [
+      "Sansan Dateng Heavy Industry Company",
+      "Sansan Daten Heavy Industry Company"
+    ],
+    "华鲁润源科技股份公司": "Hualu Runyuan Technology Co., Ltd.",
+    "连机创机机床公司": "Lianji Chuangji Machine Tool Company",
+    "康盛安健生物医药公司": "Kangsheng Anjian Biopharmaceutical Company",
+    "恒逸昌化科技股份公司": "Hengyi Changhua Technology Co., Ltd.",
+    "东车科信系统公司": "Dongche Kexin Systems Company",
+    "众课软创软件公司": "Zhongke Ruanchuang Software Company",
+    "永惠昌达批发公司": "Yonghui Changda Wholesale Company",
+    "锌锗金泽材料公司": "Xin Ge Jinze Materials Company",
+    "健名安元医疗科技公司": "Jianming Anyuan Medical Technology Company",
+    "德西锦锦智能电气公司": "Dexi Jinjin Intelligent Electrical Company",
+    "永惠泽盛连锁公司": "Yonghui Zesheng Chain Company",
+    "步步盛锦商贸公司": "Bubusheng Jin Commerce Company",
+    "联花通泽商贸公司": [
+      "Lianhua Tongze Commerce Company",
+      "Lianhua Tongze Trading Company"
+    ],
+    "恒逸昌化精细化工公司": "Hengyi Changhua Fine Chemical Company",
+    "恒逸源锦精细化工公司": "Hengyi Yuanjin Fine Chemical Company",
+    "荣盛锦盛化学公司": "Rongsheng Jinsheng Chemical Company",
+    "物丽昌源批发公司": [
+      "Wulichangyuan Wholesale Company",
+      "Wu Li Chang Yuan Wholesale Co., Ltd.",
+      "Wu Li Chang Yuan Pi Fa Co., Ltd."
+    ],
+    "果投泽源新能源公司": "Guotouzeyuan New Energy Company",
+    "华谊昌泽科技股份公司": "Huayichangze Technology Co., Ltd.",
+    "药石生康医疗器械公司": "Yaoshi Shenkang Medical Equipment Company",
+    "宝新软联软件公司": "Baoxin Ruanlian Software Company",
+    "快克创锦设备公司": "Kuaike Chuangjin Equipment Company",
+    "浪集软创信息技术公司": "Langji Ruanchuang Information Technology Company",
+    "以山生辰医疗科技公司": "Yishan Shengchen Medical Technology Company",
+    "华仁泰泽药业股份公司": "Huaren Taize Pharmaceutical Co., Ltd.",
+    "复河辰泽生物医药公司": "Fuhe Chenze Biopharmaceutical Company",
+    "临公航腾重工公司": "Lingong Hangteng Heavy Industry Company",
+    "恒逸盛盛科技股份公司": "Hengyi Shengsheng Technology Co., Ltd.",
+    "以山泽元药业股份公司": "Yishan Zeyuan Pharmaceutical Co., Ltd.",
+    "瑞行泰元医疗器械公司": "Ruiying Taiyuan Medical Equipment Co., Ltd.",
+    "青青锦饮食品公司": "Qingqing Jinyin Food Company",
+    "安步尚昌品牌公司": "Anbu Shangchang Brand Company",
+    "普各瑞健生物医药公司": "Puge Ruijian Biopharmaceutical Company",
+    "贝壳": "KE Holdings",
+    "万国数据": "GDS Holdings",
+    "北汽路远新能源汽车公司": "Bei Qi Lu Yuan Xin Neng Yuan Qi Che Co., Ltd.",
+    "康盛康健药业股份公司": "Kangsheng Kangjian Pharmaceutical Co., Ltd.",
+    "众集达昌铜业公司": "Zhong Ji Da Chang Tong Ye Co., Ltd.",
+    "众海工筑锦建筑设计公司": "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd.",
+    "陵有色冶达资源公司": "Ling You Se Ye Da Zi Yuan Co., Ltd.",
+    "用丰信创科技公司": "Yong Feng Xin Chuang Ke Ji Co., Ltd.",
+    "润会数智系统公司": "Run Hui Shu Zhi Xi Tong Co., Ltd.",
+    "众通捷通运输公司": "Zhong Tong Jie Tong Yun Shu Co., Ltd.",
+    "云达航昌快递公司": "Yun Da Hang Chang Kuai Di Co., Ltd.",
+    "环星锦雅时尚公司": "Huan Xing Jin Ya Shi Shang Co., Ltd.",
+    "李丁盛尚纺织公司": "Li Ding Sheng Shang Fang Zhi Co., Ltd.",
+    "宝新科慧软件公司": "Bao Xin Ke Hui Ruan Jian Co., Ltd.",
+    "众课创信软件公司": "Zhong Ke Chuang Xin Ruan Jian Co., Ltd.",
+    "灿芯辉芯半导体公司": "Can Xin Hui Xin Semiconductor Co., Ltd.",
+    "瑞芯芯耀材料公司": "Rui Xin Xin Yao Materials Co., Ltd.",
+    "创玮耀耀电气公司": "Chuang Wei Yao Yao Dian Qi Co., Ltd.",
+    "美能电锦科技公司": "Mei Neng Dian Jin Technology Co., Ltd.",
+    "用丰信软网络公司": "Yong Feng Xin Ruan Network Co., Ltd.",
+    "金飞数软数据服务公司": "Jin Fei Shu Ruan Data Services Co., Ltd.",
+    "物丽汇锦零售公司": [
+      "Wu Li Hui Jin Retail Co., Ltd.",
+      "Wu Li Hui Jin Ling Shou Co., Ltd."
+    ],
+    "美能炫锦电气公司": "Mei Neng Xuan Jin Dian Qi Co., Ltd.",
+    "丽信耀悦电器公司": "Li Xin Yao Yue Dian Qi Co., Ltd.",
+    "海丽创耀家电公司": "Hai Li Chuang Yao Jia Dian Co., Ltd.",
+    "三夏泽能电力公司": "San Xia Ze Neng Dian Li Co., Ltd.",
+    "景能电热燃气公司": "Jing Neng Dian Re Ran Qi Co., Ltd.",
+    "以山泰泰医疗器械公司": "Yi Shan Tai Tai Medical Devices Co., Ltd.",
+    "光盛昌泽集团公司": "Guang Sheng Chang Ze Group Co., Ltd.",
+    "浪集云慧科技公司": "Lang Ji Yun Hui Technology Co., Ltd.",
+    "普各健辰药业股份公司": "Pu Ge Jian Chen Pharmaceutical Co., Ltd.",
+    "金湖地产建设开发公司": "Jin Hu Real Estate Construction Development Co., Ltd.",
+    "华城盛源综合开发公司": [
+      "Hua Cheng Sheng Yuan Integrated Development Co., Ltd.",
+      "Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd."
+    ],
+    "龙河置锦置业公司": [
+      "Long He Zhi Jin Real Estate Co., Ltd.",
+      "Long He Zhi Jin Zhi Ye Co., Ltd."
+    ],
+    "十阳智光电器公司": "Shi Yang Zhi Guang Dian Qi Co., Ltd.",
+    "徐业智工科技公司": "Xu Ye Zhi Gong Technology Co., Ltd.",
+    "原通盛盛供应链公司": "Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd.",
+    "神轴务锦咨询公司": "Shen Zhou Wu Jin Zi Xun Co., Ltd.",
+    "众邮政运运港口公司": "Zhong You Zheng Yun Yun Port Co., Ltd.",
+    "用丰科联软件公司": "Yong Feng Ke Lian Software Co., Ltd.",
+    "海丽炫悦电气公司": "Hai Li Xuan Yue Electric Co., Ltd.",
+    "恒逸润恒科技股份公司": "Heng Yi Run Heng Technology Co., Ltd.",
+    "连机机锦机床公司": "Lian Ji Ji Jin Ji Chuang Co., Ltd.",
+    "平汝港通运物流公司": "Ping Ru Gang Tong Yun Wu Liu Co., Ltd.",
+    "卫星润锦科技股份公司": "Wei Xing Run Jin Ke Ji Co., Ltd.",
+    "高尹泽通批发公司": [
+      "Gao Yin Ze Tong Pi Fa Co., Ltd.",
+      "Gaoyin Zetong Wholesale Co., Ltd."
+    ],
+    "永惠泽汇批发公司": "Yong Hui Ze Hui Pi Fa Co., Ltd.",
+    "麻钢泰锦材料公司": [
+      "Ma Gang Tai Jin Cai Liao Co., Ltd.",
+      "Magang Taijin Materials Co., Ltd."
+    ],
+    "麻钢钢盛不锈钢公司": [
+      "Ma Gang Gang Sheng Bu Xiu Gang Co., Ltd.",
+      "Magang Gangsheng Stainless Steel Co., Ltd."
+    ],
+    "创玮耀盛电器公司": "Chuangwei Yaosheng Electric Co., Ltd.",
+    "丽信智创家电公司": "Lixin Zhichuang Home Appliances Co., Ltd.",
+    "碧园产华地产控股公司": "Biyuan Chanhua Real Estate Holdings Co., Ltd.",
+    "花润置锦建设开发公司": "Huarun Zhijin Construction Development Co., Ltd.",
+    "健名生康医疗器械公司": "Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd.",
+    "康盛康瑞药业股份公司": "Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd.",
+    "创玮锦智电器公司": "Chuang Wei Jin Zhi Electrical Appliances Co., Ltd.",
+    "龙河产置地产控股公司": "Long He Chan Zhi Di Chan Holdings Co., Ltd.",
+    "花盈泰盛财富管理公司": [
+      "Hua Ying Tai Sheng Wealth Management Co., Ltd.",
+      "Huaying Taisheng Wealth Management Co., Ltd.",
+      "Hua Ying Tai Sheng Cai Fu Management Co., Ltd."
+    ],
+    "用丰联创系统公司": [
+      "Yong Feng Lian Chuang Xi Tong Co., Ltd.",
+      "Yongfeng Lianchuang Systems Co., Ltd."
+    ],
+    "众白达贸批发公司": "Zhongbai Damao Wholesale Co., Ltd.",
+    "鲁西润恒化工公司": "Luxi Runheng Chemical Co., Ltd.",
+    "美能炫悦电气公司": [
+      "Meineng Xuanyue Electric Co., Ltd.",
+      "Mei Neng Xuan Yue Dian Qi Co., Ltd."
+    ],
+    "包铁源昌金属制品公司": [
+      "Baotie Yuanchang Metal Products Co., Ltd.",
+      "Bao Tie Yuan Chang Jin Shu Zhi Pin Co., Ltd."
+    ],
+    "保禾华昌建设开发公司": "Bao He Hua Chang Jian She Kai Fa Co., Ltd.",
+    "丽群通通电子商务公司": "Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd.",
+    "亚玮泽智科技公司": "Ya Wei Ze Zhi Technology Co., Ltd.",
+    "创芯耀锐集成电路公司": "Chuang Xin Yao Rui Integrated Circuit Co., Ltd.",
+    "连机智盛机械公司": "Lian Ji Zhi Sheng Ji Xie Co., Ltd.",
+    "和联创航设备公司": "He Lian Chuang Hang She Bei Co., Ltd.",
+    "山拉达创智能装备公司": "Shan La Da Chuang Zhi Neng Zhuang Bei Co., Ltd.",
+    "三三工智科技公司": "Sansan Gongzhi Technology Co., Ltd.",
+    "杰杰达航设备公司": "Jiejie Dahang Equipment Co., Ltd.",
+    "众金冶冶资源公司": "Zhongjin Yeye Resources Co., Ltd.",
+    "海山味香餐饮管理公司": "Haishan Weixiang Catering Management Co., Ltd.",
+    "三松食锦调味品公司": "Sansong Shijin Condiment Co., Ltd.",
+    "美能电光家电公司": "Meineng Dianguang Home Appliances Co., Ltd.",
+    "丽信盛悦智能科技公司": "Lixin Shengyue Intelligent Technology Co., Ltd.",
+    "惠金锦瑞财富管理公司": [
+      "Huijin Jinrui Wealth Management Co., Ltd.",
+      "Huijin Jinrui Wealth Management Company",
+      "Hui Jin Jin Rui Cai Fu Management Co., Ltd."
+    ],
+    "众课智云数据服务公司": [
+      "Zhongke Zhiyun Data Services Co., Ltd.",
+      "Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.",
+      "Zhongke Zhiyun Data Services Company"
+    ],
+    "长桥锦创科技公司": [
+      "Changqiao Jinchuang Technology Co., Ltd.",
+      "Zhang Qiao Jin Chuang Technology Co., Ltd.",
+      "Zhangqiao Jinchuang Technology Co., Ltd.",
+      "Changqiao Jinchuang Technology Company"
+    ],
+    "物丽汇达连锁公司": [
+      "Wuli Huida Chain Co., Ltd.",
+      "Wu Li Hui Da Chain Co., Ltd.",
+      "Wuli Huida Chain Company",
+      "Wu Li Hui Da Lian Suo Co., Ltd."
+    ],
+    "众课科数软件公司": [
+      "Zhong Ke Shu Ruan Software Co., Ltd.",
+      "Zhong Ke Ke Shu Software Co., Ltd.",
+      "Zhongke Keshu Software Company",
+      "Zhong Ke Ke Shu Ruan Jian Co., Ltd."
+    ],
+    "碧园盛华建设开发公司": [
+      "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.",
+      "Biyuan Shenghua Construction Development Co., Ltd."
+    ],
+    "碧园置泽城市发展公司": [
+      "Bi Yuan Zhi Ze Urban Development Co., Ltd.",
+      "Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.",
+      "Biyuan Zhize Urban Development Co., Ltd.",
+      "Biyuan Zhize Urban Development Company"
+    ],
+    "招业华昌房地产开发公司": [
+      "Zhao Ye Hua Chang Real Estate Development Co., Ltd.",
+      "Zhaoye Huachang Real Estate Development Co., Ltd.",
+      "Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.",
+      "Zhaoye Huachang Real Estate Development Company"
+    ],
+    "铜通泽鸿证券公司": [
+      "Tong Tong Ze Hong Securities Co., Ltd.",
+      "Tongtong Zehong Securities Co., Ltd.",
+      "Tong Tong Ze Hong Zheng Quan Co., Ltd."
+    ],
+    "爱健颐康复众心公司": [
+      "Aijian Yikang Fuzhongxin Co., Ltd.",
+      "Ai Jian Yi Kang Fu Zhong Xin Co., Ltd.",
+      "Aijian Yikang Fuzhongxin Company"
+    ],
+    "宝新慧慧网络公司": [
+      "Baoxin Huihui Network Co., Ltd.",
+      "Bao Xin Hui Hui Wang Luo Co., Ltd.",
+      "Baoxin Huihui Network Company"
+    ],
+    "众车远泽船舶公司": [
+      "Zhong Che Yuan Ze Shipbuilding Co., Ltd.",
+      "Zhongche Yuanze Shipbuilding Co., Ltd.",
+      "Zhongche Yuanze Shipbuilding Company"
+    ],
+    "花图文教在线教育公司": [
+      "Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.",
+      "Huatu Wenjiao Online Education Co., Ltd."
+    ],
+    "金制鸿盛资产管理公司": [
+      "Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.",
+      "Jinzhi Hongsheng Asset Management Co., Ltd.",
+      "Jinzhi Hongsheng Asset Management Company",
+      "Jin Zhi Hong Sheng Zi Chan Management Co., Ltd."
+    ],
+    "健帆宁泽养老服务公司": [
+      "Jianfan Ningze Elderly Care Services Co., Ltd.",
+      "Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd."
+    ],
+    "一海昌锦商务公司": [
+      "Yihai Changjin Business Co., Ltd.",
+      "Yi Hai Chang Jin Shang Wu Co., Ltd.",
+      "Yihai Changjin Business Company"
+    ],
+    "浪集慧软科技公司": [
+      "Lang Ji Hui Ruan Technology Co., Ltd.",
+      "Langji Huiruan Technology Co., Ltd.",
+      "Langji Huiruan Technology Company"
+    ],
+    "花新源石新材料公司": [
+      "Hua Xin Yuan Shi New Materials Co., Ltd.",
+      "Huaxin Yuanshi New Materials Company",
+      "Hua Xin Yuan Shi Xin Cai Liao Co., Ltd."
+    ],
+    "碧园产锦不动产公司": [
+      "Bi Yuan Chan Jin Bu Dong Chan Co., Ltd.",
+      "Biyuan Chanjin Real Estate Company"
+    ],
+    "众海工昌锦建筑设计公司": [
+      "Zhonghai Gongchangjin Architectural Design Company",
+      "Zhong Hai Gong Chang Jin Jian Zhu She Ji Co., Ltd."
+    ],
+    "招业泽锦地产控股公司": [
+      "Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.",
+      "Zhao Ye Ze Jin Real Estate Holdings Co., Ltd."
+    ],
+    "瑞行健康制药公司": [
+      "Rui Xing Jian Kang Zhi Yao Co., Ltd.",
+      "Ruixing Health Pharmaceutical Company"
+    ],
+    "北控泽净水务公司": [
+      "Beikong Zejing Water Company",
+      "Bei Kong Ze Jing Water Co., Ltd."
+    ],
+    "恒丽科智软件公司": [
+      "Hengli Kezhi Software Company",
+      "Heng Li Ke Zhi Ruan Jian Co., Ltd."
+    ],
+    "十阳锦锦电器公司": [
+      "Shiyang Jinjin Electrical Appliances Company",
+      "Shi Yang Jin Jin Electrical Appliances Co., Ltd."
+    ],
+    "星酷文工艺美术品公司": [
+      "Xingkuwen Arts and Crafts Company",
+      "Xing Ku Wen Gong Yi Mei Shu Pin Co., Ltd."
+    ],
+    "新花源通连锁公司": "Xin Hua Yuan Tong Lian Suo Co., Ltd.",
+    "花电能锦水电公司": [
+      "Hua Dian Neng Jin Shui Dian Co., Ltd.",
+      "Hua Dian Neng Jin Hydropower Co., Ltd."
+    ],
+    "花能泽泽新能源公司": "Hua Neng Ze Ze Xin Neng Yuan Co., Ltd.",
+    "航发铁船航空科技公司": "Hang Fa Tie Chuan Hang Kong Technology Co., Ltd.",
+    "润会数科科技公司": "Run Hui Shu Ke Technology Co., Ltd.",
+    "众防昌达重工公司": "Zhong Fang Chang Da Zhong Gong Co., Ltd.",
+    "宝新智智系统公司": "Bao Xin Zhi Zhi Xi Tong Co., Ltd.",
+    "大族锦精设备公司": "Da Zu Jin Jing She Bei Co., Ltd.",
+    "锡份冶锦金属公司": "Xi Fen Ye Jin Jin Shu Co., Ltd.",
+    "绿泰洁循环保科技公司": "Lv Tai Jie Xun Huan Bao Technology Co., Ltd.",
+    "丰火创泽网络设备公司": "Feng Huo Chuang Ze Wang Luo She Bei Co., Ltd.",
+    "环丘泰锦智能电气公司": "Huan Qiu Tai Jin Zhi Neng Dian Qi Co., Ltd.",
+    "绿山置锦房地产开发公司": "Lv Shan Zhi Jin Real Estate Development Co., Ltd.",
+    "瑞芯耀澜集成电路公司": "Rui Xin Yao Lan Ji Cheng Dian Lu Co., Ltd.",
+    "晶芯锐辉微电子公司": "Jing Xin Rui Hui Wei Dian Zi Co., Ltd.",
+    "药石元泽生物医药公司": "Yao Shi Yuan Ze Sheng Wu Yi Yao Co., Ltd.",
+    "众集昌源钢铁公司": "Zhong Ji Chang Yuan Gang Tie Co., Ltd.",
+    "浪集联创信息技术公司": "Lang Ji Lian Chuang Xin Xi Ji Shu Co., Ltd.",
+    "潞安富昌煤炭公司": "Lu An Fu Chang Mei Tan Co., Ltd.",
+    "恒丽云创信息技术公司": "Heng Li Yun Chuang Xin Xi Ji Shu Co., Ltd.",
+    "万会盛置建设开发公司": "Wan Hui Sheng Zhi Construction Development Co., Ltd.",
+    "航发远锦航空科技公司": "Hang Fa Yuan Jin Hang Kong Technology Co., Ltd.",
+    "华鲁荣荣化学公司": "Hua Lu Rong Rong Hua Xue Co., Ltd.",
+    "三三工机科技公司": "San San Gong Ji Technology Co., Ltd.",
+    "众聚悦饮食品公司": "Zhong Ju Yue Yin Shi Pin Co., Ltd.",
+    "华城锦锦综合开发公司": "Hua Cheng Jin Jin Zong He Kai Fa Co., Ltd.",
+    "绿山产锦置业公司": "Lv Shan Chan Jin Zhi Ye Co., Ltd.",
+    "花新泽昌新材料公司": "Hua Xin Ze Chang Xin Cai Liao Co., Ltd.",
+    "万会锦盛房地产开发公司": [
+      "Wan Hui Jin Sheng Real Estate Development Co., Ltd.",
+      "Wan Hui Jin Sheng Fang Di Chan Kai Fa Co., Ltd."
+    ],
+    "包金金昌铜业公司": "Bao Jin Jin Chang Tong Ye Co., Ltd.",
+    "众邮政达锦运输公司": "Zhong You Zheng Da Jin Yun Shu Co., Ltd.",
+    "环星锦雅服饰公司": "Huan Xing Jin Ya Apparel Co., Ltd.",
+    "豪美公司": "Haomei Company"
+  },
+  "policy_items": {
+    "原材料工业\"三品\"实施方案": "Implementation Plan for \"Three Products\" in Raw Materials Industry",
+    "支持技工强省建设若干政策": "Several Policies on Supporting the Construction of a Strong Province of Skilled Workers",
+    "安徽省人民政府关于印发支持技工强省建设若干政策的通知": "Notice of Anhui Provincial People's Government on Several Policies Supporting the Construction of a Strong Province of Skilled Workers",
+    "广东省人民政府办公厅关于印发广东省进一步促进工业经济平稳增长若干措施的通知": "Notice of the General Office of Guangdong Provincial People's Government on Printing and Distributing Several Measures of Guangdong Province for Further Promoting Steady Growth of Industrial Economy",
+    "关于组织申报生物医药产业优秀青年人才首套房购买补贴的通知": "Notice on Organizing Applications for First Home Purchase Subsidies for Outstanding Young Talents in the Biomedicine Industry",
+    "东数西算": "Eastern Data Western Computing",
+    "广东省人民政府办公厅关于印发广东省进一步 促进工业经济平稳增长若干措施的通知": "Notice of Guangdong Provincial Development and Reform Commission on Issuing the Implementation Plan for Building Guangdong's Modern Circulation System during the 14th Five-Year Plan",
+    "广东省发展改革委关于印发《广东省“十四五”现代流通体系建设实施方案》": "Notice of the General Office of the People's Government of Guangdong Province on Issuing Several Measures to Further Promote Stable Growth of the Industrial Economy in Guangdong Province",
+    "关于上海市浦东新区有关研发机构适用进口税收政策资格认定事项的通知": "Notice on Qualification Recognition Matters for Relevant R&D Institutions in Pudong New Area of Shanghai Applicable to Import Tax Policies",
+    "上海市人民政府办公厅关于印发培育“元宇宙”新赛道行动方案的通知": "Notice of the General Office of the Shanghai Municipal People's Government on Issuing the Action Plan for Cultivating the New Metaverse Track",
+    "黑龙江省人民政府办公厅关于印发黑龙江省科技振兴行动计划（2022—2026年）的通知": "Notice of the General Office of the People's Government of Heilongjiang Province on Issuing the Heilongjiang Province Science and Technology Revitalization Action Plan (2022-2026)",
+    "黑龙江省支持大型国际合作交流活动对接引进优秀人才经费补助实施细则（试行）": "Detailed Rules for the Implementation of Funding Subsidies for Supporting Large-scale International Cooperation and Exchange Activities to Connect and Introduce Outstanding Talents in Heilongjiang Province (Trial)"
+  }
+}
\ No newline at end of file
diff --git a/assets/database/enterprise/company_core.csv b/assets/database/enterprise/company_core.csv
new file mode 100644
index 0000000000000000000000000000000000000000..18dfb4baa130e96e902d21f45143e1445242b8f9
--- /dev/null
+++ b/assets/database/enterprise/company_core.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e7fadec7a8833ef4c5736d638c62bd1033247e50a47afb2fa36b0af81fa6a0f
+size 27411510
diff --git a/assets/database/enterprise/company_operation_status.csv b/assets/database/enterprise/company_operation_status.csv
new file mode 100644
index 0000000000000000000000000000000000000000..0fe66b20323f934fbb5227d33ac511edd8b5b5e8
--- /dev/null
+++ b/assets/database/enterprise/company_operation_status.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d30f8659da51de379500f0314d5d6322d534267e92377ca9eea47336e9fff94
+size 34717208
diff --git a/assets/database/enterprise/company_operation_status_detail.csv b/assets/database/enterprise/company_operation_status_detail.csv
new file mode 100644
index 0000000000000000000000000000000000000000..f5bff45e0ce45339e2353ad73786f55c040c58a1
--- /dev/null
+++ b/assets/database/enterprise/company_operation_status_detail.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:016e39c51abb7f940506d2855bbc0419a6ba50dbbe14e62684285d6c7985d569
+size 70022647
diff --git a/assets/database/enterprise/company_operation_yearly_status.csv b/assets/database/enterprise/company_operation_yearly_status.csv
new file mode 100644
index 0000000000000000000000000000000000000000..59b1a325701408a9bcfb2df9f9cb6308d72afd5a
--- /dev/null
+++ b/assets/database/enterprise/company_operation_yearly_status.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df4aa7baff8cea1d9eadec7b6650d57a827ea0520cb3fe636d708afadcb0fd15
+size 65472033
diff --git a/assets/database/enterprise/company_profile.csv b/assets/database/enterprise/company_profile.csv
new file mode 100644
index 0000000000000000000000000000000000000000..82242376788729fbd1cdb43a85ff960cc4f9f7e9
--- /dev/null
+++ b/assets/database/enterprise/company_profile.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ecd68737f4f453746c3271b146ef5f6b6e741536aab4fc8ae2a32be89085e3c2
+size 1783742
diff --git a/assets/database/enterprise/company_profile_as.csv b/assets/database/enterprise/company_profile_as.csv
new file mode 100644
index 0000000000000000000000000000000000000000..e6e8a20cfb42d4d44c998c2dd1e80b4763d0ae23
--- /dev/null
+++ b/assets/database/enterprise/company_profile_as.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14488f5e36b9eec554bc1776513135206dc83b3fc834441bc7eff4ca5171a3cc
+size 242816
diff --git a/assets/database/enterprise/company_profile_eu.csv b/assets/database/enterprise/company_profile_eu.csv
new file mode 100644
index 0000000000000000000000000000000000000000..0ec35b3775dd8d1f0783b97a14d149807ee14cff
--- /dev/null
+++ b/assets/database/enterprise/company_profile_eu.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:448bd2077692933fb41398dff37f7767c2cac726d852bf54d5ea9ead2b1c4a8e
+size 228500
diff --git a/assets/database/enterprise/company_profile_na.csv b/assets/database/enterprise/company_profile_na.csv
new file mode 100644
index 0000000000000000000000000000000000000000..7bf9be04c95f8fa415e06563a16a40693c69e856
--- /dev/null
+++ b/assets/database/enterprise/company_profile_na.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83323ec3a56cb36abbb86263c95ee39e0ad5b2bff5101e90d95a69d5efe158cf
+size 152112
diff --git a/assets/database/enterprise/company_profile_oc.csv b/assets/database/enterprise/company_profile_oc.csv
new file mode 100644
index 0000000000000000000000000000000000000000..a210a2d78cfc87bb0dd7ecd8910b89cf7e1e6f47
--- /dev/null
+++ b/assets/database/enterprise/company_profile_oc.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f174f6be68dfd8ccfabfc50fed443faa3ffc339b23012a6e54a262b72d19981
+size 79246
diff --git a/assets/database/industry/national_industry_status.csv b/assets/database/industry/national_industry_status.csv
new file mode 100644
index 0000000000000000000000000000000000000000..a98cdec5a5905957e1270c584c8a2777ed75ff65
--- /dev/null
+++ b/assets/database/industry/national_industry_status.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:723ce91a24f67b5a3d032a297d40417a1df32c2d584c8215047de92f7c204fd6
+size 1472926
diff --git a/assets/database/industry/national_industry_status_detail.csv b/assets/database/industry/national_industry_status_detail.csv
new file mode 100644
index 0000000000000000000000000000000000000000..d054deb1de7d78c6fd23893be805cb479c08e7c1
--- /dev/null
+++ b/assets/database/industry/national_industry_status_detail.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f040fb21b4208bc280acc3bb6c81feb286bbd63150a30ca22677ed802d529b7
+size 3631383
diff --git a/assets/database/industry/national_industry_yearly_status.csv b/assets/database/industry/national_industry_yearly_status.csv
new file mode 100644
index 0000000000000000000000000000000000000000..aaba03d6527f62128665cf93e262f2b5a52d5a62
--- /dev/null
+++ b/assets/database/industry/national_industry_yearly_status.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6a2f88f64e19168d8adb70b0edd9dbeebd8df44c4db472c5b23af7a8b0d0f85
+size 55980
diff --git a/assets/database/industry/regional_industry_status.csv b/assets/database/industry/regional_industry_status.csv
new file mode 100644
index 0000000000000000000000000000000000000000..d7ead6a5c7ae031d4065960a6388442742f3f973
--- /dev/null
+++ b/assets/database/industry/regional_industry_status.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6314f5198f67c8ebf90e03f891f042d8244a5cee5538bd4731678820b289d7e
+size 22504920
diff --git a/assets/database/industry/regional_industry_status_detail.csv b/assets/database/industry/regional_industry_status_detail.csv
new file mode 100644
index 0000000000000000000000000000000000000000..e87e88382f46871fe16c55d065add42f3b23e5d3
--- /dev/null
+++ b/assets/database/industry/regional_industry_status_detail.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a08132df31975bcfbb56ea2d2a3b79dc3b052b97d0f88b63ebe1542a0d6f822d
+size 287796
diff --git a/assets/database/industry/regional_industry_yearly_status.csv b/assets/database/industry/regional_industry_yearly_status.csv
new file mode 100644
index 0000000000000000000000000000000000000000..b2fbe4e895f7375634f76e8fd0b91402d3080c0e
--- /dev/null
+++ b/assets/database/industry/regional_industry_yearly_status.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab96eb7f4ca90cf54cd75515beb6031ed72ff26d2df8d3264a969b1c2d8aab41
+size 289546
diff --git a/assets/database/internal_metrics.csv b/assets/database/internal_metrics.csv
new file mode 100644
index 0000000000000000000000000000000000000000..88586d4ca7a79b8e075ee89f9f97168f7ce34c1a
--- /dev/null
+++ b/assets/database/internal_metrics.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f584fb7cacf19fcdc50e59b398fc65ad041d9651860d9cfe1d7afb277dbeb9dc
+size 2238
diff --git a/assets/database/policy/policy_release_status.csv b/assets/database/policy/policy_release_status.csv
new file mode 100644
index 0000000000000000000000000000000000000000..9c7c74835cad3ed7f747ed8d8997ea593d3dd2c7
--- /dev/null
+++ b/assets/database/policy/policy_release_status.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a784a9301e6a8feb63d7b5e2fae7da0f4a3fee3d1871703cf1560d36163184ef
+size 1109101
diff --git a/assets/database/policy/policy_resource.csv b/assets/database/policy/policy_resource.csv
new file mode 100644
index 0000000000000000000000000000000000000000..bb3d91018d81c62c449d6add3043b22b824a274c
--- /dev/null
+++ b/assets/database/policy/policy_resource.csv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:041c1e89ed2a24164075e92c9b8e7f74490042ef52dcfbd2c0e8f5c17764a029
+size 20833614
diff --git a/assets/qa_gold/comprehensive_decision/easy001.json b/assets/qa_gold/comprehensive_decision/easy001.json
new file mode 100644
index 0000000000000000000000000000000000000000..e1c1fe71deaa41c3ca8afd64805e9f43ed7cbccd
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/easy001.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy001",
+  "question": "In which province is the enterprise with the highest R&D investment ratio nationwide in 2022 located?",
+  "guidelines": "The answer must be \"yes\" or \"no\". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "Jiangsu Province",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all records with year=2022 from company_operation_status.csv, and extract the enterprise name and R&D investment ratio fields.",
+    "Filter enterprise records with non-null R&D investment ratio, sort by R&D investment ratio in descending order, and identify the enterprise with the highest R&D investment ratio as \"Yaoshi Shenkang Medical Equipment Company\" with an R&D investment ratio of 934642.80%.",
+    "Look up the record for enterprise name \"Yaoshi Shenkang Medical Equipment Company\" in company_profile.csv, extract the province field, and obtain the province where the enterprise is located as \"Jiangsu Province\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Enterprise with highest R&D investment ratio": "Yaoshi Shenkang Medical Equipment Company",
+    "R&D investment ratio (%)": 934642.8,
+    "Province of location": "Jiangsu Province"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/easy002.json b/assets/qa_gold/comprehensive_decision/easy002.json
new file mode 100644
index 0000000000000000000000000000000000000000..bffb186ffd8d589ad624f55fbc4b9a8bdc741dfd
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/easy002.json
@@ -0,0 +1,36 @@
+{
+  "id": "easy002",
+  "question": "In 2022, nationwide, how is the chemical raw materials and chemical products manufacturing industry ranked by asset scale? Please list the top five provinces.",
+  "guidelines": "Answer format: [Province A, Province B, ...]. Output only province names, do not add any other explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": [
+    "Shandong Province",
+    "Zhejiang Province",
+    "Jiangsu Province",
+    "Shanghai",
+    "Guangdong Province"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all records with industry=\"Chemical Raw Materials and Chemical Products Manufacturing\" from regional_industry_status.csv, and extract the province and total assets fields.",
+    "Sort all provincial data by total assets in descending order to determine the asset scale ranking of each province. Extract the top five provinces and their total assets: Shandong Province (544109686731.85), Zhejiang Province (401644798867.80), Jiangsu Province (250743006622.33), Shanghai (230472528900.23), Guangdong Province (176926240169.03)."
+  ],
+  "steps_num": 2,
+  "milestone": {
+    "Total assets in Shandong Province (yuan)": 544109686731.85,
+    "Total assets in Zhejiang Province (yuan)": 401644798867.8,
+    "Total assets in Jiangsu Province (yuan)": 250743006622.33,
+    "Total assets in Shanghai (yuan)": 230472528900.23,
+    "Total assets in Guangdong Province (yuan)": 176926240169.03,
+    "Top five provinces": [
+      "Shandong Province",
+      "Zhejiang Province",
+      "Jiangsu Province",
+      "Shanghai",
+      "Guangdong Province"
+    ]
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/easy003.json b/assets/qa_gold/comprehensive_decision/easy003.json
new file mode 100644
index 0000000000000000000000000000000000000000..b8101b7f95189ac53af865d55a1b9dccebd0e565
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/easy003.json
@@ -0,0 +1,23 @@
+{
+  "id": "easy003",
+  "question": "In 2022, in the Qilu region (Shandong Province), how many policies support Zhongbai Jinmao Chain Company in its industry?",
+  "guidelines": "The answer must be an exact number. Output only the number, do not add units, commas, or any textual explanation. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 2,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Search for records of \"Zhongbai Jinmao Chain Company\" in company_profile.csv, extract the industry field, and determine that its industry is \"Wholesale and Retail Trade\".",
+    "Filter from policy_release_status.csv for all policy records with province=\"Shandong Province\", and policies where the industry field contains \"Wholesale and Retail Trade\". Found 2 policies: 1 from \"Local Policy - Shandong Provincial People's Government General Office Policy Count\" and 1 from \"Local Policy - Shandong Provincial Development and Reform Commission Policy Count\".",
+    "Count the number of policies meeting the criteria, which is 2."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Zhongbai Jinmao Chain Company": "Wholesale and Retail Trade",
+    "Shandong Province Local Policy - Shandong Provincial People's Government General Office Policy Count": 1,
+    "Shandong Province Local Policy - Shandong Provincial Development and Reform Commission Policy Count": 1,
+    "Number of policies in Shandong Province related to Wholesale and Retail Trade": 2
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/easy004.json b/assets/qa_gold/comprehensive_decision/easy004.json
new file mode 100644
index 0000000000000000000000000000000000000000..f5e49d3926c84bead2ce7082d7e291e2c36675ca
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/easy004.json
@@ -0,0 +1,36 @@
+{
+  "id": "easy004",
+  "question": "In 2022, nationwide, what is the ranking of provinces by asset size in the Information Transmission, Software and Information Technology Services industry? Please list the top five provinces.",
+  "guidelines": "Answer format: [Province A, Province B, ...]. Output only the province ranking, do not add any other explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": [
+    "Beijing",
+    "Zhejiang Province",
+    "Guangdong Province",
+    "Shanghai",
+    "Jiangsu Province"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter from regional_industry_status.csv all records with industry=\"Information Transmission, Software and Information Technology Services\", and extract the province and total assets fields.",
+    "Sort all provincial data in descending order by total assets to determine the asset size ranking of each province. Extract the top five provinces and their total assets: Beijing (10297490896006.5), Zhejiang Province (4115693929492.25), Guangdong Province (2262247330030.01), Shanghai (1282711003966.55), Jiangsu Province (177568006242.47)."
+  ],
+  "steps_num": 2,
+  "milestone": {
+    "Total assets of Beijing (CNY)": 10297490896006.5,
+    "Total assets of Zhejiang Province (CNY)": 4115693929492.25,
+    "Total assets of Guangdong Province (CNY)": 2262247330030.01,
+    "Total assets of Shanghai (CNY)": 1282711003966.55,
+    "Total assets of Jiangsu Province (CNY)": 177568006242.47,
+    "Top five provinces by ranking": [
+      "Beijing",
+      "Zhejiang Province",
+      "Guangdong Province",
+      "Shanghai",
+      "Jiangsu Province"
+    ]
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/easy005.json b/assets/qa_gold/comprehensive_decision/easy005.json
new file mode 100644
index 0000000000000000000000000000000000000000..e146c847d12aca4696a77782ad0db2c7b8be7cf0
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/easy005.json
@@ -0,0 +1,37 @@
+{
+  "id": "easy005",
+  "question": "In 2022, nationwide, what is the ranking of provinces by profitability in the Real Estate industry? Please list the top five provinces.",
+  "guidelines": "Answer format: [Province A, Province B, ...]. Output only the province ranking, do not add any other explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": [
+    "Hong Kong SAR",
+    "Guangdong Province",
+    "Zhejiang Province",
+    "Beijing",
+    "Jilin Province"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter from regional_industry_status.csv all records with industry=\"Real Estate\", extract province and total net profit amount fields, finding data for 34 provinces/regions in the Real Estate industry.",
+    "Filter province records where total net profit amount is not null, totaling 34 valid records. Sort all provincial data in descending order by total net profit amount to determine the profitability ranking of each province.",
+    "Extract the top five provinces and their total net profit amounts: Hong Kong SAR (86497465420.52 CNY), Guangdong Province (70559018502.22 CNY), Zhejiang Province (12297928184.06 CNY), Beijing (8975618268.55 CNY), Jilin Province (675521026.00 CNY)."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Total net profit amount of Hong Kong SAR (CNY)": 86497465420.52,
+    "Total net profit amount of Guangdong Province (CNY)": 70559018502.22,
+    "Total net profit amount of Zhejiang Province (CNY)": 12297928184.06,
+    "Total net profit amount of Beijing (CNY)": 8975618268.55,
+    "Total net profit amount of Jilin Province (CNY)": 675521026.0,
+    "Top five provinces by ranking": [
+      "Hong Kong SAR",
+      "Guangdong Province",
+      "Zhejiang Province",
+      "Beijing",
+      "Jilin Province"
+    ]
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/easy006.json b/assets/qa_gold/comprehensive_decision/easy006.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc13845e0aa6466c12baecddcaf1557eed0a6892
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/easy006.json
@@ -0,0 +1,21 @@
+{
+  "id": "easy006",
+  "question": "In 2022, what is Sichuan Province's national ranking by average R&D investment in the Information Transmission, Software and Information Technology Services industry?",
+  "guidelines": "The answer must be an exact number representing the ranking. Output only the number, do not add units, commas, or any textual explanation. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 10,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter from regional_industry_status.csv all records with industry=\"Information Transmission, Software and Information Technology Services\", finding data for 34 provinces/regions.",
+    "Extract the \"mean R&D investment amount\" field for each province to obtain mean R&D investment data for all provinces. Sort all provinces by mean R&D investment amount in descending order to determine the R&D investment ranking of each province.",
+    "Locate Sichuan Province's position in the sorted list. Sichuan Province's mean R&D investment amount is 147274301.44 CNY, and its ranking is 10th."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Mean R&D investment amount of Sichuan Province (CNY)": 147274301.44,
+    "Sichuan Province ranking": 10
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard001.json b/assets/qa_gold/comprehensive_decision/hard001.json
new file mode 100644
index 0000000000000000000000000000000000000000..7fb858fce7bfc34749c84f4cf631e9b55b84ae47
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard001.json
@@ -0,0 +1,31 @@
+{
+  "id": "hard001",
+  "question": "In 2022, a strategic consulting firm was commissioned by a provincial government to quantitatively rank the comprehensive attractiveness of pharmaceutical manufacturing across provinces, in order to identify priority target regions for attracting leading enterprises. The company designed a four-dimensional weighted scoring system: four original indicators—enterprise agglomeration level (weight 30%), R&D expenditure as a share of revenue (weight 30%), regional policy coverage intensity (weight 20%), and R&D human resource penetration rate (weight 20%)—were normalized (min-max) and then weighted to produce a composite score. Among these, agglomeration level is measured by the proportion of enterprises in each province to the national total in pharmaceutical manufacturing; policy intensity is measured by the ratio of relevant policy items in each province to the total number of relevant policies nationwide; and human resource penetration rate is the total number of R&D personnel in each province divided by total employees. What is the specific composite score value of the province with the highest weighted composite score after normalization across provinces?",
+  "guidelines": "The answer should be a numerical value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 0.92,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter records with industry=\"Pharmaceutical Manufacturing\" from regional_industry_status.csv, extract province, total enterprises, total R&D expenditure, total operating revenue, total R&D personnel, and total employees; 34 provincial records were found.",
+    "Filter records with industry=\"Pharmaceutical Manufacturing\" from national_industry_status.csv to obtain the national total of 449 pharmaceutical manufacturing enterprises.",
+    "Filter policy records from policy_release_status.csv where the industry field contains \"Pharmaceutical Manufacturing\"; 80 relevant policies were found across 22 provinces. Group by province to count policy numbers per province.",
+    "Calculate four original indicators for each province: industry agglomeration = total enterprises/449, R&D intensity = total R&D expenditure/total operating revenue, policy support = province policy count/80, talent density = total R&D personnel/total employees. Filter to 16 valid provinces with all four indicators non-null.",
+    "Apply min-max normalization to each of the four indicators across the 16 valid provinces: normalized value = (original value - min)/(max - min).",
+    "Calculate composite score = normalized industry agglomeration × 0.3 + normalized R&D intensity × 0.3 + normalized policy support × 0.2 + normalized talent density × 0.2.",
+    "Sort by composite score in descending order. Shanghai has the highest composite score, with industry agglomeration 0.1203, R&D intensity 0.2548, policy support 0.1375, talent density 0.1620, and composite score = 0.9160."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "National total pharmaceutical manufacturing enterprises": 449.0,
+    "National total pharmaceutical manufacturing-related policies": 80,
+    "Number of valid provinces": 16,
+    "Shanghai industry agglomeration": 0.1203,
+    "Shanghai R&D intensity": 0.2548,
+    "Shanghai policy support": 0.1375,
+    "Shanghai talent density": 0.162,
+    "Shanghai composite score": 0.916
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard002.json b/assets/qa_gold/comprehensive_decision/hard002.json
new file mode 100644
index 0000000000000000000000000000000000000000..2591fff913d5b0959540853062a22aeca297cf7a
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard002.json
@@ -0,0 +1,28 @@
+{
+  "id": "hard002",
+  "question": "In 2022, conduct a quantitative assessment of the investment value of the semiconductor industry across provinces. The evaluation framework requires incorporating three dimensions: first, industry scale (40% weight), measured by the inter-provincial rank percentile of total operating revenue in each province; second, profitability quality (30% weight), reflected by the inter-provincial rank percentile of operating profit margin (total operating profit divided by total operating revenue) in each province; third, technology output intensity (30% weight), measured by the inter-provincial rank percentile of the ratio of total patent applications to R&D expenditure (converted to 100 million yuan) in each province. The rank percentile for each indicator is calculated by sorting values from low to high, using the formula (rank - 1) / (total number of provinces - 1). Note that only provinces with complete data for all three indicators are included in the calculation. Under this weighted scoring system, what is the final score of the province ranked first in comprehensive investment value?",
+  "guidelines": "The answer should be a numerical value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 0.67,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter records with industry=\"Semiconductor Industry\" from regional_industry_status.csv, extract province, total operating revenue, total operating profit, and total R&D expenditure; 34 records were found.",
+    "Filter enterprises with industry=\"Semiconductor Industry\" from company_profile.csv, extract company name, bmCode, and province; 172 enterprises were found.",
+    "Join with company_operation_status.csv via bmCode to extract annual domestic invention patent applications. Filter 149 valid records with non-null values, group by province to sum annual domestic invention patent applications; 22 provinces have patent data.",
+    "Inner join regional data with enterprise-level patent summary by province, filter records with total operating revenue > 0 and total R&D expenditure > 0; 13 valid provinces. Calculate three original indicators: industry scale = total operating revenue, profitability = total operating profit / total operating revenue, innovation output = total annual domestic invention patent applications / total R&D expenditure (in 100 million yuan).",
+    "Rank each indicator by value from low to high (rank method = min), calculate rank percentile = (rank - 1) / (13 - 1).",
+    "Calculate investment value composite score = industry scale rank percentile × 0.4 + profitability rank percentile × 0.3 + innovation output rank percentile × 0.3.",
+    "Sort by composite score in descending order. Zhejiang Province has the highest composite score, with industry scale rank percentile 0.6667, profitability rank percentile 0.5000, innovation output rank percentile 0.8333, and composite score = 0.6667."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Number of valid provinces": 13,
+    "Zhejiang Province industry scale rank percentile": 0.6667,
+    "Zhejiang Province profitability rank percentile": 0.5,
+    "Zhejiang Province innovation output rank percentile": 0.8333,
+    "Zhejiang Province composite score": 0.6667
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard003.json b/assets/qa_gold/comprehensive_decision/hard003.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d018c0741e278dccb0a59ac2cdd42bf2ac9809c
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard003.json
@@ -0,0 +1,36 @@
+{
+  "id": "hard003",
+  "question": "In 2022, an automotive manufacturing enterprise commissioned a third-party institution to score and rate the industrial supporting capacity of each province before selecting a site for a new plant. The scoring rules are as follows: first, rank provinces in descending order by the number of government policies related to automotive manufacturing, and take the top five provinces by policy count as the candidate pool; then, within the candidate provinces, calculate the industrial supporting composite index, which is a weighted combination of three components—upstream and downstream supply chain density (weight 0.4), local labor reserve (weight 0.3), and government subsidy intensity per enterprise (weight 0.3). Supply chain density is defined as the ratio of total automotive manufacturing enterprises in the province to the national total in the industry; labor reserve is defined as the ratio of total industry employees in the province to the national total in the industry; subsidy intensity is defined as total government rewards and subsidies for automotive manufacturing in the province divided by the number of enterprises in the province (subsidy intensity must be normalized across all provinces before being used in the formula). Among the top five provinces by policy ranking, what is the composite index value of the province with the highest industrial supporting composite index?",
+  "guidelines": "Answer format: numerical value (4 decimal places). Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 0.3187,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter policy records from policy_release_status.csv where industry field contains \"Automotive Manufacturing\"; 69 records found. Group by province to count policy numbers per province (excluding \"National\" level), sort by policy count descending; top 5 provinces by policy support are ['Guangdong Province', 'Shanghai', 'Hunan Province', 'Sichuan Province', 'Chongqing'].",
+    "Filter records with industry=\"Automotive Manufacturing\" from regional_industry_status.csv, extract province, total enterprises, total employees, and total government rewards and subsidies; 34 records found.",
+    "Filter records with industry=\"Automotive Manufacturing\" from national_industry_status.csv to obtain national total of 230 automotive manufacturing enterprises and 3,254,510 total employees.",
+    "Calculate three original indicators for each province: upstream-downstream enterprise density = total enterprises/230, talent reserve = total employees/3,254,510, subsidy intensity = total government rewards and subsidies/total enterprises.",
+    "Apply min-max normalization to subsidy intensity: normalized value = (subsidy intensity - 5468453.88)/(476294284.71 - 5468453.88).",
+    "Calculate industrial supporting composite index = upstream-downstream enterprise density × 0.4 + talent reserve × 0.3 + normalized subsidy intensity × 0.3.",
+    "Filter among top 5 policy provinces ['Guangdong Province', 'Shanghai', 'Hunan Province', 'Sichuan Province', 'Chongqing'] (Chongqing excluded due to missing data for composite index calculation). Guangdong Province has the highest composite index: upstream-downstream enterprise density 0.1174, talent reserve 0.4499, normalized subsidy intensity 0.4558, composite index = 0.3187."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "National total automotive manufacturing enterprises": 230.0,
+    "National total automotive manufacturing employees": 3254510.0,
+    "Top 5 provinces by policy": [
+      "Guangdong Province",
+      "Shanghai",
+      "Hunan Province",
+      "Sichuan Province",
+      "Chongqing"
+    ],
+    "Guangdong Province upstream-downstream enterprise density": 0.1174,
+    "Guangdong Province talent reserve": 0.4499,
+    "Guangdong Province normalized subsidy intensity": 0.4558,
+    "Guangdong Province composite index": 0.3187
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard004.json b/assets/qa_gold/comprehensive_decision/hard004.json
new file mode 100644
index 0000000000000000000000000000000000000000..daa00ac93dfab02184d08fc57f5332035286da56
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard004.json
@@ -0,0 +1,31 @@
+{
+  "id": "hard004",
+  "question": "In 2022, a provincial development and reform commission, when reviewing the effectiveness of fiscal subsidies for the chemical raw materials and chemical products manufacturing industry, needed to identify enterprises with misallocated subsidy resources. Specifically, analysts must first define the scope: only examine enterprises located in provinces that have policy entries for \"Chemical Raw Materials and Chemical Products Manufacturing\" in the policy release status data; then use the industry-wide median of government subsidies and the median operating profit margin (profit margin = operating profit ÷ operating revenue × 100%) as dual thresholds to identify \"capital misallocation\" enterprises—those that simultaneously have \"subsidy amount above the industry median\" but \"profit margin below the industry median\". Among the valid enterprises in the policy-covered provinces, what is the proportion of capital misallocation enterprises as a percentage of total valid enterprises in those provinces (express the result as a percentage with 2 decimal places, without the % symbol)?",
+  "guidelines": "The answer should be a percentage value with 2 decimal places. Output only the number, without the % symbol or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 23.18,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter policy records from policy_release_status.csv where industry field contains \"Chemical Raw Materials and Chemical Products Manufacturing\"; 60 records found. Extract province field, remove nulls and exclude \"National\" level; 23 policy-covered provinces: ['Shanghai', 'Yunnan Province', 'Inner Mongolia Autonomous Region', 'Sichuan Province', 'Ningxia Hui Autonomous Region', 'Anhui Province', 'Shandong Province', 'Shanxi Province', 'Guangdong Province', 'Guangxi Zhuang Autonomous Region', 'Xinjiang Uygur Autonomous Region', 'Jiangxi Province', 'Hebei Province', 'Henan Province', 'Hainan Province', 'Hubei Province', 'Hunan Province', 'Gansu Province', 'Fujian Province', 'Guizhou Province', 'Liaoning Province', 'Shaanxi Province', 'Heilongjiang Province'].",
+    "Filter enterprise records with industry=\"Chemical Raw Materials and Chemical Products Manufacturing\" from company_profile.csv, extract company name, bmCode, and province; 364 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract government rewards and subsidies, operating profit, and operating revenue; 364 records after merge.",
+    "Filter valid samples with non-null values for government rewards and subsidies, operating profit, and operating revenue; 362 enterprises.",
+    "Calculate industry-wide median benchmarks for valid enterprises: operating profit margin = operating profit/operating revenue × 100%; median government subsidy is 10,019,029.08 yuan, median operating profit margin is 10.00%.",
+    "Filter valid enterprises whose province is in the policy-covered province list; 233 enterprises.",
+    "Among the 233 valid enterprises in policy-covered provinces, filter \"high subsidy, low output\" enterprises with government subsidy > 10,019,029.08 and operating profit margin < 10.00%; 54 enterprises, proportion = 54/233 × 100% = 23.18%."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Number of policy-covered provinces": 23,
+    "Total chemical enterprises": 364,
+    "Number of valid samples": 362,
+    "Median government subsidy (yuan)": 10019029.08,
+    "Median operating profit margin (%)": 10.0,
+    "Valid enterprises in policy-covered provinces": 233,
+    "High subsidy low output enterprises": 54,
+    "Proportion (%)": 23.18
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard005.json b/assets/qa_gold/comprehensive_decision/hard005.json
new file mode 100644
index 0000000000000000000000000000000000000000..7aea754da28a467f2c0a40807658e438a0fc31a4
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard005.json
@@ -0,0 +1,30 @@
+{
+  "id": "hard005",
+  "question": "In 2022, for the information transmission, software and information technology services industry, an industry research institute sought to obtain a policy-adjusted comprehensive innovation efficiency indicator by superimposing the incentive effect of local policy support on top of raw innovation efficiency. The calculation logic is as follows: first, exclude from enterprise microdata any samples with missing R&D expenditure or annual domestic invention patent grants; for the remaining valid enterprises, aggregate by province and calculate the ratio of total invention patent grants to total R&D expenditure (converted to 100 million yuan) for each province as the province's raw innovation efficiency benchmark; then use the proportion of policy items in that province out of all information technology policies as the policy support coefficient, and multiply the raw efficiency benchmark by (1 plus the policy support coefficient) to obtain the final policy-adjusted innovation efficiency. Among all provinces with data, what is the specific value of the province with the highest adjusted efficiency?",
+  "guidelines": "The answer should be a numerical value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 63.74,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter policy records from policy_release_status.csv where industry field contains \"Information Transmission, Software and Information Technology Services\"; 206 records found. Group by province to count policy numbers per province.",
+    "Filter enterprise records with industry=\"Information Transmission, Software and Information Technology Services\" from company_profile.csv, extract company name, bmCode, and province; 644 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract R&D expenditure and annual domestic invention patent grants; 644 records after merge.",
+    "Filter valid enterprises with non-null values for both R&D expenditure and annual domestic invention patent grants; 432 enterprises.",
+    "Group by province to sum R&D expenditure and annual domestic invention patent grants; 28 provinces have valid data.",
+    "Convert total R&D expenditure per province to 100 million yuan, calculate raw innovation efficiency = total patent grants / total R&D expenditure (100 million yuan). Merge with policy data, calculate policy support coefficient = province policy count / 206.",
+    "Calculate policy-adjusted innovation efficiency = raw innovation efficiency × (1 + policy support coefficient), sort by adjusted efficiency descending. Hong Kong Special Administrative Region has the highest: raw efficiency 63.7360, policy support coefficient 0.0000, adjusted efficiency = 63.7360."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Total information technology-related policies": 206,
+    "Total information technology services enterprises": 644,
+    "Number of valid enterprises": 432,
+    "Number of valid provinces": 28,
+    "Hong Kong SAR raw innovation efficiency": 63.736,
+    "Hong Kong SAR policy support coefficient": 0.0,
+    "Hong Kong SAR adjusted innovation efficiency": 63.74
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard006.json b/assets/qa_gold/comprehensive_decision/hard006.json
new file mode 100644
index 0000000000000000000000000000000000000000..f511de3e6611c0142d338121bdbe1630897b1180
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard006.json
@@ -0,0 +1,32 @@
+{
+  "id": "hard006",
+  "question": "In 2022, to measure the impact of different ownership backgrounds on the operating performance of specialized equipment manufacturing enterprises, an analysis team compared each enterprise's return on equity (ROE) level with the industry-wide return level in its province to calculate \"excess ROE\" as a relative performance indicator. Specifically: enterprise ROE is calculated as net profit divided by net assets (total assets minus total liabilities) multiplied by 100%; provincial industry benchmark ROE is extracted from provincial industry summary tables, calculated as total industry net profit divided by total industry net assets (total assets minus total liabilities) multiplied by 100%; each enterprise's excess ROE is the difference between its own ROE and its province's benchmark ROE. After grouping by ownership type, which ownership category has the highest mean excess ROE among enterprises? What is that mean value in percentage points?",
+  "guidelines": "The answer should be a percentage value with 2 decimal places. Output only the number, without the % symbol or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": [
+    "Collective Enterprise",
+    5.17
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter enterprise records with industry=\"Specialized Equipment Manufacturing\" from company_profile.csv, extract company name, bmCode, ownership, and province; 447 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract net profit, total assets, total liabilities, and operating revenue; 447 records after merge.",
+    "Filter records with industry=\"Specialized Equipment Manufacturing\" from regional_industry_status.csv, calculate provincial industry average ROE = total net profit / (total assets - total liabilities) × 100%; 15 valid provinces.",
+    "Filter valid enterprises with total assets > total liabilities and non-null operating revenue; 444 enterprises.",
+    "Calculate net assets = total assets - total liabilities for each enterprise, then ROE = net profit / net assets × 100%.",
+    "Inner join enterprise data with provincial industry average ROE by province; 389 enterprises matched. Calculate excess ROE = enterprise ROE - provincial industry average ROE for each enterprise.",
+    "Group by ownership to calculate mean excess ROE for each ownership type; 6 ownership types. Collective enterprises (2 enterprises) have the highest mean excess ROE = 5.17%."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Total specialized equipment manufacturing enterprises": 447,
+    "Number of valid enterprises": 444,
+    "Enterprises matched with provincial data": 389,
+    "Number of ownership types": 6,
+    "Collective enterprise count": 2,
+    "Collective enterprise mean excess ROE (%)": 5.17
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard007.json b/assets/qa_gold/comprehensive_decision/hard007.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa0a7999a0fa46e0f359972e8adf8e45b8991810
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard007.json
@@ -0,0 +1,39 @@
+{
+  "id": "hard007",
+  "question": "In 2022, when a research institution was reviewing the implementation effectiveness of provincial industrial policies in the food and beverage industry, it found that although some provinces had issued many support policies, the profitability of enterprises within their jurisdictions was not ideal. To identify such \"policy-heavy, low-return\" provinces, the institution planned to analyze separately those provinces with a relatively large number of policies (including national-level policies, totaling 3 or more): sum the operating profit amounts of all food and beverage industry enterprises in these provinces and divide by the sum of operating revenue amounts to obtain the comprehensive operating profit margin for each province, then identify the province with the lowest profit margin. What is the profit margin value (as a percentage, rounded to two decimal places) for the province with the lowest comprehensive operating profit margin?",
+  "guidelines": "The answer should be a percentage value with 2 decimal places. Output only the number, without the % symbol or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 0.59,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter policy records with industry field containing \"Food and Beverage Industry\" from policy_release_status.csv, 16 records in total. Group by province field to count policy numbers per province; 4 national-level policies need to be added to each province. After adding national-level policy count to each province's count, filter provinces with policy count >= 3, totaling 9 provinces: ['Yunnan', 'Sichuan', 'Ningxia', 'Hebei', 'Henan', 'Hainan', 'Hunan', 'Gansu', 'Guizhou'].",
+    "Filter all enterprise records with industry=\"Food and Beverage Industry\" from company_profile.csv, extract company name, bmCode, and province fields; 247 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract operating profit amount and operating revenue amount fields.",
+    "Filter valid enterprises with non-null operating revenue amount; 247 enterprises in total.",
+    "Group by province field, aggregate the sum of operating profit amounts and sum of operating revenue amounts for each province.",
+    "Calculate comprehensive operating profit margin for each province = sum of operating profit amounts / sum of operating revenue amounts × 100%.",
+    "Among the 9 provinces with >=3 policies, sort by operating profit margin in ascending order; the province with the lowest operating profit margin is Hainan, with total operating profit of 78,834,917.61 yuan, total operating revenue of 13,274,274,000.99 yuan, and operating profit margin = 0.59%."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Total food and beverage industry policies": 16,
+    "Provinces with >=3 policies": [
+      "Yunnan",
+      "Sichuan",
+      "Ningxia",
+      "Hebei",
+      "Henan",
+      "Hainan",
+      "Hunan",
+      "Gansu",
+      "Guizhou"
+    ],
+    "Number of valid enterprises": 247,
+    "Hainan total operating profit": 78834917.61,
+    "Hainan total operating revenue": 13274274000.99,
+    "Hainan operating profit margin (%)": 0.59
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard008.json b/assets/qa_gold/comprehensive_decision/hard008.json
new file mode 100644
index 0000000000000000000000000000000000000000..efec213fdb26ad3548fde4390b48f2123e80cc96
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard008.json
@@ -0,0 +1,29 @@
+{
+  "id": "hard008",
+  "question": "In 2022, a private equity institution sought to identify high-quality provinces in the electricity, heat, gas and water production and supply industry that combine growth potential, market undervaluation, and innovation resilience. The screening logic has three layers: The first layer requires that the median year-over-year change in operating revenue of enterprises within the province be positive (>0%), to exclude regions where revenue is already shrinking; The second layer, based on the first layer results, further requires that the province's market valuation level be relatively low, i.e., the P/S ratio of all enterprises in the province must be lower than the median P/S ratio across all provinces nationwide (national median is calculated from the provincial P/S ratio series); The third layer adds an innovation requirement, i.e., the mean R&D investment ratio of enterprises in the province must be higher than the mean of all enterprises in the industry with R&D investment ratio records. How many provinces satisfy all three conditions simultaneously?",
+  "guidelines": "The answer should be an integer. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 4,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all provincial records with industry=\"Electricity, Heat, Gas and Water Production and Supply\" from regional_industry_status.csv, extract province, median year-over-year change in operating revenue, total company market cap, total operating revenue amount, and number of enterprises; 34 records in total.",
+    "Filter enterprises with industry=\"Electricity, Heat, Gas and Water Production and Supply\" from company_profile.csv, extract company name, bmCode, and province fields; 189 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract R&D investment ratio field. 122 enterprises have non-null R&D investment ratio; national average R&D investment ratio is 1.1470%. Group by province to calculate average R&D investment ratio per province.",
+    "Filter 16 valid provinces with non-null and >0 total operating revenue amount; calculate P/S ratio per province = total company market cap / total operating revenue amount (converted to 100 million yuan); national median P/S ratio is 1.024358.",
+    "Filter high-growth provinces with median year-over-year change in operating revenue > 0%; 15 provinces in total.",
+    "Among high-growth provinces, filter low-valuation provinces with P/S ratio < national median 1.024358; 7 provinces in total.",
+    "Among high-growth, low-valuation provinces, further filter provinces where average enterprise R&D investment ratio > 1.1470%; 4 provinces ultimately satisfy all three conditions: ['Guangdong', 'Shanghai', 'Henan', 'Hebei']."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Number of valid provinces": 16,
+    "National median P/S ratio": 1.024358,
+    "National average enterprise R&D investment ratio (%)": 1.147,
+    "Number of high-growth provinces": 15,
+    "Number of high-growth, low-valuation provinces": 7,
+    "Number of provinces meeting all conditions": 4
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard009.json b/assets/qa_gold/comprehensive_decision/hard009.json
new file mode 100644
index 0000000000000000000000000000000000000000..b49de89b9e7a9bc7506165b2954ad873ff4dd322
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard009.json
@@ -0,0 +1,31 @@
+{
+  "id": "hard009",
+  "question": "In 2022, an investment manager at a merger and acquisition fund was seeking \"high R&D, low valuation\" M&A targets in the textile, footwear and apparel industry, but the scope was limited to provinces covered by textile, footwear and apparel industry-related policies. The prerequisite for screening valid enterprises is: net profit amount strictly greater than zero, and both R&D investment ratio and company market cap fields have data records. On this basis, first use all valid enterprises in the industry as the benchmark population to calculate the median R&D investment ratio and the median P/E ratio respectively; then from the subset of valid enterprises located in policy-covered provinces, filter enterprises whose R&D investment ratio is higher than the industry median and whose P/E ratio is lower than the industry median. How many enterprises satisfy the above dual screening conditions? (P/E ratio = company market cap (100 million yuan) ÷ net profit amount (100 million yuan))",
+  "guidelines": "The answer should be an integer. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 9,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter policy records with industry field containing \"Textile, Footwear and Apparel\" from policy_release_status.csv, 21 records in total. Extract province field, remove nulls and exclude \"National\" level, to obtain 11 policy-covered provinces: ['Shanghai', 'Sichuan', 'Shandong', 'Guangdong', 'Guangxi', 'Xinjiang', 'Hebei', 'Hunan', 'Fujian', 'Liaoning', 'Shaanxi'].",
+    "Filter all enterprise records with industry=\"Textile, Footwear and Apparel\" from company_profile.csv, extract company name, bmCode, and province fields; 177 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract R&D investment ratio, net profit amount, and company market cap fields; 177 records after merge.",
+    "Filter valid enterprises with net profit amount > 0 and both R&D investment ratio and company market cap non-null; 81 enterprises in total.",
+    "Confirm data units: company market cap is in 100 million yuan, net profit amount is in yuan. Unify units for P/E calculation: P/E = company market cap (100 million yuan) ÷ (net profit amount (yuan) ÷ 100000000), i.e., P/E = company market cap (100 million yuan) ÷ net profit amount (100 million yuan).",
+    "Calculate industry-wide median benchmarks for valid enterprises: median R&D investment ratio is 2.8, median P/E = company market cap (100 million yuan) / net profit amount (100 million yuan) is 23.13.",
+    "Among valid enterprises, filter those whose province is in the policy-covered province list; 29 enterprises in total.",
+    "Among the 29 valid enterprises in policy-covered provinces, filter enterprises with R&D investment ratio > 2.8 and P/E < 23.13; 9 enterprises in total."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "Number of policy-covered provinces": 11,
+    "Total textile, footwear and apparel enterprises": 177,
+    "Number of valid enterprises": 81,
+    "Median R&D investment ratio": 2.8,
+    "Median P/E": 23.13,
+    "Valid enterprises in policy-covered provinces": 29,
+    "Number of enterprises meeting conditions": 9
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard010.json b/assets/qa_gold/comprehensive_decision/hard010.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb55ccd7473174db94c3c0f91e380853ffb4c510
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard010.json
@@ -0,0 +1,30 @@
+{
+  "id": "hard010",
+  "question": "In 2022, to quantify the comprehensive competitive strength of the construction industry across regions, an industry association constructed a provincial competitiveness index system. The index is composed of four weighted sub-dimensions: market size share of national total (weight 30%), asset operation efficiency i.e. operating profit to total assets ratio (weight 30%), technology accumulation level i.e. cumulative invention patent grants to number of enterprises in jurisdiction ratio (weight 20%), and talent structure i.e. R&D personnel as share of total employees (weight 20%). The four raw indicators are each min-max normalized across all valid provinces, then weighted and summed to obtain the final index. Only provinces with data records for all four indicators are included in the calculation. Finally, please calculate the index difference between the first-ranked province and the last-ranked province in the competitiveness index ranking (rounded to two decimal places).",
+  "guidelines": "The answer should be a numeric value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 0.71,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all provincial records with industry=\"Construction\" from regional_industry_status.csv, extract province, total operating revenue amount, total operating profit amount, total assets, total cumulative Chinese invention patent grants, number of enterprises, total R&D personnel count, and total employee count fields; 34 records in total.",
+    "Filter records with industry=\"Construction\" from national_industry_status.csv to obtain national construction industry total operating revenue amount of 12,683,425,500,139.00 yuan and total number of enterprises 148.",
+    "Filter enterprises with industry=\"Construction\" from company_profile.csv, 148 enterprises in total, for cross-validation of provincial-level data.",
+    "Calculate four raw indicators per province: scale index = total operating revenue amount / 12,683,425,500,139.00, efficiency index = total operating profit amount / total assets, innovation index = total cumulative Chinese invention patent grants / number of enterprises, talent index = total R&D personnel count / total employee count. Filter 16 valid provinces with all four indicators non-null.",
+    "Apply min-max normalization to each of the four indicators: normalized value = (raw value - min) / (max - min).",
+    "Calculate competitiveness index = normalized scale index × 0.3 + normalized efficiency index × 0.3 + normalized innovation index × 0.2 + normalized talent index × 0.2.",
+    "Sort by competitiveness index in descending order; the difference between first-ranked Beijing (0.7287) and last-ranked Liaoning (0.0142) = 0.7145."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Number of valid provinces": 16,
+    "National total construction enterprises": 148,
+    "First-ranked province": "Beijing",
+    "First-ranked score": 0.7287,
+    "Last-ranked province": "Liaoning",
+    "Last-ranked score": 0.0142,
+    "Difference": 0.7145
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard011.json b/assets/qa_gold/comprehensive_decision/hard011.json
new file mode 100644
index 0000000000000000000000000000000000000000..d8bfc1e06aff2a73aa1fefbec33657de2de12590
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard011.json
@@ -0,0 +1,50 @@
+{
+  "id": "hard011",
+  "question": "In 2022, a think tank was commissioned to study the impact of policy intervention on R&D behavior in the communication transmission equipment industry. The research design divides all enterprises with R&D investment ratio data records into two groups: one group from provinces that have appeared in policy release information with \"Communication Transmission Equipment\" related policy entries (\"National\" level entries do not count as provinces and are not included in either group); the other group from provinces that have never appeared in the above policy entries. After grouping, calculate the arithmetic mean of R&D investment ratio for each group respectively, then compute the difference between them (policy-covered provinces mean minus non-policy-covered provinces mean). This difference reflects the association between policy coverage and R&D intensity of communication transmission equipment enterprises within the jurisdiction. What is this difference in percentage points?",
+  "guidelines": "The answer should be a numeric value with 2 decimal places. A positive number indicates policy-covered provinces are higher; a negative number indicates non-policy-covered provinces are higher. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 4.9,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter policy records with industry field containing \"Communication Transmission Equipment\" from policy_release_status.csv, 70 records in total. Extract unique province values and exclude \"National\" to obtain 17 policy-covered provinces: ['Anhui', 'Shandong', 'Guangdong', 'Sichuan', 'Hubei', 'Fujian', 'Jiangxi', 'Chongqing', 'Hunan', 'Yunnan', 'Guizhou', 'Henan', 'Shaanxi', 'Hainan', 'Beijing', 'Shanghai', 'Xinjiang'].",
+    "Filter enterprise records with industry=\"Communication Transmission Equipment\" from company_profile.csv, extract company name, bmCode, and province fields; 120 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract R&D investment ratio field; 120 records after merge.",
+    "Filter valid enterprises with non-null R&D investment ratio; 117 enterprises in total.",
+    "Based on the policy-covered province list from step 1, divide valid enterprises into two groups: 86 enterprises in policy-covered provinces, 31 enterprises in non-policy-covered provinces.",
+    "Calculate mean R&D investment ratio for each group: policy-covered provinces average = 14.35%, non-policy-covered provinces average = 9.45%.",
+    "Calculate difference = policy-covered average R&D ratio - non-policy-covered average R&D ratio = 14.35 - 9.45 = 4.90 percentage points."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Number of communication transmission equipment policies": 70,
+    "Policy-covered provinces": [
+      "Anhui",
+      "Shandong",
+      "Guangdong",
+      "Sichuan",
+      "Hubei",
+      "Fujian",
+      "Jiangxi",
+      "Chongqing",
+      "Hunan",
+      "Yunnan",
+      "Guizhou",
+      "Henan",
+      "Shaanxi",
+      "Hainan",
+      "Beijing",
+      "Shanghai",
+      "Xinjiang"
+    ],
+    "Number of communication transmission equipment enterprises": 120,
+    "Number of valid enterprises": 117,
+    "Enterprises in policy-covered provinces": 86,
+    "Enterprises in non-policy-covered provinces": 31,
+    "Policy-covered average R&D ratio (%)": 14.35,
+    "Non-policy-covered average R&D ratio (%)": 9.45,
+    "Difference (percentage points)": 4.9
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard012.json b/assets/qa_gold/comprehensive_decision/hard012.json
new file mode 100644
index 0000000000000000000000000000000000000000..6f80bd634bce8eea5fb55261382f7510633c5ddd
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard012.json
@@ -0,0 +1,29 @@
+{
+  "id": "hard012",
+  "question": "In 2022, an antitrust research team analyzed the provincial market structure of the metal smelting and rolling processing industry. To ensure statistical reliability, only provinces with operating revenue amount records and at least 5 enterprises in the industry within the jurisdiction were included. Among qualifying provinces, the Herfindahl-Hirschman Index (HHI) was used to measure market concentration in each province: calculate each enterprise's operating revenue as a share of total operating revenue of all valid enterprises in the province, sum the squares of these shares and multiply by 100% to obtain the province's HHI value. Higher HHI indicates more concentrated markets and greater monopoly risk. After identifying the province with the highest HHI, extract the province's total operating profit amount and total operating revenue amount from provincial industry summary data, and calculate the corresponding operating profit margin. What is the operating profit margin of the province with the highest HHI?",
+  "guidelines": "The answer should be a percentage value with 2 decimal places. Output only the number, without the % symbol or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 4.14,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter enterprise records with industry=\"Metal Smelting and Rolling Processing\" from company_profile.csv, extract company name, bmCode, and province fields; 145 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract operating revenue amount field; 145 records after merge.",
+    "Filter 111 enterprises with non-null operating revenue amount; group by province to count enterprises per province; retain 13 provinces with enterprise count >= 5: ['Shanghai', 'Yunnan', 'Beijing', 'Sichuan', 'Anhui', 'Shandong', 'Guangdong', 'Jiangsu', 'Jiangxi', 'Henan', 'Zhejiang', 'Liaoning', 'Hong Kong'].",
+    "Within each valid province, calculate each enterprise's market share = enterprise operating revenue amount / sum of operating revenue amounts of all enterprises in the province.",
+    "Calculate Herfindahl-Hirschman Index (HHI) for each province = sum of squares of enterprise market shares × 100%; sort by HHI in descending order.",
+    "Among qualifying provinces, using the same valid enterprise sample as for HHI calculation (non-null operating revenue and province enterprise count ≥ 5), aggregate total operating profit amount and total operating revenue amount per province; calculate operating profit margin per province = sum of operating profit amounts / sum of operating revenue amounts × 100%.",
+    "The province with the highest HHI is Shanghai, HHI = 88.47. Total operating profit of valid enterprises in this province is 16,186,839,594.21 yuan, total operating revenue is 391,233,407,230.44 yuan; operating profit margin = 4.14%."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Number of metal smelting and rolling processing enterprises": 145,
+    "Enterprises with non-null operating revenue": 111,
+    "Number of valid provinces (enterprises >= 5)": 13,
+    "Province with highest HHI": "Shanghai",
+    "HHI value": 88.47,
+    "Shanghai operating profit margin (%)": 4.14
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard013.json b/assets/qa_gold/comprehensive_decision/hard013.json
new file mode 100644
index 0000000000000000000000000000000000000000..9060172dae0a9c60a0b96424ae42ee0a6e2d5966
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard013.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard013",
+  "question": "In 2022, a provincial industry and information department sought to evaluate the government subsidy utilization efficiency of enterprises of different sizes in the rubber and plastic products industry. After dividing enterprises into three groups by total assets—large (top 1/3 rounded up), medium (middle 1/3 rounded up), and small (bottom 1/3)—which enterprise size group has the highest subsidy utilization efficiency? What is its efficiency value?",
+  "guidelines": "The answer should be a numeric value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": [
+    "Large",
+    166.33
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all enterprise records with industry=\"Rubber and Plastic Products\" from company_profile.csv, extract company name, bmCode, and province fields; 107 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract total assets, operating revenue amount, and government reward funds and subsidies fields; 107 records after merge.",
+    "Filter records with industry=\"Rubber and Plastic Products\" from national_industry_status.csv to obtain national benchmark data: total enterprises 107, total operating revenue 313,571,405,678.69 yuan, total government subsidies 1,966,394,638.00 yuan, total assets 476,387,663,666.96 yuan.",
+    "Filter valid enterprises with all three fields (total assets, operating revenue amount, government reward funds and subsidies) non-null and government reward funds and subsidies > 0; 107 enterprises in total.",
+    "Sort by total assets in descending order; divide 107 enterprises into three groups: large enterprise group (top 36, highest 1/3 by total assets), medium enterprise group (middle 36), small enterprise group (bottom 35, lowest 1/3 by total assets).",
+    "Calculate subsidy utilization efficiency per group = sum of operating revenue amounts of enterprises in the group / sum of government reward funds and subsidies of enterprises in the group. Small group efficiency = 95.58, medium group efficiency = 161.13, large group efficiency = 166.33.",
+    "The group with the highest subsidy utilization efficiency is the large enterprise group, efficiency value = 166.33."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "National total rubber and plastic enterprises": 107,
+    "Number of valid enterprises": 107,
+    "Small group enterprise count": 35,
+    "Medium group enterprise count": 36,
+    "Large group enterprise count": 36,
+    "Small group efficiency": 95.58,
+    "Medium group efficiency": 161.13,
+    "Large group efficiency": 166.33,
+    "Highest efficiency group": "Large"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard014.json b/assets/qa_gold/comprehensive_decision/hard014.json
new file mode 100644
index 0000000000000000000000000000000000000000..4954a22f6bb983ef72012b41ecaaf1f3f1259914
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard014.json
@@ -0,0 +1,30 @@
+{
+  "id": "hard014",
+  "question": "In 2022, a technology innovation fund evaluated the \"R&D-patent conversion\" full-chain efficiency of the consumer electronics and electrical industry across provinces, seeking to identify the province with optimal conversion efficiency (only provinces with valid enterprise count >= 3 are included). What is the R&D-patent conversion efficiency value of that province? (R&D-patent conversion efficiency = sum of annual Chinese invention patent grants / sum of annual Chinese invention patent applications × R&D output density; R&D output density = sum of annual Chinese invention patent applications / sum of R&D investment amount (100 million yuan))",
+  "guidelines": "The answer should be a numeric value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 47.29,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all enterprise records with industry=\"Consumer Electronics and Electrical\" from company_profile.csv, extract company name, bmCode, and province fields; 358 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract annual Chinese invention patent applications, annual Chinese invention patent grants, and R&D investment amount fields; 358 records after merge.",
+    "Filter records with industry=\"Consumer Electronics and Electrical\" from national_industry_status.csv to obtain national benchmark data: total enterprises 358, total R&D investment amount 245,156,000,000.00 yuan, total annual invention patent applications 63,940, total annual invention patent grants 43,780.",
+    "Filter 266 valid enterprises with all three fields (annual Chinese invention patent applications, annual Chinese invention patent grants, R&D investment amount) non-null; retain 13 provinces with valid enterprise count >= 3 after grouping by province.",
+    "Aggregate by province: sum of annual Chinese invention patent applications, sum of annual Chinese invention patent grants, and sum of R&D investment amount.",
+    "Calculate R&D output density per province = sum of annual Chinese invention patent applications / sum of R&D investment amount (100 million yuan); conversion efficiency = (sum of annual Chinese invention patent grants / sum of annual Chinese invention patent applications) × R&D output density.",
+    "Sort by conversion efficiency in descending order; the province with the highest conversion efficiency is Shandong, with 24,694 patent applications, 15,143 patent grants, R&D output density 77.1105, conversion efficiency = 47.29."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "National total consumer electronics and electrical enterprises": 358,
+    "Number of valid enterprises": 266,
+    "Number of valid provinces": 13,
+    "Shandong patent applications": 24694.0,
+    "Shandong patent grants": 15143.0,
+    "Shandong R&D output density": 77.1105,
+    "Shandong conversion efficiency": 47.29
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard015.json b/assets/qa_gold/comprehensive_decision/hard015.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a01d7a0920ecca42aedb7ecaf9a7a3b5783db48
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard015.json
@@ -0,0 +1,31 @@
+{
+  "id": "hard015",
+  "question": "In 2022, a provincial government evaluated the comprehensive financial health of real estate enterprises to decide which provinces (where the province has an effective enterprise count >= 3) should face strengthened risk supervision for real estate firms. What is the health score of the province with the lowest financial health? (Financial health = Profitability score × 0.4 + Solvency score × 0.3 + Growth capability score × 0.3; Profitability is measured by the average net profit margin of enterprises in that province, where net profit margin = net profit amount / operating revenue amount; Solvency is measured as 1 − the arithmetic mean of enterprises' asset-liability ratio in that province / 100; Growth capability is measured as the median of enterprises' year-over-year change in operating revenue in that province / 100; each indicator is min-max normalized across all valid provinces before being substituted into the formula.)",
+  "guidelines": "The answer should be a numerical value with 2 decimal places. Output only the number without units or text explanation. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 0.07,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter records with industry=\"房地产业\" from company_profile.csv; extract enterprise name, bmCode, and province — 295 enterprises.",
+    "Join with company_operation_status.csv on enterprise name and bmCode for year=2022; extract net profit amount, operating revenue amount, asset-liability ratio, and year-over-year change in operating revenue — 295 rows after merge.",
+    "Keep records with operating revenue amount > 0 and non-null net profit amount and asset-liability ratio — 295 rows; do not additionally exclude rows by asset-liability ratio; all participate in within-province arithmetic means and subsequent normalization.",
+    "Count enterprises by province; define valid provinces as those with effective enterprise count >= 3; in this data there are 17 valid provinces, and growth capability (median YoY operating revenue change) can be computed for all of them.",
+    "On those 17 valid provinces only, compute three raw indicators per province: Profitability = arithmetic mean over enterprises of (net profit amount / operating revenue amount); Solvency = 1 − arithmetic mean of asset-liability ratio / 100; Growth capability = median of year-over-year change in operating revenue / 100.",
+    "Apply min-max normalization to profitability, solvency, and growth capability separately across all 17 valid provinces (i.e. \"all valid provinces\" means this set only, excluding provinces with only 1–2 enterprises).",
+    "Compute financial health = normalized profitability × 0.4 + normalized solvency × 0.3 + normalized growth capability × 0.3.",
+    "Sort the 17 valid provinces by financial health ascending; the lowest is Beijing: raw profitability ≈ −1.0774, raw solvency ≈ −21.9888, raw growth capability ≈ −0.1341, health score ≈ 0.0713, rounded to two decimals as 0.07."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "real_estate_enterprises_profile_merged_2022": 295,
+    "enterprise_records_province_aggregation_no_al_ratio_exclusion": 295,
+    "valid_provinces_count_enterprises_ge_3": 17,
+    "Beijing_raw_profitability": -1.0774,
+    "Beijing_raw_solvency": -21.9888,
+    "Beijing_raw_growth_capability": -0.1341,
+    "Beijing_financial_health": 0.0713
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard016.json b/assets/qa_gold/comprehensive_decision/hard016.json
new file mode 100644
index 0000000000000000000000000000000000000000..24538a8c2e9df53f06797608a0f1cc4858108ccf
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard016.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard016",
+  "question": "In 2022, an institutional investor plans to build an equity portfolio among Haishan Chang Industrial Equipment Company, Zhongbai Jinmao Chain Company, and Sansan Dateng Heavy Industry Company. The total portfolio weight must equal 1, the portfolio-weighted asset-liability ratio must equal exactly 45%, and the portfolio-weighted year-on-year operating revenue change must equal exactly 0%. Based on these three companies' 2022 operating data, find their portfolio weights and compute the portfolio-weighted ROE. Note: each company's asset-liability ratio is computed as total liabilities ÷ total assets × 100%.",
+  "guidelines": "Answer format: weight of Haishan Chang Industrial Equipment Company, weight of Zhongbai Jinmao Chain Company, weight of Sansan Dateng Heavy Industry Company, weighted ROE. The first three weights to four decimal places; weighted ROE to three decimal places. Output numbers and commas only, with no explanatory text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": [
+    0.1318,
+    0.4954,
+    0.3728,
+    10.376
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From company_operation_status.csv, filter by bmCode for Haishan Chang Industrial Equipment Company (100071), Zhongbai Jinmao Chain Company (100120), and Sansan Dateng Heavy Industry Company (100260) to obtain their 2022 records; 3 valid rows found. Extract total assets, total liabilities, year-on-year change in operating revenue, and net profit amount.",
+    "Compute each firm's asset-liability ratio from total assets and total liabilities using total liabilities / total assets × 100%. Results: Haishan Chang Industrial Equipment Company 86.3880%, Zhongbai Jinmao Chain Company 57.6074%, Sansan Dateng Heavy Industry Company 13.6112%.",
+    "Derive shareholder equity from total assets minus total liabilities; then compute ROE = net profit amount / shareholder equity × 100%. ROE for the three firms is 15.2002%, 11.5040%, and 7.1713%, respectively.",
+    "Let the three firms' weights be w1, w2, and w3. Set up the simultaneous equations w1+w2+w3=1, 86.3880w1+57.6074w2+13.6112w3=45, and −19.28w1+12.21w2−9.41w3=0.",
+    "Solve the system of three linear equations to obtain w1=0.13180488, w2=0.49541694, w3=0.37277818.",
+    "Compute portfolio-weighted ROE=15.2002%×0.13180488+11.5040%×0.49541694+7.1713%×0.37277818; the result is 10.376%."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Firm count": 3,
+    "Haishan Chang Industrial Equipment Company asset-liability ratio": 86.388,
+    "Zhongbai Jinmao Chain Company asset-liability ratio": 57.6074,
+    "Sansan Dateng Heavy Industry Company asset-liability ratio": 13.6112,
+    "Haishan Chang Industrial Equipment Company weight": 0.1318,
+    "Zhongbai Jinmao Chain Company weight": 0.4954,
+    "Sansan Dateng Heavy Industry Company weight": 0.3728,
+    "Portfolio-weighted ROE": 10.376
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard017.json b/assets/qa_gold/comprehensive_decision/hard017.json
new file mode 100644
index 0000000000000000000000000000000000000000..277c0ee25a6c43688a146b83851b0a631b62741f
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard017.json
@@ -0,0 +1,28 @@
+{
+  "id": "hard017",
+  "question": "In 2022, in the chemical raw materials and chemical products manufacturing industry covered by the Implementation Plan for \"Three Products\" in Raw Materials Industry, Hualu Runyuan Technology Co., Ltd. plans to restore profitability through product upgrade and price increases. After implementing the \"Three Products\" reforms, the company can obtain two types of certain profit improvements: one from process and quality improvement, equal to 1.5% of that year's operating revenue; the other from special support and subsidies. Assuming sales volume is unchanged, price increases have no effect on costs, and the goal is to bring net profit exactly to zero, find the minimum price increase rate required based on the company's 2022 operating data and policy information.",
+  "guidelines": "Answer format: minimum price increase rate. Four decimal places. Output the number only, no percent sign or text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 14.2792,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From policy_resource.jsonl, confirm that the Implementation Plan for \"Three Products\" in Raw Materials Industry specifies special support and subsidy of 5,000,000 yuan.",
+    "From company_profile.csv, filter record with bmCode=100639; confirm Hualu Runyuan Technology belongs to chemical raw materials and chemical products manufacturing.",
+    "From company_operation_status.csv, filter 2022 record with bmCode=100639; extract operating revenue and net profit: operating revenue 664,310,105.79 yuan, net profit −109,823,137.00 yuan.",
+    "Process and quality improvement profit gain = 664,310,105.79 × 1.5% = 9,964,651.59 yuan; add special support and subsidy 5,000,000.00 yuan.",
+    "Remaining profit gap = 109,823,137.00 − 9,964,651.59 − 5,000,000.00 = 94,858,485.41 yuan.",
+    "Under the assumption that sales volume is unchanged and all price-increase revenue flows to profit, minimum price increase rate = 94,858,485.41 / 664,310,105.79 = 14.2792%."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Operating revenue (yuan)": 664310105.79,
+    "Net profit (yuan)": -109823137,
+    "Process improvement profit gain (yuan)": 9964651.59,
+    "Subsidy profit gain (yuan)": 5000000,
+    "Remaining profit gap (yuan)": 94858485.41,
+    "Minimum price increase rate (%)": 14.2792
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard018.json b/assets/qa_gold/comprehensive_decision/hard018.json
new file mode 100644
index 0000000000000000000000000000000000000000..423babf34c049bca1ff539097b5dfbda6511f806
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard018.json
@@ -0,0 +1,30 @@
+{
+  "id": "hard018",
+  "question": "In 2022, does Lianji Chuangji Machine Tool Company meet the basic conditions under the Several Policies on Supporting the Construction of a Strong Province of Skilled Workers? If yes, treat it as a sample firm eligible for skills training subsidies and apply the following: all R&D personnel receive individual skill improvement subsidies at the \"senior technician\" rate; all non-R&D employees at the \"technician\" rate; assume all new subsidies are used to offset that year's R&D investment. Calculate by how many basis points the adjusted R&D investment ratio is lower than the disclosed 2022 R&D investment ratio.",
+  "guidelines": "Answer format: the number of basis points of decline. Two decimal places. Output the number only, no units or text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "43.33",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From company_profile.csv, confirm Lianji Chuangji Machine Tool Company is located in Anhui Province, satisfying the geographic condition for Anhui's local skills policy.",
+    "From policy_resource.jsonl, locate the Notice of Anhui Provincial People's Government on Several Policies Supporting the Construction of a Strong Province of Skilled Workers; extract individual skill improvement subsidy rates: technician 3,500 yuan per person, senior technician 5,000 yuan per person.",
+    "From company_operation_status.csv, filter records with bmCode=100791 and year=2022; extract total employees, R&D personnel count, R&D investment amount, operating revenue, and disclosed R&D investment ratio. Results: 1,276 employees, 290 R&D personnel, R&D investment 58,357,098.35 yuan, operating revenue 1,082,827,667.36 yuan, disclosed R&D ratio 5.37%.",
+    "Non-R&D employees = 1,276 − 290 = 986.",
+    "Total new subsidies = 290 × 5,000 + 986 × 3,500 = 4,901,000.00 yuan.",
+    "Adjusted R&D investment = 58,357,098.35 − 4,901,000.00 = 53,456,098.35 yuan.",
+    "Adjusted R&D ratio = 53,456,098.35 / 1,082,827,667.36 × 100% = 4.9367%. Compared with disclosed 5.37%, decline = 5.37% − 4.9367% = 0.4333%, or 43.33 basis points."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Total employees": 1276,
+    "R&D personnel count": 290,
+    "Non-R&D employee count": 986,
+    "Total new subsidies (yuan)": 4901000,
+    "Adjusted R&D investment (yuan)": 53456098.35,
+    "Adjusted R&D ratio (%)": 4.9367,
+    "Decline (basis points)": 43.33
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/hard019.json b/assets/qa_gold/comprehensive_decision/hard019.json
new file mode 100644
index 0000000000000000000000000000000000000000..d812ad8434724dda2bb8ce1a02e39540ff73a516
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/hard019.json
@@ -0,0 +1,33 @@
+{
+  "id": "hard019",
+  "question": "2022年，以山安泽医疗科技公司具有“微生态活菌业务”和“高温合金业务”双主业特征。某基金经理希望用以山安泽医疗科技公司与三三达腾重工公司构建一个两股票组合，来替代连机创机机床公司的增长暴露，并进一步检验该替代组合在剔除补贴后的盈利质量与研发强度溢价。若组合要求加权营业收入同比增减幅恰好等于连机创机机床公司2022年的对应指标，请基于本地数据计算：以山安泽医疗科技公司的组合权重、剔除政府奖励资金和补贴后的组合加权净利率，以及该组合研发投入占比相对连机创机机床公司高出的基点数。",
+  "guidelines": "答案格式为：以山安泽医疗科技公司权重,剔除补贴后的组合加权净利率,研发投入占比高出的基点数。前两项按百分比口径保留2位小数，最后一项保留2位小数。仅输出数字和逗号，不要添加单位或文字说明。如无法找到相关数据，请回答“未查询到相关数据”。",
+  "answer": [
+    38.58,
+    14.82,
+    708.28
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "从company_core.jsonl中定位以山安泽医疗科技公司，确认其核心竞争力描述同时覆盖微生态活菌业务与高温合金业务，满足题目中的双主业前提。",
+    "从company_operation_status.csv中筛选bmCode分别为764661、100260、100791且year=2022的记录，提取营业收入同比增减幅、净利润金额、营业收入金额、政府奖励资金、补贴和研发投入占比字段。其中，以山安泽医疗科技公司的研发投入占比字段值为18.11%，三三达腾重工公司的研发投入占比字段值为8.90%，连机创机机床公司的研发投入占比字段值为5.37%。",
+    "得到以山安泽医疗科技公司营业收入同比增减幅20.92%，三三达腾重工公司营业收入同比增减幅-9.41%，连机创机机床公司营业收入同比增减幅2.29%。设以山安泽医疗科技公司权重为w，则20.92%×w+(-9.41%)×(1-w)=2.29%，解得w=38.5757%，三三达腾重工公司权重为61.4243%。",
+    "分别计算剔除补贴后的净利率。以山安泽医疗科技公司剔除补贴后的净利润=105565919.71-27049179.17=78516740.54元，剔除补贴后的净利率=78516740.54/793847988.60×100%=9.8907%。三三达腾重工公司剔除补贴后的净利润=155125599.81-11937935.97=143187663.84元，剔除补贴后的净利率=143187663.84/799392301.84×100%=17.9121%。",
+    "按组合权重计算剔除补贴后的组合加权净利率=38.5757%×9.8907%+61.4243%×17.9121%=14.8178%。",
+    "按组合权重计算组合研发投入占比=38.5757%×18.11%+61.4243%×8.90%=12.4528%。连机创机机床公司研发投入占比字段值为5.37%，因此组合研发投入占比高出12.4528%-5.37%=7.0828个百分点，即708.28个基点。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "以山安泽医疗科技公司权重(%)": 38.58,
+    "三三达腾重工公司权重(%)": 61.42,
+    "以山安泽医疗科技公司剔除补贴后净利率(%)": 9.89,
+    "三三达腾重工公司剔除补贴后净利率(%)": 17.91,
+    "剔除补贴后的组合加权净利率(%)": 14.82,
+    "组合研发投入占比(%)": 12.45,
+    "相对连机创机机床公司高出的基点数": 708.28
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium001.json b/assets/qa_gold/comprehensive_decision/medium001.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d50b792a1f26f3de1d1ef963b233c642ccf611e
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium001.json
@@ -0,0 +1,32 @@
+{
+  "id": "medium001",
+  "question": "For 2022 pharmaceutical manufacturing industry data by province, if R&D funding intensity is measured as each province's total R&D expenditure as a percentage of its total operating revenue, among all provinces with complete data records, what is the specific value of this ratio for the province with the highest level? Which company has the highest R&D funding intensity in that province?",
+  "guidelines": "The first answer is a numeric value (2 decimal places), unit is %; the second answer is the full company name, which must exactly match the \"Company Name\" field in company_profile.csv. If either question cannot be answered, respond with \"No relevant data found\".",
+  "answer": [
+    25.48,
+    "Kangsheng Anjian Biopharmaceutical Company"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter industry=\"Pharmaceutical Manufacturing\", extract province, total R&D expenditure, total operating revenue; exclude records with missing total R&D expenditure or total operating revenue, or with total operating revenue ≤ 0, yielding 16 valid province records.",
+    "For each province, compute R&D funding intensity = total R&D expenditure ÷ total operating revenue × 100%, sort by intensity descending. Shanghai has the highest: total R&D expenditure 40,798,081,760.73 yuan, total operating revenue 160,133,198,188.25 yuan, intensity = 40,798,081,760.73 ÷ 160,133,198,188.25 × 100% = 25.48%.",
+    "From company_profile.csv, filter province=\"Shanghai\" and industry=\"Pharmaceutical Manufacturing\", obtain company names and bmCode list for pharmaceutical manufacturing firms in that city.",
+    "From company_operation_status.csv, filter year=2022 and bmCode in the above set, with non-null R&D expenditure and operating revenue, and operating revenue > 0.",
+    "For each company, compute R&D funding intensity = R&D expenditure ÷ operating revenue × 100% using the same formula. The highest is \"Kangsheng Anjian Biopharmaceutical Company\" (bmCode=505404): R&D expenditure 809,733,452.00 yuan, operating revenue 12,792,315.00 yuan, intensity = 809,733,452.00 ÷ 12,792,315.00 × 100% = 6329.84%."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province with highest R&D funding intensity": "Shanghai",
+    "Shanghai total R&D expenditure (yuan)": 40798081760.73,
+    "Shanghai total operating revenue (yuan)": 160133198188.25,
+    "Provincial R&D funding intensity (%)": 25.48,
+    "Company with highest R&D funding intensity in that province": "Kangsheng Anjian Biopharmaceutical Company",
+    "Company R&D expenditure (yuan)": 809733452.0,
+    "Company operating revenue (yuan)": 12792315.0,
+    "Company R&D funding intensity (%)": 6329.84
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium002.json b/assets/qa_gold/comprehensive_decision/medium002.json
new file mode 100644
index 0000000000000000000000000000000000000000..ab47029445a044a5174498ad7b10486d2ce39511
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium002.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium002",
+  "question": "In 2022, a semiconductor company plans to expand production and wishes to locate in the province with the highest enterprise concentration to gain industrial synergy effects. What is the proportion of that province's semiconductor industry enterprise count to the national total? What proportion does that province's total operating profit in the semiconductor industry account for of the national semiconductor industry's total operating profit?",
+  "guidelines": "Two answers required, both numeric values (2 decimal places), unit is %. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": [
+    31.4,
+    6.22
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter records with industry=\"Semiconductor Industry\", extract province, total enterprise count, total operating profit; 34 province records obtained.",
+    "From national_industry_status.csv, filter records with industry=\"Semiconductor Industry\", extract total enterprise count and total operating profit: national total 172 enterprises, national total operating profit 411,298,557,285.26 yuan.",
+    "Sort by total enterprise count in regional_industry_status.csv descending; Guangdong Province has the highest with 54 enterprises; corresponding total operating profit 25,562,691,329.46 yuan.",
+    "Compute enterprise concentration = Guangdong enterprise count / national enterprise count × 100% = 54 / 172 × 100% = 31.40%.",
+    "Compute that province's semiconductor industry operating profit share = Guangdong total operating profit / national total operating profit × 100% = 25,562,691,329.46 / 411,298,557,285.26 × 100% = 6.22%."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province with highest enterprise concentration": "Guangdong Province",
+    "Guangdong semiconductor industry enterprise count": 54,
+    "National semiconductor industry enterprise count": 172,
+    "Enterprise concentration (%)": 31.4,
+    "Guangdong semiconductor industry total operating profit (yuan)": 25562691329.46,
+    "National semiconductor industry total operating profit (yuan)": 411298557285.26,
+    "Guangdong operating profit share (%)": 6.22
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium003.json b/assets/qa_gold/comprehensive_decision/medium003.json
new file mode 100644
index 0000000000000000000000000000000000000000..cac69c121b2e6b51ae2a529e2cab88e50183a461
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium003.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium003",
+  "question": "In 2022, measuring per capita output efficiency of automobile manufacturing by province using revenue per capita (total operating revenue ÷ total employee count), among all provinces nationwide, what is the specific value in yuan per person for the province with the highest indicator? Compared to the national average, by what percentage (1 decimal place) is that province's per capita revenue higher?",
+  "guidelines": "Two answers required: first is a numeric value (2 decimal places), unit yuan/person; second is a percentage (1 decimal place), unit %. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": [
+    3898878.23,
+    175.0
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter all province records with industry=\"Automobile Manufacturing\", extract province, total operating revenue, total employee count; 34 records obtained.",
+    "Filter province records with non-null total operating revenue and total employee count, and total employee count > 0; 14 valid provinces obtained.",
+    "Compute per capita revenue for each valid province = total operating revenue / total employee count, sort by this indicator descending; Beijing has the highest: 341,596,317,171.50 / 87,614 = 3,898,878.23 yuan/person.",
+    "From national_industry_status.csv, filter national records with industry=\"Automobile Manufacturing\", extract total operating revenue and total employee count: national total operating revenue 4,614,449,954,119.47 yuan, national total employee count 3,254,510, national average per capita revenue = 1,417,863.20 yuan/person.",
+    "Compute Beijing's excess over national average = (3,898,878.23 - 1,417,863.20) / 1,417,863.20 × 100% = 174.98%, rounded to 1 decimal place per requirement: 175.0%."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Beijing total operating revenue (yuan)": 341596317171.5,
+    "Beijing total employee count": 87614,
+    "Beijing per capita revenue (yuan/person)": 3898878.23,
+    "National total operating revenue (yuan)": 4614449954119.47,
+    "National total employee count": 3254510,
+    "National average per capita revenue (yuan/person)": 1417863.2,
+    "Beijing excess over national average (%)": 175.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium004.json b/assets/qa_gold/comprehensive_decision/medium004.json
new file mode 100644
index 0000000000000000000000000000000000000000..de8996b7a0d43025452ef338ce0f42942893db12
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium004.json
@@ -0,0 +1,32 @@
+{
+  "id": "medium004",
+  "question": "In 2022, rank provinces by profitability in chemical raw materials and chemical products manufacturing. Provincial operating profit margin is computed as total operating profit divided by total operating revenue. Using this as the ranking criterion, what is Guangdong Province's rank? Apply the same ranking to all relevant enterprises within Guangdong Province—which enterprise ranks first?",
+  "guidelines": "Two answers required: first is Guangdong Province's rank in the provincial ranking (integer, e.g. \"6\" means 6th place); second is the full name of the top-ranked enterprise in Guangdong, which must match the \"Company Name\" field in company_profile.csv. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": [
+    6,
+    "Hengyi Changhua Technology Co., Ltd."
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter province records with industry=\"Chemical Raw Materials and Chemical Products Manufacturing\", extract province, total operating profit, total operating revenue; 34 records obtained.",
+    "Filter records with non-null total operating profit and total operating revenue, and total operating revenue ≠ 0; 16 valid provinces. Compute provincial operating profit margin = total operating profit / total operating revenue × 100%, sort by this indicator descending.",
+    "In the provincial ranking, Guangdong Province has total operating profit 11,690,448,651.68 yuan, total operating revenue 101,800,752,670.91 yuan, operating profit margin = 11.4837%, rank 6th.",
+    "From company_profile.csv, filter enterprises with province=\"Guangdong Province\" and industry=\"Chemical Raw Materials and Chemical Products Manufacturing\", obtain company names and bmCode; from company_operation_status.csv, filter year=2022 and bmCode in the above set, extract operating profit and operating revenue.",
+    "For Guangdong enterprises, compute operating profit margin = operating profit / operating revenue × 100% (requiring operating revenue ≠ 0), sort by operating profit margin descending. First place: \"Hengyi Changhua Technology Co., Ltd.\" (bmCode=533611), operating profit margin = 2,190,338,633.59 / 3,466,111,075.75 × 100% = 63.19%."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Guangdong total operating profit (yuan)": 11690448651.68,
+    "Guangdong total operating revenue (yuan)": 101800752670.91,
+    "Guangdong operating profit margin (%)": 11.4837,
+    "Guangdong provincial rank": 6,
+    "Top-ranked enterprise in Guangdong": "Hengyi Changhua Technology Co., Ltd.",
+    "Top enterprise operating profit (yuan)": 2190338633.59,
+    "Top enterprise operating revenue (yuan)": 3466111075.75,
+    "Top enterprise operating profit margin (%)": 63.19
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium005.json b/assets/qa_gold/comprehensive_decision/medium005.json
new file mode 100644
index 0000000000000000000000000000000000000000..003b8ee6423cb5472084739e4cb758204243dd08
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium005.json
@@ -0,0 +1,23 @@
+{
+  "id": "medium005",
+  "question": "In 2022, from all private enterprises in the food and beverage industry, aggregate government rewards and subsidy amounts by province of registration to find the province with the highest provincial subsidy total. How many hundred million yuan in government subsidies did private enterprises in that province receive in total?",
+  "guidelines": "The answer should be a numerical value (2 decimal places), unit is hundred million yuan. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 12.6,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all enterprise records with industry=\"Food and Beverage\" and ownership=\"Private Enterprise\" from company_profile.csv, extract enterprise name, bmCode and province fields, finding 161 food and beverage private enterprises.",
+    "Join with company_operation_status.csv via bmCode to obtain 2022 data for these enterprises, extract government reward funds and subsidy fields.",
+    "Filter records where government reward funds and subsidy are non-null, resulting in 157 enterprises with subsidy data.",
+    "Group by province and calculate total government subsidies received by private enterprises in each province.",
+    "Sort provinces by total government subsidies in descending order. The province with the highest total subsidies is Inner Mongolia, with total subsidies of 1,259,874,619.23 yuan; convert from yuan to hundred million yuan (divide by 100,000,000), yielding 12.60 hundred million yuan."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Inner Mongolia food and beverage private enterprise total subsidies (yuan)": 1259874619.23,
+    "Inner Mongolia food and beverage private enterprise total subsidies (hundred million yuan)": 12.6
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium006.json b/assets/qa_gold/comprehensive_decision/medium006.json
new file mode 100644
index 0000000000000000000000000000000000000000..d07f84c821682269d86b43bb69a689e3b82b00fd
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium006.json
@@ -0,0 +1,22 @@
+{
+  "id": "medium006",
+  "question": "In 2022, a local government planned to introduce support policies for the specialized equipment manufacturing industry and needed to understand the province's R&D personnel investment level in this industry. Is the average proportion of R&D personnel to total employees in specialized equipment manufacturing enterprises in Zhejiang Province higher than the national average for this industry?",
+  "guidelines": "The answer should be \"Yes\" or \"No\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter records with industry=\"Specialized Equipment Manufacturing\" and province=\"Zhejiang\" from regional_industry_status.csv, extract average R&D personnel proportion field; Zhejiang's average R&D personnel proportion is 18.25%.",
+    "Filter records with industry=\"Specialized Equipment Manufacturing\" from national_industry_status.csv, extract average R&D personnel proportion field; national average R&D personnel proportion is 20.10%.",
+    "Compare Zhejiang's average R&D personnel proportion (18.25%) with the national average (20.10%).",
+    "Zhejiang's value is not greater than the national value, therefore the answer is \"No\"."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Zhejiang average R&D personnel proportion (%)": 18.25,
+    "National average R&D personnel proportion (%)": 20.1
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium007.json b/assets/qa_gold/comprehensive_decision/medium007.json
new file mode 100644
index 0000000000000000000000000000000000000000..760ec6cb0bb7491913eb833e281602513047dd30
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium007.json
@@ -0,0 +1,32 @@
+{
+  "id": "medium007",
+  "question": "In 2022, analyze government subsidy leverage in the information transmission, software and information technology services industry. Define government subsidy leverage effect as the ratio of each province's total operating profit to total government subsidies. What is the specific ratio for the province with the highest government subsidy leverage effect? Which enterprise in that province has the highest leverage effect?",
+  "guidelines": "Two answers required: first is a numeric value (2 decimal places, unitless ratio); second is the full company name. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": [
+    67.19,
+    "Dongche Kexin Systems Company"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From company_profile.csv, filter industry=\"Information Transmission, Software and Information Technology Services\", extract company name, bmCode, and province.",
+    "Inner-join with company_operation_status.csv on year=2022 by company name and bmCode to obtain the 2022 enterprise sample for this industry (644 records, covering 31 provincial-level regions).",
+    "Group by province: sum operating profit to get each province's total operating profit; sum government rewards and subsidies to get each province's total government subsidies.",
+    "Keep provinces with total government subsidies > 0 and non-null total operating profit; compute government subsidy leverage effect = total operating profit / total government subsidies — 31 valid provinces; sort descending: first is Jiangxi Province — total operating profit 93,463,073.13 yuan, total government rewards and subsidies 1,391,005.09 yuan, provincial leverage = 67.19 (2 decimal places).",
+    "Sort enterprises by government subsidy leverage effect descending; top enterprise: Dongche Kexin Systems Company (bmCode=591984); operating profit 20,080,241.24 yuan, government rewards and subsidies 1,391,005.09 yuan, enterprise-level leverage ≈ 14.44; Zhongke Ruanchuang Software Company is excluded from enterprise-level ranking due to missing subsidy data."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Provincial aggregation note": "Aggregate by province from the enterprise table; do not use Jiangxi totals missing from the regional table",
+    "Jiangxi total operating profit (yuan)": 93463073.13,
+    "Jiangxi total government rewards and subsidies (yuan)": 1391005.09,
+    "Jiangxi government subsidy leverage effect": 67.19,
+    "Enterprise with highest government subsidy leverage effect": "Dongche Kexin Systems Company",
+    "Enterprise operating profit (yuan)": 20080241.24,
+    "Enterprise government rewards and subsidies (yuan)": 1391005.09,
+    "Enterprise government subsidy leverage effect": 14.44
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium008.json b/assets/qa_gold/comprehensive_decision/medium008.json
new file mode 100644
index 0000000000000000000000000000000000000000..35a98394ae6d8cf70c5ecf4e8741feeaa8391330
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium008.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium008",
+  "question": "In 2022, to study the capital turnover of central state-owned enterprises in the electricity, heat, gas and water production and supply industry, calculate the asset turnover ratio for each enterprise by dividing its annual operating revenue by its total assets. Find the arithmetic mean of the asset turnover ratios for these enterprises.",
+  "guidelines": "The answer should be a numerical value (rounded to 4 decimal places). If the relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 0.3266,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter records from company_profile.csv where industry=\"Electricity, heat, gas and water production and supply\" and ownership=\"Central state-owned enterprise\", extract company name and bmCode fields, finding 44 enterprises in total.",
+    "Join with company_operation_status.csv by bmCode to retrieve 2022 data for these enterprises, extract operating revenue amount and total assets fields.",
+    "Filter records where both operating revenue amount and total assets are not empty and total assets is greater than 0, resulting in 44 valid enterprises.",
+    "Calculate asset turnover ratio for each enterprise = operating revenue amount / total assets.",
+    "Calculate the average asset turnover ratio for all qualifying central state-owned enterprises = sum of asset turnover ratios (14.369275) / number of enterprises (44) = 0.3266."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Sum of Asset Turnover Ratios": 14.369275,
+    "Number of Valid Enterprises": 44,
+    "Average Asset Turnover Ratio": 0.3266
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium009.json b/assets/qa_gold/comprehensive_decision/medium009.json
new file mode 100644
index 0000000000000000000000000000000000000000..4b46dadf567b2e179fd7a91b0d150f8d3863523c
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium009.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium009",
+  "question": "In 2022, a mining industry enterprise plans to go public for financing and hopes to list on the exchange with the highest average market capitalization among enterprises in this industry. Among the four exchanges—Shenzhen Stock Exchange, Hong Kong Stock Exchange, Shanghai Stock Exchange, and Beijing Stock Exchange—which exchange has the highest average market capitalization of listed mining enterprises?",
+  "guidelines": "The answer should be the exchange name (e.g., \"Shenzhen Stock Exchange\", \"Hong Kong Stock Exchange\", \"Shanghai Stock Exchange\", \"Beijing Stock Exchange\"). If the relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "Shanghai Stock Exchange",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter records from company_profile.csv where industry=\"Mining\", extract company name, bmCode and exchange fields, finding 143 mining enterprises in total.",
+    "Join with company_operation_status.csv by bmCode to retrieve 2022 data for these enterprises, extract company market capitalization field.",
+    "Filter records where company market capitalization is not empty, resulting in 142 enterprises with market capitalization data.",
+    "Group by exchange field and calculate the average market capitalization of mining enterprises for each exchange.",
+    "Sort by average market capitalization in descending order. Shanghai Stock Exchange has the highest average market capitalization, with 49 mining enterprises, total market capitalization of 4,771.8 billion CNY, and average market capitalization of 973.84 billion CNY."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Shanghai Stock Exchange Total Market Capitalization of Mining Enterprises (billion CNY)": 47718.0,
+    "Number of Mining Enterprises on Shanghai Stock Exchange": 49,
+    "Shanghai Stock Exchange Average Market Capitalization of Mining Enterprises (billion CNY)": 973.84
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium010.json b/assets/qa_gold/comprehensive_decision/medium010.json
new file mode 100644
index 0000000000000000000000000000000000000000..13427c381dfa6bc48929bcef8b0b93b297f4ab21
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium010.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium010",
+  "question": "In 2022, in the provincial data for the construction industry, each province has an indicator reflecting the average asset-liability ratio (financial leverage level) of enterprises in that province's industry (considering only enterprises with valid total assets and total liabilities). Among the provinces covered by valid data, which province has the lowest value for this mean indicator, and what is that value?",
+  "guidelines": "The answer should be a numerical value (rounded to 2 decimal places), in units of %. If the relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": [
+    "Shanxi Province",
+    27.17
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter records from company_profile.csv where industry=\"Construction\", extract company name, bmCode and province fields, finding 148 enterprises in total.",
+    "Join with company_operation_status.csv by bmCode, extract total liabilities and total assets fields, resulting in 148 records after merging.",
+    "Filter enterprise records where both total liabilities and total assets are not empty and total assets is greater than 0, resulting in 148 valid enterprises covering 22 provinces.",
+    "Calculate asset-liability ratio for each enterprise = total liabilities / total assets × 100%, then group by province and calculate the mean asset-liability ratio for each province.",
+    "Sort all provinces by mean asset-liability ratio in ascending order. Shanxi Province has the lowest mean asset-liability ratio of 27.17%."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Total Number of Construction Enterprises": 148,
+    "Number of Valid Enterprises (total liabilities and total assets not empty)": 148,
+    "Number of Valid Provinces": 22,
+    "Shanxi Province Mean Asset-Liability Ratio (%)": 27.17
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium011.json b/assets/qa_gold/comprehensive_decision/medium011.json
new file mode 100644
index 0000000000000000000000000000000000000000..0412fb929f06ab3ce4dbec49869d16018779bd2f
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium011.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium011",
+  "question": "In 2022, among all enterprises in the rubber and plastic products industry with R&D investment records, what is the R&D concentration CR5?",
+  "guidelines": "The answer should be a numerical value (rounded to 2 decimal places), in units of %. If the relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 34.66,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter records from company_profile.csv where industry=\"Rubber and plastic products\", extract company name and bmCode fields, finding 107 enterprises in total.",
+    "Join with company_operation_status.csv by bmCode to retrieve 2022 data for these enterprises, extract R&D investment amount field.",
+    "Filter records where R&D investment amount is not empty, resulting in 106 enterprises with R&D investment data. Sort by R&D investment amount in descending order, extract the top 5 enterprises and their R&D investment amounts.",
+    "Calculate the sum of R&D investment amounts of the top 5 enterprises as 4,221,126,553.77 CNY. Calculate the total R&D investment amount of all valid enterprises in the industry as 12,179,847,530.98 CNY.",
+    "Calculate R&D concentration CR5 = (4,221,126,553.77 / 12,179,847,530.98) × 100% = 34.66%."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Sum of R&D Investment Amount of Top 5 Enterprises (CNY)": 4221126553.77,
+    "Total R&D Investment Amount of the Industry (CNY)": 12179847530.98,
+    "R&D Concentration CR5 (%)": 34.66
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium012.json b/assets/qa_gold/comprehensive_decision/medium012.json
new file mode 100644
index 0000000000000000000000000000000000000000..e11ac36777106f75cde766104ecf1db821dee895
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium012.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium012",
+  "question": "In 2022, among all enterprises in Guangdong Province belonging to the wholesale and retail trade industry, using each enterprise's net profit margin as the comparison standard, what is the indicator value for the enterprise with the highest net profit margin? What is that enterprise's rank among all enterprises in this industry nationwide?",
+  "guidelines": "Two answers required: first is a numeric value (2 decimal places, unit %); second is the rank number (integer, e.g. \"7\" means 7th place). If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": [
+    31.25,
+    7
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From company_profile.csv, filter all enterprise records with industry=\"Wholesale and Retail Trade\" and province=\"Guangdong Province\", extract company name and bmCode; 43 Guangdong wholesale and retail enterprises found.",
+    "From company_operation_status.csv, join 2022 data by bmCode for these enterprises, extract net profit and operating revenue; 2022 data found for 43 enterprises.",
+    "Filter enterprise records with non-null net profit and operating revenue, and operating revenue ≠ 0; 43 valid records.",
+    "Compute each enterprise's net profit margin = net profit / operating revenue × 100%. Sort all enterprises by net profit margin descending; the enterprise with the highest net profit margin is \"Yonghui Changda Wholesale Company\", net profit 292,221,119.71 yuan, operating revenue 935,248,730.59 yuan, net profit margin = 31.25%.",
+    "From company_profile.csv, filter all enterprises nationwide with industry=\"Wholesale and Retail Trade\"; from company_operation_status.csv, take year=2022 with operating revenue > 0 and non-null net profit. Sort nationwide valid enterprises by net profit margin descending; Yonghui Changda Wholesale Company ranks 7th."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Yonghui Changda Wholesale Company net profit (yuan)": 292221119.71,
+    "Yonghui Changda Wholesale Company operating revenue (yuan)": 935248730.59,
+    "Yonghui Changda Wholesale Company net profit margin (%)": 31.25,
+    "Yonghui Changda Wholesale Company nationwide rank": 7
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium013.json b/assets/qa_gold/comprehensive_decision/medium013.json
new file mode 100644
index 0000000000000000000000000000000000000000..8adc1fd1b1977bbd746bf566f468381eb775852c
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium013.json
@@ -0,0 +1,29 @@
+{
+  "id": "medium013",
+  "question": "In 2022, a scientific research and technical services enterprise wishes to identify the province with the fastest net profit growth in the industry to guide market expansion. What is the indicator value for the province with the highest median year-on-year net profit growth rate in the national scientific research and technical services industry? What is the nationwide rank of the enterprise with the highest such indicator in that province among all enterprises in this industry?",
+  "guidelines": "Two answers required: first is a numeric value (2 decimal places, unit %), i.e. the indicator value for the province with the highest \"median year-on-year net profit growth rate\" in this industry nationwide; second is a rank number (integer, e.g. \"23\" means 23rd place), i.e. the nationwide rank of the enterprise with the highest \"year-on-year net profit growth rate\" in that province among all enterprises in this industry. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": [
+    13.81,
+    23
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter all records with industry=\"Scientific Research and Technical Services\", extract province and median year-on-year net profit growth rate; 34 province records found.",
+    "Filter province records with non-null median year-on-year net profit growth rate; 16 valid provinces.",
+    "Sort by median year-on-year net profit growth rate descending; first place: Anhui Province, median year-on-year net profit growth rate 13.81%. From company_profile.csv, filter enterprises with province=\"Anhui Province\" and industry=\"Scientific Research and Technical Services\", obtain enterprise bmCode set.",
+    "From company_operation_status.csv, filter year=2022 and bmCode in that set, extract \"year-on-year net profit growth rate\", sort by this indicator descending. Highest enterprise in that province: Zhongqi Shengyuan Technology Research Institute, year-on-year net profit growth rate 32.58%.",
+    "From company_profile.csv, filter nationwide enterprises with industry=\"Scientific Research and Technical Services\"; from company_operation_status.csv, filter year=2022 with valid year-on-year net profit growth rate. Sort by year-on-year net profit growth rate descending; Zhongqi Shengyuan Technology Research Institute ranks 23rd nationwide."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Valid province count for scientific research and technical services (non-null median YoY net profit growth)": 16,
+    "Anhui Province median year-on-year net profit growth rate (%)": 13.81,
+    "Enterprise with highest indicator in Anhui Province": "Zhongqi Shengyuan Technology Research Institute",
+    "Enterprise year-on-year net profit growth rate (%)": 32.58,
+    "Enterprise nationwide rank": 23
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium014.json b/assets/qa_gold/comprehensive_decision/medium014.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f2091aefed6befaf99ff2e2739720801983a639
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium014.json
@@ -0,0 +1,30 @@
+{
+  "id": "medium014",
+  "question": "In 2022, for the metal smelting and rolling processing industry, among provinces with valid records for both total government subsidies and total industry employee count, per capita subsidy is computed as each province's total government rewards and subsidies divided by that province's industry employee count. What is the per capita subsidy in yuan for the province with the highest per capita subsidy? What is the nationwide rank of the enterprise with the highest such indicator in that province among all enterprises in this industry?",
+  "guidelines": "Two answers required: first is a numeric value (2 decimal places, unit yuan/person), i.e. the highest provincial per capita subsidy; second is a rank number (integer), indicating the nationwide rank of the enterprise with the highest indicator in that province among all enterprises in this industry. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": [
+    17569.95,
+    12
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter all province records with industry=\"Metal Smelting and Rolling Processing\", extract province, total government rewards and subsidies, total employee count; 34 province records obtained.",
+    "Filter records with non-null total government rewards and subsidies and total employee count, total employee count > 0 and total government rewards and subsidies > 0; 16 valid provinces.",
+    "Compute each province's per capita subsidy (yuan/person) = total government rewards and subsidies / total employee count, sort by per capita subsidy descending. Province with highest per capita subsidy: Shanghai; 893,080,778.37 / 50,830 = 17,569.95 yuan/person.",
+    "Within Shanghai: from company_profile.csv, filter enterprises with industry=\"Metal Smelting and Rolling Processing\" and province=\"Shanghai\", obtain bmCode set; from company_operation_status.csv, filter year=2022 and bmCode in that set, with government rewards and subsidies > 0 and total employee count > 0. Compute each enterprise's per capita subsidy = government rewards and subsidies / total employee count, sort descending; first place: Xin Ge Jinze Materials Company.",
+    "Nationwide, for all enterprises in this industry (year=2022, government rewards and subsidies > 0, total employee count > 0), compute per capita subsidy using the same formula and sort descending; Xin Ge Jinze Materials Company ranks 12th nationwide."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Shanghai total government rewards and subsidies (yuan)": 893080778.37,
+    "Shanghai total employee count": 50830.0,
+    "Shanghai per capita subsidy (yuan/person)": 17569.95,
+    "Enterprise with highest indicator in Shanghai": "Xin Ge Jinze Materials Company",
+    "Enterprise per capita subsidy (yuan/person)": 31941.25,
+    "Enterprise nationwide rank": 12
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium015.json b/assets/qa_gold/comprehensive_decision/medium015.json
new file mode 100644
index 0000000000000000000000000000000000000000..e3c64d519ab04776ed23bf7f4f48fe728dcc832e
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium015.json
@@ -0,0 +1,65 @@
+{
+  "id": "medium015",
+  "question": "List the 2022 indicators for which Shandong Province's financial industry enterprise averages are below the national financial industry medians.",
+  "guidelines": "The answer must list all qualifying indicator names, separated by semicolons. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": [
+    "Year-on-year R&D personnel growth rate",
+    "Year-on-year operating profit growth rate",
+    "Year-on-year net profit growth rate",
+    "Year-on-year employee growth rate",
+    "Capitalized R&D expenditure",
+    "Year-on-year capitalized R&D expenditure growth rate",
+    "Annual PCT patent applications",
+    "Annual PCT invention patent applications",
+    "Provincial/ministerial science and technology progress award",
+    "Participation in drafting national standards",
+    "Participation in drafting industry standards",
+    "Annual Chinese patent applications",
+    "Annual Chinese invention patent applications",
+    "Annual Chinese patent grants",
+    "Cumulative Chinese invention patent applications",
+    "Annual Chinese invention patent grants",
+    "Cumulative PCT patent applications",
+    "Cumulative PCT invention patent applications",
+    "Cumulative Chinese patent applications",
+    "Cumulative Chinese invention patent grants",
+    "Cumulative patent citations",
+    "R&D personnel ratio",
+    "R&D personnel count",
+    "Year-on-year R&D expenditure growth rate",
+    "Cumulative Chinese invention patent lapses",
+    "Company market value",
+    "Asset-liability ratio",
+    "Total employee count"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter records with industry=\"Financial Industry\" and province=\"Shandong Province\", extract total enterprise count; 12 Shandong financial industry enterprises found.",
+    "From national_industry_status.csv, filter records with industry=\"Financial Industry\" and district=\"National\", extract total enterprise count; 297 national financial industry enterprises found.",
+    "From regional_industry_status.csv, extract all indicator averages for Shandong financial industry enterprises, including average total assets, average net profit, average total employee count, average operating revenue, average asset-liability ratio, average R&D expenditure, etc.",
+    "From national_industry_status.csv, extract all indicator medians for national financial industry enterprises, including median total assets, median net profit, median total employee count, median operating revenue, median asset-liability ratio, median R&D expenditure, etc.",
+    "Compare each Shandong financial industry indicator average with the corresponding national financial industry median; exclude indicators such as enterprise count and totals that are not suitable for comparison. Filter indicators where Shandong average is below national median; 28 indicators found, mainly including year-on-year R&D personnel growth rate, year-on-year operating profit growth rate, year-on-year net profit growth rate, annual Chinese patent applications, cumulative Chinese patent applications, R&D personnel count, R&D personnel ratio, etc. Final result: Shandong Province financial industry enterprises have 28 indicators with averages below the national financial industry medians."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Shandong Province financial industry enterprise count": 12,
+    "National financial industry enterprise count": 297,
+    "Indicator count below national median": 28,
+    "Main indicator list": [
+      "Year-on-year R&D personnel growth rate",
+      "Year-on-year operating profit growth rate",
+      "Year-on-year net profit growth rate",
+      "Year-on-year employee growth rate",
+      "Capitalized R&D expenditure",
+      "Year-on-year capitalized R&D expenditure growth rate",
+      "Annual PCT patent applications",
+      "Annual PCT invention patent applications",
+      "Provincial/ministerial science and technology progress award",
+      "Participation in drafting national standards"
+    ]
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium016.json b/assets/qa_gold/comprehensive_decision/medium016.json
new file mode 100644
index 0000000000000000000000000000000000000000..9f6d99bbdd1148c510caa4ba0794e6b176ee47b9
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium016.json
@@ -0,0 +1,22 @@
+{
+  "id": "medium016",
+  "question": "In 2022, between Sichuan Province's top enterprise by operating revenue and Shandong Province's top enterprise by net profit in the pharmaceutical manufacturing industry, which one has the highest tax payment?",
+  "guidelines": "The answer must be the company name. Output only the company name, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": "No relevant data found",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From company_profile.csv, filter all enterprise records with province=\"Sichuan Province\", extract company name and bmCode; 198 Sichuan enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by company name and bmCode, extract operating revenue; sort by operating revenue descending. Top enterprise by operating revenue: Dexi Jinjin Intelligent Electrical Company (bmCode: 451895), operating revenue 142,422,544,758.90 yuan.",
+    "From company_profile.csv, filter all enterprise records with province=\"Shandong Province\" and industry=\"Pharmaceutical Manufacturing\", extract company name and bmCode; 22 Shandong pharmaceutical manufacturing enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by company name and bmCode, extract net profit; sort by net profit descending. Top enterprise by net profit: Jianming Anyuan Medical Technology Company (bmCode: 199134), net profit 2,950,153,469.00 yuan.",
+    "Checked all fields in company_operation_status.csv, company_profile.csv, company_core.csv, etc.; no tax payment related fields found. Since the data files do not contain tax payment information, the tax payments of the two enterprises cannot be compared."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Tax payment field exists": false
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium017.json b/assets/qa_gold/comprehensive_decision/medium017.json
new file mode 100644
index 0000000000000000000000000000000000000000..d75643a7ec7c74b218cad8b7ef538ef38b2c8b87
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium017.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium017",
+  "question": "In 2022, the industry where Zhongbai Jinmao Chain Company operates, is the enterprise with the highest year-on-year R&D expenditure growth rate also the one with the highest R&D expenditure?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From company_profile.csv, find the record for Zhongbai Jinmao Chain Company, extract industry field; industry is \"Wholesale and Retail Trade\". From company_profile.csv, filter all enterprise records with industry=\"Wholesale and Retail Trade\", extract company name and bmCode; 273 wholesale and retail enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by company name and bmCode, extract R&D expenditure and year-on-year R&D expenditure growth rate; 2022 data found for 273 enterprises.",
+    "Filter enterprise records with non-null year-on-year R&D expenditure growth rate (129 records); sort all enterprises by year-on-year R&D expenditure growth rate descending. Enterprise with highest year-on-year R&D expenditure growth rate: Yonghui Zesheng Chain Company, growth rate 2221.3%, R&D expenditure 516,766.56 yuan.",
+    "Filter enterprise records with non-null R&D expenditure; sort all enterprises by R&D expenditure descending. Enterprise with highest R&D expenditure: Bubusheng Jin Commerce Company, R&D expenditure 2,800,235,364.00 yuan, year-on-year R&D expenditure growth rate 11.8%.",
+    "Compare the enterprise with highest year-on-year R&D expenditure growth rate (Yonghui Zesheng Chain Company) and the enterprise with highest R&D expenditure (Bubusheng Jin Commerce Company); they are not the same enterprise, so the answer is \"No\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Enterprise with highest year-on-year R&D expenditure growth rate": "Yonghui Zesheng Chain Company",
+    "Highest growth rate enterprise YoY R&D expenditure growth rate (%)": 2221.3,
+    "Highest growth rate enterprise R&D expenditure (yuan)": 516766.56,
+    "Enterprise with highest R&D expenditure": "Bubusheng Jin Commerce Company",
+    "Highest R&D enterprise R&D expenditure (yuan)": 2800235364.0,
+    "Highest R&D enterprise YoY R&D expenditure growth rate (%)": 11.8,
+    "Is same enterprise": false
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium018.json b/assets/qa_gold/comprehensive_decision/medium018.json
new file mode 100644
index 0000000000000000000000000000000000000000..48676a13a2e8d266f473a44a9ad3683ed082c15c
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium018.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium018",
+  "question": "In 2022 Nationwide, is the province with the highest R&D expenditure growth rate also the province with the lowest R&D expenditure?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From regional_industry_status.csv, read R&D expenditure data for all provinces, including province, total R&D expenditure, average year-on-year R&D expenditure growth rate, etc.",
+    "Aggregate total R&D expenditure and average year-on-year R&D expenditure growth rate by province; exclude provinces with total R&D expenditure of 0 (missing data).",
+    "Identify the province with the highest average year-on-year R&D expenditure growth rate: Hong Kong Special Administrative Region, R&D expenditure growth rate 212.35%, total R&D expenditure 14,248,827,635.96 yuan.",
+    "Identify the province with the lowest total R&D expenditure (excluding 0): Jilin Province, total R&D expenditure 8,574,294,268.49 yuan, R&D expenditure growth rate 16.95%.",
+    "Compare whether the province with the highest R&D expenditure growth rate (Hong Kong SAR) and the province with the lowest R&D expenditure (Jilin Province) are the same province.Conclusion: The province with the highest R&D expenditure growth rate is not the province with the lowest R&D expenditure; the answer is No."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province with highest R&D expenditure growth rate": "Hong Kong Special Administrative Region",
+    "Highest R&D expenditure growth rate (%)": 212.35,
+    "Highest growth province total R&D expenditure (yuan)": 14248827635.96,
+    "Province with lowest R&D expenditure": "Jilin Province",
+    "Lowest R&D expenditure (yuan)": 8574294268.49,
+    "Lowest R&D province R&D expenditure growth rate (%)": 16.95,
+    "Is same province": false
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium019.json b/assets/qa_gold/comprehensive_decision/medium019.json
new file mode 100644
index 0000000000000000000000000000000000000000..c6a725cad0997466cdb8d447e375bc5b674c3282
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium019.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium019",
+  "question": "In 2022, Sichuan Province, is Zhongbai Jinmao Chain Company's R&D expenditure higher than the R&D expenditure of the enterprise ranked 15th nationwide in its industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From company_profile.csv, find the record for company name=\"Zhongbai Jinmao Chain Company\", extract industry and province; industry is \"Wholesale and Retail Trade\", province is \"Sichuan Province\".",
+    "From company_operation_status.csv, filter 2022 data for company name=\"Zhongbai Jinmao Chain Company\", extract R&D expenditure; Zhongbai Jinmao Chain Company's R&D expenditure is 11,270,987.0 yuan.",
+    "From company_profile.csv, filter all enterprise records with industry=\"Wholesale and Retail Trade\", extract company name or bmCode; 273 wholesale and retail enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by company name or bmCode, extract R&D expenditure; 2022 R&D expenditure data found for 273 wholesale and retail enterprises.",
+    "Filter enterprise records with non-null R&D expenditure; sort all enterprises by R&D expenditure descending to determine R&D expenditure ranking. Enterprise ranked 15th: Lianhua Tongze Commerce Company, R&D expenditure 265,616,054.7 yuan, province Beijing. Compare Zhongbai Jinmao Chain Company's R&D expenditure (11,270,987.0 yuan) with the 15th-ranked nationwide wholesale and retail enterprise's R&D expenditure (265,616,054.7 yuan). Since 11,270,987.0 < 265,616,054.7, the answer is No."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Zhongbai Jinmao Chain Company R&D expenditure (yuan)": 11270987.0,
+    "15th-ranked nationwide wholesale and retail enterprise R&D expenditure (yuan)": 265616054.7,
+    "Comparison result": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium020.json b/assets/qa_gold/comprehensive_decision/medium020.json
new file mode 100644
index 0000000000000000000000000000000000000000..b52aae9d3cfa821e9189ca3a05f199b625607502
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium020.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium020",
+  "question": "In 2022, the chemical raw materials and chemical products manufacturing industry in Shandong Province, is the market capitalization of the leading enterprises by operating revenue among the top 3 in this industry nationwide?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From company_profile.csv, filter all enterprise records with province=\"Shandong Province\" and industry=\"Chemical raw materials and chemical products manufacturing\", extract company names, identify enterprises in the chemical raw materials and chemical products manufacturing industry in Shandong Province.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by bmCode or company name, extract operating revenue amount and company market capitalization, obtain operating revenue and market capitalization for each enterprise.",
+    "Filter enterprises with non-null operating revenue amount and company market capitalization, sort by operating revenue amount descending, determine the leading enterprise by operating revenue as \"Hengyi Changhua Fine Chemical Company\" with operating revenue of 165,565,462,711.66 yuan and company market capitalization of 2,754.0.",
+    "From company_profile.csv, filter all enterprise records with industry=\"Chemical raw materials and chemical products manufacturing\", from company_operation_status.csv obtain 2022 company market capitalization for these enterprises by bmCode or company name.",
+    "Filter enterprises with non-null company market capitalization, sort by company market capitalization descending, determine nationwide market capitalization ranking for this industry; identify top 3: No.1 \"Hengyi Changhua Fine Chemical Company\" (Shandong Province) 2,754.0, No.2 \"Rongsheng Jinsheng Chemical Company\" (Qinghai Province) 1,040.0, No.3 \"Hengyi Yuanjin Fine Chemical Company\" (Ningxia Hui Autonomous Region) 927.0. Determine whether the Shandong Province revenue-leading enterprise \"Hengyi Changhua Fine Chemical Company\" ranks among the nationwide top 3 by market capitalization: this enterprise is No.1 nationwide by market capitalization, thus it is among the top 3, conclusion: Yes."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Shandong Province operating revenue leading enterprise name": "Hengyi Changhua Fine Chemical Company",
+    "Shandong Province operating revenue leading enterprise operating revenue (yuan)": 165565462711.66,
+    "Shandong Province operating revenue leading enterprise company market capitalization": 2754.0,
+    "Chemical raw materials and chemical products manufacturing nationwide market cap No.1 enterprise": "Hengyi Changhua Fine Chemical Company",
+    "Chemical raw materials and chemical products manufacturing nationwide market cap No.3 enterprise": "Hengyi Yuanjin Fine Chemical Company",
+    "Chemical raw materials and chemical products manufacturing nationwide market cap No.3 enterprise market capitalization": 927.0,
+    "Comparison result": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium021.json b/assets/qa_gold/comprehensive_decision/medium021.json
new file mode 100644
index 0000000000000000000000000000000000000000..202b74062e7a48dbd25f6fe834b4f956b8338881
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium021.json
@@ -0,0 +1,25 @@
+{
+  "id": "medium021",
+  "question": "In 2022, the chemical raw materials and chemical products manufacturing industry in Shandong Province, is the market capitalization of the enterprise with the highest operating revenue among the top 3 in this industry nationwide?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From company_profile.csv, filter all enterprise records with province=\"Shandong Province\" and industry=\"Chemical raw materials and chemical products manufacturing\", extract company names, identify enterprises in the chemical raw materials and chemical products manufacturing industry in Shandong Province.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by bmCode or company name, extract operating revenue amount and company market capitalization, obtain operating revenue and market capitalization for each enterprise.",
+    "Filter enterprises with non-null operating revenue amount and company market capitalization, sort by operating revenue amount descending, determine the leading enterprise by operating revenue as \"Hengyi Changhua Fine Chemical Company\" with operating revenue of 165,565,462,711.66 yuan and company market capitalization of 2,754.0.",
+    "From company_profile.csv, filter all enterprise records with industry=\"Chemical raw materials and chemical products manufacturing\", from company_operation_status.csv obtain 2022 company market capitalization for these enterprises by bmCode or company name.",
+    "Filter enterprises with non-null company market capitalization, sort by company market capitalization descending, determine nationwide market capitalization ranking for this industry; identify top 3: No.1 \"Hengyi Changhua Fine Chemical Company\" (Shandong Province) 2,754.0, No.2 \"Rongsheng Jinsheng Chemical Company\" (Qinghai Province) 1,040.0, No.3 \"Hengyi Yuanjin Fine Chemical Company\" (Ningxia Hui Autonomous Region) 927.0. Determine whether the Shandong Province revenue-leading enterprise \"Hengyi Changhua Fine Chemical Company\" ranks among the nationwide top 3 by market capitalization: this enterprise is No.1 nationwide by market capitalization, thus it is among the top 3, conclusion: Yes."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Shandong Province operating revenue leading enterprise operating revenue (yuan)": 165565462711.66,
+    "Shandong Province operating revenue leading enterprise company market capitalization": 2754.0,
+    "Chemical raw materials and chemical products manufacturing nationwide market cap No.3 enterprise market capitalization": 927.0,
+    "Comparison result": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium022.json b/assets/qa_gold/comprehensive_decision/medium022.json
new file mode 100644
index 0000000000000000000000000000000000000000..6f0ce01ab4d9d766860b51a1460dfb93618f328e
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium022.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium022",
+  "question": "In 2022 is the region with the highest average R&D investment growth rate also the one with the most policy quantity?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter all provincial records, extract province and year-on-year R&D investment growth rate mean fields, obtain R&D investment growth rate data for each province and industry.",
+    "Aggregate R&D investment growth rate data by province, calculate the average R&D investment growth rate for each province (average of year-on-year R&D investment growth rate mean across all industries).",
+    "Identify the region with the highest average R&D investment growth rate as \"Hong Kong Special Administrative Region\" with an average R&D investment growth rate of 212.35%.",
+    "From policy_release_status.csv, filter all provincial records, count policy quantity by province, obtain policy quantity for each province.",
+    "Identify the region with the most policies as \"Guangdong Province\" with 59 policies. Compare the region with the highest average R&D investment growth rate (Hong Kong Special Administrative Region) with the region with the most policies (Guangdong Province), determine they are not the same, conclusion: No."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Region with highest average R&D investment growth rate": "Hong Kong Special Administrative Region",
+    "Region with most policies": "Guangdong Province",
+    "Comparison result": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium023.json b/assets/qa_gold/comprehensive_decision/medium023.json
new file mode 100644
index 0000000000000000000000000000000000000000..c2c6d241ccd0c6a83f97f7b7b890792704f99eda
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium023.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium023",
+  "question": "In 2022 is the average year-on-year net profit growth rate for the industry where Haishan Changgong Equipment Company operates higher than the average R&D investment growth rate?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From company_profile.csv, find the record with company name=\"Haishan Changgong Equipment Company\", extract the industry field, determine its industry as \"General equipment manufacturing\".",
+    "From national_industry_status.csv, filter records with industry=\"General equipment manufacturing\" and district=\"National\", extract net profit year-on-year growth rate mean and R&D investment year-on-year growth rate mean fields.",
+    "Obtain the net profit year-on-year growth rate mean for general equipment manufacturing as -79.47492958%, and the R&D investment year-on-year growth rate mean as 10.1399523809524%.",
+    "Compare the average year-on-year net profit growth rate (-79.47492958%) with the R&D investment growth rate (10.1399523809524%), determine that -79.47492958 is less than 10.1399523809524, conclusion: No."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry where Haishan Changgong Equipment Company operates": "General equipment manufacturing",
+    "General equipment manufacturing net profit year-on-year growth rate mean (%)": -79.47492958,
+    "General equipment manufacturing R&D investment year-on-year growth rate mean (%)": 10.1399523809524,
+    "Comparison result": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium024.json b/assets/qa_gold/comprehensive_decision/medium024.json
new file mode 100644
index 0000000000000000000000000000000000000000..eea8a6e2dab0b59fe24fa24dfda5202adde39b85
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium024.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium024",
+  "question": "In 2022, is the region with the highest total operating revenue nationwide also the one with the largest total enterprise market capitalization?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter all provincial records (excluding \"National\"), extract province, total operating revenue amount, and total company market capitalization fields.",
+    "Aggregate total operating revenue amount (yuan) across all industries by province, obtain total operating revenue amount (yuan) for each province. Note: operating revenue amount in the file is in yuan, no conversion needed.",
+    "Identify the region with the highest total operating revenue amount (yuan) as \"Beijing\" with total operating revenue of 54,657,099,005,195.0 yuan.",
+    "Convert total company market capitalization from 10,000 yuan unit to yuan unit (multiply by 10,000), aggregate total company market capitalization (yuan) across all industries by province, obtain total enterprise market capitalization (yuan) for each province.",
+    "Identify the region with the largest total enterprise market capitalization (yuan) as \"Beijing\" with total enterprise market capitalization of 3,959,736,000.0 yuan. Compare the region with the highest total operating revenue amount (yuan) (Beijing) with the region with the largest total enterprise market capitalization (yuan) (Beijing), determine they are the same, conclusion: Yes."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Region with highest total operating revenue amount (yuan)": "Beijing",
+    "Region with largest total enterprise market capitalization (yuan)": "Beijing",
+    "Comparison result": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium025.json b/assets/qa_gold/comprehensive_decision/medium025.json
new file mode 100644
index 0000000000000000000000000000000000000000..5c484bfe9ea839321bb12082ca5f543fafcebd0c
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium025.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium025",
+  "question": "In 2022, is the industry with the highest R&D investment growth rate also the industry with the largest R&D investment?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Only output the answer, do not add any explanation. If the relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all records with district=\"National\" from national_industry_status.csv, extract the fields: industry, year-over-year change in R&D investment (average), and total R&D investment amount.",
+    "Filter industry records where both the year-over-year change in R&D investment (average) and total R&D investment amount are not empty (notna). Note: The total R&D investment amount in the file is in yuan (CNY), no conversion is needed.",
+    "Sort all industries by the year-over-year change in R&D investment (average) field in descending order. The industry with the highest R&D investment growth rate is \"Information transmission, software and information technology services\", with an R&D investment growth rate of 179.78%.",
+    "Sort all industries by the total R&D investment amount field in descending order. The industry with the largest total R&D investment amount is \"Information transmission, software and information technology services\", with a total R&D investment amount of 616535246522.23 yuan.",
+    "Compare the industry with the highest R&D investment growth rate (Information transmission, software and information technology services) with the industry with the largest total R&D investment amount (Information transmission, software and information technology services). Conclude that they are the same, so the answer is: Yes."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry with the highest R&D investment growth rate": "Information transmission, software and information technology services",
+    "Maximum R&D investment growth rate (%)": 179.78,
+    "Industry with the largest total R&D investment amount": "Information transmission, software and information technology services",
+    "Maximum total R&D investment amount (yuan)": 616535246522.23,
+    "Comparison result": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium026.json b/assets/qa_gold/comprehensive_decision/medium026.json
new file mode 100644
index 0000000000000000000000000000000000000000..51bc5baf5f2ab45b62547c875f15c3e54de66ad8
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium026.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium026",
+  "question": "In 2022, does the industry with the highest average asset-liability ratio in Guangdong Province belong to the industry with the highest average asset-liability ratio nationwide?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Only output the answer, do not add any explanation. If the relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all records with province=\"Guangdong Province\" from regional_industry_status.csv, extract the fields: industry and average asset-liability ratio.",
+    "Filter industry records where the average asset-liability ratio is not empty. Sort by the average asset-liability ratio field in descending order. The industry with the highest average asset-liability ratio in Guangdong Province is \"Textile, footwear and apparel industry\", with an average asset-liability ratio of 724.30%.",
+    "Filter all records with district=\"National\" from national_industry_status.csv, extract the fields: industry and average asset-liability ratio.",
+    "Filter industry records where the average asset-liability ratio is not empty. Sort by the average asset-liability ratio field in descending order. The industry with the highest average asset-liability ratio nationwide is \"Financial industry\", with an average asset-liability ratio of 518.42%.",
+    "Compare the industry with the highest average asset-liability ratio in Guangdong Province (Textile, footwear and apparel industry) with the industry with the highest average asset-liability ratio nationwide (Financial industry). Conclude that they are not the same, so the answer is: No."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry with the highest average asset-liability ratio in Guangdong Province": "Textile, footwear and apparel industry",
+    "Industry with the highest average asset-liability ratio nationwide": "Financial industry",
+    "Comparison result": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium027.json b/assets/qa_gold/comprehensive_decision/medium027.json
new file mode 100644
index 0000000000000000000000000000000000000000..12e60be03cacbeb78d626f226d68548039e1fe86
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium027.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium027",
+  "question": "In 2022, Compare the enterprise with the highest R&D input-output ratio in Guangdong Province and the enterprise with the highest R&D input-output ratio in Sichuan Province (R&D input-output ratio = operating revenue amount / R&D investment amount). Which enterprise has higher operating revenue? The unit of operating revenue is yuan.",
+  "guidelines": "The answer must be \"yes\" or \"no\". Output only the answer, without any additional explanation. If the relevant data cannot be found, please answer \"No relevant data retrieved\".",
+  "answer": "The enterprise with the highest R&D input-output ratio in Guangdong Province (Wulichangyuan Wholesale Company) has higher operating revenue",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all enterprise records with province=\"Guangdong Province\" and province=\"Sichuan Province\" from company_profile.csv, and extract the enterprise name and province fields.",
+    "Filter all records with year=2022 from company_operation_status.csv, and extract the enterprise name, operating revenue amount, and R&D investment amount fields. Note: The operating revenue amount in the file is in yuan; no conversion is needed.",
+    "Filter enterprise records where both operating revenue amount and R&D investment amount are not null and R&D investment amount is greater than 0. Calculate R&D input-output ratio = operating revenue amount / R&D investment amount.",
+    "Identify the enterprise with the highest R&D input-output ratio in Guangdong Province as \"Wulichangyuan Wholesale Company\", with an R&D input-output ratio of 8100.17 and operating revenue of 73,443,148,744.17 yuan.",
+    "Identify the enterprise with the highest R&D input-output ratio in Sichuan Province as \"Guotouzeyuan New Energy Company\", with an R&D input-output ratio of 56708.37 and operating revenue of 2,377,118,263.90 yuan. Compare the enterprise with the highest R&D input-output ratio in Guangdong Province (Wulichangyuan Wholesale Company, operating revenue 73,443,148,744.17 yuan) with the enterprise with the highest R&D input-output ratio in Sichuan Province (Guotouzeyuan New Energy Company, operating revenue 2,377,118,263.90 yuan), and determine that the Guangdong Province enterprise has higher operating revenue."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Enterprise with highest R&D input-output ratio in Guangdong Province": "Wulichangyuan Wholesale Company",
+    "Enterprise with highest R&D input-output ratio in Sichuan Province": "Guotouzeyuan New Energy Company",
+    "Enterprise with higher operating revenue": "Enterprise with highest R&D input-output ratio in Guangdong Province (Wulichangyuan Wholesale Company)"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium028.json b/assets/qa_gold/comprehensive_decision/medium028.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b3829cd692b0e5e4584b84604827041a55877b7
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium028.json
@@ -0,0 +1,59 @@
+{
+  "id": "medium028",
+  "question": "In 2022, for the enterprise ranked first in year-over-year growth rate of R&D investment in the chemical raw materials and chemical products manufacturing industry, which of its indicators perform below the industry average?",
+  "guidelines": "Output only the indicator name(s) as the answer, without any additional explanation. If the relevant data cannot be found, please answer \"No relevant data retrieved\"",
+  "answer": [
+    "Total assets",
+    "R&D investment ratio",
+    "Capitalized R&D investment",
+    "Year-over-year change in capitalized R&D investment",
+    "Company market capitalization",
+    "Cumulative China patent applications",
+    "Cumulative China invention patent grants",
+    "Cumulative China invention patent invalidations",
+    "Annual China patent applications",
+    "Annual China invention patent applications",
+    "Annual China patent grants",
+    "Annual China invention patent grants",
+    "R&D personnel ratio",
+    "Total liabilities",
+    "Cumulative citations of all patents",
+    "Cumulative China invention patent applications",
+    "Participation in drafting national standards"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all enterprise records with industry=\"chemical raw materials and chemical products manufacturing\" from company_profile.csv, and extract the enterprise name and industry fields.",
+    "Filter all records with year=2022 from company_operation_status.csv, and extract the enterprise name, year-over-year change in R&D investment, and various indicator fields. Note: All amount-related indicators are in yuan; no conversion is needed.",
+    "Filter enterprise records where the year-over-year change in R&D investment is not null, sort by the year-over-year change in R&D investment in descending order, and identify the enterprise ranked first in R&D investment year-over-year growth rate as \"Huayichangze Technology Co., Ltd.\", with an R&D investment year-over-year growth rate of 2766.85%.",
+    "Calculate the average of each indicator for all enterprises in the chemical raw materials and chemical products manufacturing industry (excluding the year-over-year change in R&D investment itself).",
+    "Compare each indicator of this enterprise (Huayichangze Technology Co., Ltd.) with the industry average, and identify the indicators that are below the industry average. A total of 17 indicators were found: total assets, R&D investment ratio, capitalized R&D investment, year-over-year change in capitalized R&D investment, company market capitalization, cumulative China patent applications, cumulative China invention patent grants, cumulative China invention patent invalidations, annual China patent applications, annual China invention patent applications, annual China patent grants, annual China invention patent grants, R&D personnel ratio, total liabilities, cumulative citations of all patents, cumulative China invention patent applications, participation in drafting national standards."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Enterprise ranked first in R&D investment year-over-year growth rate": "Huayichangze Technology Co., Ltd.",
+    "Indicators below industry average": [
+      "Total assets",
+      "R&D investment ratio",
+      "Capitalized R&D investment",
+      "Year-over-year change in capitalized R&D investment",
+      "Company market capitalization",
+      "Cumulative China patent applications",
+      "Cumulative China invention patent grants",
+      "Cumulative China invention patent invalidations",
+      "Annual China patent applications",
+      "Annual China invention patent applications",
+      "Annual China patent grants",
+      "Annual China invention patent grants",
+      "R&D personnel ratio",
+      "Total liabilities",
+      "Cumulative citations of all patents",
+      "Cumulative China invention patent applications",
+      "Participation in drafting national standards"
+    ]
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium029.json b/assets/qa_gold/comprehensive_decision/medium029.json
new file mode 100644
index 0000000000000000000000000000000000000000..d519594065d21285bacc7a70753623f21060b0e8
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium029.json
@@ -0,0 +1,43 @@
+{
+  "id": "medium029",
+  "question": "Total number of all enterprises affected by the policy 'Notice of the General Office of Guangdong Provincial People's Government on Printing and Distributing Several Measures of Guangdong Province for Further Promoting Steady Growth of Industrial Economy' in 2022",
+  "guidelines": "The answer must be \"yes\" or \"no\". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 416,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Search policy_resource.csv for policy records whose policy name contains \"Guangdong Province's Promotion of Steady Industrial Economic Growth\" and \"Guangdong Province\" to identify the policy.",
+    "Extract the industries involved from the industry field of the policy record, obtaining 8 industries: Consumer Electronics and Electrical Equipment, Automobile Manufacturing, Textile and Footwear, Special Purpose Equipment Manufacturing, General Purpose Equipment Manufacturing, Communication Transmission Equipment, Semiconductor, and Other Manufacturing.",
+    "Filter all enterprise records with province=\"Guangdong Province\" from company_profile.csv, and extract the province and industry fields.",
+    "Count the number of enterprises in each involved industry in Guangdong Province: Consumer Electronics and Electrical Equipment 150, Automobile Manufacturing 27, Textile and Footwear 33, Special Purpose Equipment Manufacturing 80, General Purpose Equipment Manufacturing 20, Communication Transmission Equipment 38, Semiconductor 54, Other Manufacturing 14.",
+    "Calculate the total number of enterprises across all involved industries: 150+27+33+80+20+38+54+14=416."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Number of industries involved in the policy": 8,
+    "Involved industries": [
+      "Consumer Electronics and Electrical Equipment",
+      "Automobile Manufacturing",
+      "Textile and Footwear",
+      "Special Purpose Equipment Manufacturing",
+      "General Purpose Equipment Manufacturing",
+      "Communication Transmission Equipment",
+      "Semiconductor",
+      "Other Manufacturing"
+    ],
+    "Enterprise count per industry": {
+      "Consumer Electronics and Electrical Equipment": 150,
+      "Automobile Manufacturing": 27,
+      "Textile and Footwear": 33,
+      "Special Purpose Equipment Manufacturing": 80,
+      "General Purpose Equipment Manufacturing": 20,
+      "Communication Transmission Equipment": 38,
+      "Semiconductor": 54,
+      "Other Manufacturing": 14
+    },
+    "Total number of enterprises": 416
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium030.json b/assets/qa_gold/comprehensive_decision/medium030.json
new file mode 100644
index 0000000000000000000000000000000000000000..80f07dad141421349d708dd7b1a8e692ecf130be
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium030.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium030",
+  "question": "Did both the operating revenue and R&D investment of Haishan Changgong Equipment Company and Sansan Daten Heavy Industry Company show an upward trend in 2022?",
+  "guidelines": "The answer must be \"yes\" or \"no\". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "no",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all records with enterprise names \"Haishan Changgong Equipment Company\" and \"Sansan Daten Heavy Industry Company\" from company_operation_status.csv, and extract the enterprise name, year, operating revenue amount, R&D investment amount, year-over-year change in operating revenue, and year-over-year change in R&D investment fields. Note: amounts are in yuan and do not require conversion.",
+    "Check the data years for both companies and find that only 2022 data is available; therefore use year-over-year change in operating revenue and year-over-year change in R&D investment to determine trends.",
+    "Analyze Haishan Changgong Equipment Company: year-over-year change in operating revenue is -19.28% (decline), year-over-year change in R&D investment is -14.09% (decline); neither operating revenue nor R&D investment shows an upward trend.",
+    "Analyze Sansan Daten Heavy Industry Company: year-over-year change in operating revenue is -9.41% (decline), year-over-year change in R&D investment is 26.97% (increase); operating revenue does not show an upward trend, but R&D investment does.",
+    "Comprehensive conclusion: Neither operating revenue nor R&D investment of Haishan Changgong Equipment Company shows an upward trend; for Sansan Daten Heavy Industry Company, operating revenue does not show an upward trend (only R&D investment increased). Therefore, it is not the case that both operating revenue and R&D investment of both companies show an upward trend. Conclusion: No."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Haishan Changgong Equipment Company year-over-year change in operating revenue (%)": -19.28,
+    "Haishan Changgong Equipment Company year-over-year change in R&D investment (%)": -14.09,
+    "Haishan Changgong Equipment Company trend": "Both operating revenue and R&D investment declined",
+    "Sansan Daten Heavy Industry Company year-over-year change in operating revenue (%)": -9.41,
+    "Sansan Daten Heavy Industry Company year-over-year change in R&D investment (%)": 26.97,
+    "Sansan Daten Heavy Industry Company trend": "Operating revenue declined, R&D investment increased",
+    "Comprehensive conclusion": "Not all showing upward trend"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium031.json b/assets/qa_gold/comprehensive_decision/medium031.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f5c77648fbd16a840cb3c3bce974f09aa04ce9f
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium031.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium031",
+  "question": "In 2022, was Haishan Changgong Equipment Company the only enterprise in its region with both increased operating revenue and R&D investment?",
+  "guidelines": "The answer must be \"yes\" or \"no\". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "no",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Look up the record for enterprise name \"Haishan Changgong Equipment Company\" in company_profile.csv, extract the province field, and obtain the region where the enterprise is located as \"Shandong Province\".",
+    "Filter records with enterprise name \"Haishan Changgong Equipment Company\" and year=2022 from company_operation_status.csv, and extract the year-over-year change in operating revenue and year-over-year change in R&D investment fields.",
+    "Analyze Haishan Changgong Equipment Company's data: year-over-year change in operating revenue is -19.28% (decline), year-over-year change in R&D investment is -14.09% (decline); neither operating revenue nor R&D investment increased.",
+    "Since neither operating revenue nor R&D investment of Haishan Changgong Equipment Company increased, it cannot be the only enterprise in its region with both increased operating revenue and R&D investment. Conclusion: No."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Region where Haishan Changgong Equipment Company is located": "Shandong Province",
+    "Haishan Changgong Equipment Company year-over-year change in operating revenue (%)": -19.28,
+    "Haishan Changgong Equipment Company year-over-year change in R&D investment (%)": -14.09,
+    "Did operating revenue increase": "No",
+    "Did R&D investment increase": "No",
+    "Conclusion": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium032.json b/assets/qa_gold/comprehensive_decision/medium032.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a664dae5e5d98071848b6436b0502efef6b880a
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium032.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium032",
+  "question": "In 2022, was Haishan Changgong Equipment Company the only enterprise in its region with both declined operating revenue and R&D investment?",
+  "guidelines": "The answer must be \"yes\" or \"no\". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "no",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Look up the record for enterprise name \"Haishan Changgong Equipment Company\" in company_profile.csv, extract the province field, and obtain the region where the enterprise is located as \"Shandong Province\".",
+    "Filter records with enterprise name \"Haishan Changgong Equipment Company\" and year=2022 from company_operation_status.csv, and extract the year-over-year change in operating revenue and year-over-year change in R&D investment fields.",
+    "Analyze Haishan Changgong Equipment Company's data: year-over-year change in operating revenue is -19.28% (decline), year-over-year change in R&D investment is -14.09% (decline); both operating revenue and R&D investment declined.",
+    "Filter all enterprise records with province=\"Shandong Province\" and year=2022 from company_operation_status.csv, and extract the enterprise name, year-over-year change in operating revenue, and year-over-year change in R&D investment fields.",
+    "Count the number of enterprises in Shandong Province with both declined operating revenue and R&D investment (year-over-year change in operating revenue < 0 and year-over-year change in R&D investment < 0); a total of 61 enterprises were found. Determine whether Haishan Changgong Equipment Company is the only one: Since 61 enterprises in Shandong Province have both declined operating revenue and R&D investment, Haishan Changgong Equipment Company is not the only one. Conclusion: No."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Region where Haishan Changgong Equipment Company is located": "Shandong Province",
+    "Haishan Changgong Equipment Company year-over-year change in operating revenue (%)": -19.28,
+    "Haishan Changgong Equipment Company year-over-year change in R&D investment (%)": -14.09,
+    "Did operating revenue decline": "Yes",
+    "Did R&D investment decline": "Yes",
+    "Number of enterprises in Shandong Province with both declined operating revenue and R&D investment": 61,
+    "Conclusion": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium033.json b/assets/qa_gold/comprehensive_decision/medium033.json
new file mode 100644
index 0000000000000000000000000000000000000000..e0ac24ae4ecbe4a1a3f8034e62467c16496da236
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium033.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium033",
+  "question": "In the industry with the most invention patent grants in 2022, what is the average asset-liability ratio of enterprises? In which province is this industry concentrated?",
+  "guidelines": "Answer data should retain two decimal places. Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": [
+    85.7,
+    "Guangdong Province"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all records with district=\"National\" from national_industry_status.csv, and extract the industry and total cumulative China invention patent grants fields.",
+    "Filter industry records with non-null total cumulative China invention patent grants, sort by total cumulative China invention patent grants in descending order, and identify the industry with the most invention patent grants as \"Consumer Electronics and Electrical Equipment\" with a total of 217,188 cumulative China invention patent grants.",
+    "Filter industry \"Consumer Electronics and Electrical Equipment\" from national_industry_status.csv to obtain the average asset-liability ratio of 85.70%.",
+    "Filter enterprises with industry \"Consumer Electronics and Electrical Equipment\" from company_profile.csv, and count enterprises by region.",
+    "From the regional enterprise counts, determine the province with the maximum number as \"Guangdong Province\" with 150 enterprises."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry with the most invention patent grants": "Consumer Electronics and Electrical Equipment",
+    "Number of enterprises in this industry": 358,
+    "Average asset-liability ratio (%)": 85.7,
+    "Number of 'Consumer Electronics and Electrical Equipment' enterprises in Guangdong Province": 150
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium034.json b/assets/qa_gold/comprehensive_decision/medium034.json
new file mode 100644
index 0000000000000000000000000000000000000000..9fbe002ae9c38bf7aba0e736d0e289b28ad517f3
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium034.json
@@ -0,0 +1,35 @@
+{
+  "id": "medium034",
+  "question": "Among the provinces that performed best in cultivating high-tech enterprises in 2022 (measured by the number of high-tech enterprises), which high-tech enterprise has the highest year-over-year growth in average operating revenue?",
+  "guidelines": "The answer must be an enterprise name. Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "Langji Ruanchuang Information Technology Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Identify high-tech enterprises from company_profile.csv: enterprises whose firstClassification and secondClassification fields contain keywords such as \"technology\", \"innovation\", \"R&D\", or whose industry field contains keywords such as \"technology\", \"information\", \"software\", \"electronics\", \"communication\".",
+    "Build a mapping from enterprise name to province from company_profile.csv. Count high-tech enterprises by province and identify the top 3 provinces with the most high-tech enterprises: Guangdong Province (89), Beijing (49), Jiangsu Province (48).",
+    "Filter all records with year=2022 from company_operation_status.csv, and extract the enterprise name and year-over-year change in operating revenue fields.",
+    "Filter enterprise records that are high-tech enterprises and have non-null (notna) year-over-year change in operating revenue. For each province, identify the high-tech enterprise with the highest year-over-year growth in operating revenue: Guangdong Province — \"Kuaike Chuangjin Equipment Company\" (47.76%), Beijing — \"Baoxin Ruanlian Software Company\" (51.65%), Jiangsu Province — \"Langji Ruanchuang Information Technology Company\" (93.78%).",
+    "Compare the enterprises with the highest year-over-year growth in operating revenue across these three provinces, and find the highest as \"Langji Ruanchuang Information Technology Company\" (located in Jiangsu Province, 93.78%)."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Top 3 provinces with the most high-tech enterprises": [
+      "Guangdong Province",
+      "Beijing",
+      "Jiangsu Province"
+    ],
+    "Number of high-tech enterprises in Guangdong Province": 89,
+    "Number of high-tech enterprises in Beijing": 49,
+    "Number of high-tech enterprises in Jiangsu Province": 48,
+    "High-tech enterprise with highest year-over-year growth in operating revenue in Guangdong Province": "Kuaike Chuangjin Equipment Company",
+    "High-tech enterprise with highest year-over-year growth in operating revenue in Beijing": "Baoxin Ruanlian Software Company",
+    "High-tech enterprise with highest year-over-year growth in operating revenue in Jiangsu Province": "Langji Ruanchuang Information Technology Company",
+    "Enterprise with highest year-over-year growth in operating revenue": "Langji Ruanchuang Information Technology Company",
+    "Province where this enterprise is located": "Jiangsu Province",
+    "Year-over-year growth in operating revenue (%)": 93.78
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium035.json b/assets/qa_gold/comprehensive_decision/medium035.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c20a32ede21866cff67b405f273925c91c98558
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium035.json
@@ -0,0 +1,38 @@
+{
+  "id": "medium035",
+  "question": "In 2022, which provinces ranked in the top three in terms of market share among enterprises in the cloud computing services sector?",
+  "guidelines": "The answer must be province names. Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": [
+    "Beijing",
+    "Shanghai",
+    "Guangdong Province"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Identify cloud computing services enterprises from company_profile.csv: enterprises with secondClassification=\"Internet and cloud computing, big data services\" or containing keywords such as \"cloud computing\", \"cloud services\", \"cloud platform\".",
+    "Build a mapping from enterprise name to province from company_profile.csv. Filter all records with year=2022 from company_operation_status.csv, and extract the enterprise name and operating revenue amount fields.",
+    "Filter enterprise records in the cloud computing services sector with non-null operating revenue amount, and calculate total operating revenue by province.",
+    "Calculate total operating revenue as 81,064,346,518.77 yuan for computing market share percentages. Sort by total operating revenue by province and identify the top 3 provinces by market share: Beijing (93.92%, 3 enterprises), Shanghai (2.43%, 1 enterprise), Guangdong Province (2.13%, 2 enterprises).",
+    "Determine the provinces with leading market share positions as the top 3: Beijing, Shanghai, and Guangdong Province."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Number of enterprises in cloud computing services sector": 9,
+    "Total operating revenue (yuan)": 81064346518.77,
+    "Top 3 provinces by market share": [
+      "Beijing",
+      "Shanghai",
+      "Guangdong Province"
+    ],
+    "Beijing market share (%)": 93.92,
+    "Number of enterprises in Beijing": 3,
+    "Shanghai market share (%)": 2.43,
+    "Number of enterprises in Shanghai": 1,
+    "Guangdong Province market share (%)": 2.13,
+    "Number of enterprises in Guangdong Province": 2
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium036.json b/assets/qa_gold/comprehensive_decision/medium036.json
new file mode 100644
index 0000000000000000000000000000000000000000..9181f8fcfbeb2232af48befc829e85d51f86a1d9
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium036.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium036",
+  "question": "In 2022, what percentage of total operating revenue in the market do the top 20% enterprises by R&D investment in Sichuan Province's pharmaceutical manufacturing industry account for?",
+  "guidelines": "The answer must be an exact number with 2 decimal places. Output only the number, do not add units, commas, or any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 61.92,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all enterprise records with province=\"Sichuan Province\" and industry=\"Pharmaceutical Manufacturing\" from company_profile.csv, extract the enterprise name and bmCode fields, and find 15 pharmaceutical manufacturing enterprises in Sichuan Province.",
+    "Filter 2022 data for these enterprises from company_operation_status.csv by enterprise name and bmCode, extract the R&D investment amount and operating revenue amount fields, and find 2022 data for 15 pharmaceutical manufacturing enterprises in Sichuan Province.",
+    "Filter enterprise records with non-null (notna) R&D investment amount and operating revenue amount (15 records total), sort all enterprises by R&D investment amount in descending order, determine R&D investment rankings, and identify the top 20% enterprises as \"Huaren Taize Pharmaceutical Co., Ltd., Fuhe Chenze Biopharmaceutical Company, Yishan Shengchen Medical Technology Company\".",
+    "Extract the operating revenue amount for the 3 enterprises: 19038287881.00, 100067338.00, 3389021585.67 respectively. Calculate the sum of operating revenue for the top 20% enterprises as 22527376804.67.",
+    "Calculate the sum of operating revenue for all 15 pharmaceutical manufacturing enterprises in Sichuan Province as 36382939883.72. Calculate percentage = (sum of operating revenue of top 20% enterprises / sum of operating revenue of all enterprises) × 100%, rounded to two decimal places: 61.92%"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Sum of operating revenue of top 20% enterprises (yuan)": 22527376804.67,
+    "Sum of operating revenue of all enterprises (yuan)": 36382939883.72,
+    "Percentage": 61.92
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium037.json b/assets/qa_gold/comprehensive_decision/medium037.json
new file mode 100644
index 0000000000000000000000000000000000000000..5faf4096a6914269d5dec1eeeb6394cf15276f1a
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium037.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium037",
+  "question": "In 2022, what percentage of total operating revenue in the market do the top enterprises by R&D investment (top 3) in Sichuan Province's pharmaceutical manufacturing industry account for?",
+  "guidelines": "The answer must be an exact number with 2 decimal places. Output only the number, do not add units, commas, or any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 61.92,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter all enterprise records with province=\"Sichuan Province\" and industry=\"Pharmaceutical Manufacturing\" from company_profile.csv, extract the enterprise name and bmCode fields, and find 15 pharmaceutical manufacturing enterprises in Sichuan Province.",
+    "Filter 2022 data for these enterprises from company_operation_status.csv by enterprise name and bmCode, extract the R&D investment amount and operating revenue amount fields, and find 2022 data for 15 pharmaceutical manufacturing enterprises in Sichuan Province.",
+    "Filter enterprise records with non-null R&D investment amount and operating revenue amount (15 records total), sort all enterprises by R&D investment amount in descending order, determine R&D investment rankings, and identify the top 20% enterprises as \"Huaren Taize Pharmaceutical Co., Ltd., Fuhe Chenze Biopharmaceutical Company, Yishan Shengchen Medical Technology Company\".",
+    "Extract the operating revenue amount for the 3 enterprises: 19038287881.00, 100067338.00, 3389021585.67 respectively. Calculate the sum of operating revenue for the top 20% enterprises as 22527376804.67.",
+    "Calculate the sum of operating revenue for all 15 pharmaceutical manufacturing enterprises in Sichuan Province as 36382939883.72. Calculate percentage = (sum of operating revenue of top 20% enterprises / sum of operating revenue of all enterprises) × 100%, rounded to two decimal places: 61.92%"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Sum of operating revenue of top 20% enterprises (yuan)": 22527376804.67,
+    "Sum of operating revenue of all enterprises (yuan)": 36382939883.72,
+    "Percentage": 61.92
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium038.json b/assets/qa_gold/comprehensive_decision/medium038.json
new file mode 100644
index 0000000000000000000000000000000000000000..2caba06e9b6a54cf5e036811ede1f6b26830c16c
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium038.json
@@ -0,0 +1,23 @@
+{
+  "id": "medium038",
+  "question": "In 2022, which enterprise has the highest operating revenue in the same region and industry as Sansan Daten Heavy Industry Company?",
+  "guidelines": "The answer must be an enterprise name. Output only the enterprise name, do not add any other explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "Lingong Hangteng Heavy Industry Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Look up the record for Sansan Daten Heavy Industry Company in company_profile.csv, extract the province and industry fields, and determine its region as \"Beijing\" and industry as \"Special Purpose Equipment Manufacturing\".",
+    "Filter all enterprise records with province=\"Beijing\" and industry=\"Special Purpose Equipment Manufacturing\" from company_profile.csv, extract the enterprise name and bmCode fields, and find 38 special purpose equipment manufacturing enterprises in Beijing.",
+    "Filter 2022 data for these enterprises from company_operation_status.csv by enterprise name and bmCode, extract the operating revenue amount field, and find 2022 operating revenue data for 38 special purpose equipment manufacturing enterprises in Beijing.",
+    "Sort all enterprises by operating revenue amount in descending order to determine the operating revenue ranking. Identify the enterprise with the highest operating revenue as \"Lingong Hangteng Heavy Industry Company\" with operating revenue of 80034467972.00 yuan."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Total number of special purpose equipment manufacturing enterprises in Beijing": 38,
+    "Operating revenue of Lingong Hangteng Heavy Industry Company (yuan)": 80034467972.0,
+    "Enterprise with highest operating revenue": "Lingong Hangteng Heavy Industry Company"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium039.json b/assets/qa_gold/comprehensive_decision/medium039.json
new file mode 100644
index 0000000000000000000000000000000000000000..4fde8c0798e69952ab5396d739af76b91cebeb99
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium039.json
@@ -0,0 +1,56 @@
+{
+  "id": "medium039",
+  "question": "In 2022, list all indicators for which Guangdong Province's Information Transmission, Software and Information Technology Services industry has mean values superior to the national average, and sort them by advantage magnitude from high to low.",
+  "guidelines": "Answer format: [Indicator 1, Indicator 2, Indicator 3, ...]. Output only indicator names, separated by commas and spaces, do not add any other explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": [
+    "Mean of year-over-year change in R&D investment ratio",
+    "Mean of year-over-year change in net profit",
+    "Mean of annual PCT patent applications",
+    "Mean of annual PCT invention patent applications",
+    "Mean of government incentive funds and subsidies",
+    "Mean of net profit amount",
+    "Mean of year-over-year change in capitalized R&D investment",
+    "Mean of cumulative PCT patent applications",
+    "Mean of cumulative PCT invention patent applications",
+    "Mean of cumulative citations of all patents",
+    "Mean of cumulative China invention patent grants",
+    "Mean of participation in drafting national standards",
+    "Mean of R&D personnel ratio",
+    "Mean of R&D investment ratio",
+    "Mean of asset-liability ratio"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter records with industry=\"Information Transmission, Software and Information Technology Services\" and district=\"National\" from national_industry_status.csv, extract all fields containing \"mean\" and their values, and obtain national average data. Total number of indicator fields containing mean is 48.",
+    "Filter records with industry=\"Information Transmission, Software and Information Technology Services\" and province=\"Guangdong Province\" from regional_industry_status.csv, extract all fields containing \"mean\" and their values, and obtain Guangdong Province average data. Total number of indicator fields containing mean is 48.",
+    "Iterate through all fields containing \"mean\", compare Guangdong Province and national values. For each field, if Guangdong Province value is greater than national value, mark it as an advantage indicator. Statistics show 15 indicators where Guangdong Province mean is higher than the national average.",
+    "Sort by advantage magnitude from high to low to obtain 15 advantage indicators: mean of year-over-year change in R&D investment ratio, mean of year-over-year change in net profit, mean of annual PCT patent applications, mean of annual PCT invention patent applications, mean of government incentive funds and subsidies, mean of net profit amount, mean of year-over-year change in capitalized R&D investment, mean of cumulative PCT patent applications, mean of cumulative PCT invention patent applications, mean of cumulative citations of all patents, mean of cumulative China invention patent grants, mean of participation in drafting national standards, mean of R&D personnel ratio, mean of R&D investment ratio, mean of asset-liability ratio."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Total number of national indicator fields containing mean": 48,
+    "Total number of Guangdong Province indicator fields containing mean": 48,
+    "Number of advantage indicators in Guangdong Province": 15,
+    "List of advantage indicators": [
+      "Mean of year-over-year change in R&D investment ratio",
+      "Mean of year-over-year change in net profit",
+      "Mean of annual PCT patent applications",
+      "Mean of annual PCT invention patent applications",
+      "Mean of government incentive funds and subsidies",
+      "Mean of net profit amount",
+      "Mean of year-over-year change in capitalized R&D investment",
+      "Mean of cumulative PCT patent applications",
+      "Mean of cumulative PCT invention patent applications",
+      "Mean of cumulative citations of all patents",
+      "Mean of cumulative China invention patent grants",
+      "Mean of participation in drafting national standards",
+      "Mean of R&D personnel ratio",
+      "Mean of R&D investment ratio",
+      "Mean of asset-liability ratio"
+    ]
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium040.json b/assets/qa_gold/comprehensive_decision/medium040.json
new file mode 100644
index 0000000000000000000000000000000000000000..01095fb53098b4db9f70fcf07c37f1742ab0b999
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium040.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium040",
+  "question": "In 2022, in Guangdong Province's Raw Chemical Materials and Chemical Products Manufacturing industry, what share of the market's operating revenue does the company with the highest market capitalization account for?",
+  "guidelines": "The answer must be an exact number, rounded to 2 decimal places. Output only the number, do not add units, commas, or any textual explanation. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 21.92,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter from company_profile.csv all enterprise records with province=\"Guangdong Province\" and industry=\"Raw Chemical Materials and Chemical Products Manufacturing\", extract enterprise name and bmCode fields, finding 41 enterprises in Guangdong Province's Raw Chemical Materials and Chemical Products Manufacturing industry.",
+    "Filter from company_operation_status.csv the 2022 data of these enterprises by enterprise name and bmCode fields, extract company market capitalization and operating revenue amount fields, finding 2022 data for 41 enterprises in Guangdong Province's Raw Chemical Materials and Chemical Products Manufacturing industry.",
+    "Filter enterprise records where both company market capitalization and operating revenue amount are not null (notna), totaling 41 records. Sort all enterprises in descending order by company market capitalization to determine market capitalization ranking. The enterprise with the highest market capitalization is \"Hengyi Shengsheng Technology Co., Ltd.\". The operating revenue amount of \"Hengyi Shengsheng Technology Co., Ltd.\" is 22316919995.89 CNY.",
+    "Calculate the total operating revenue of Guangdong Province's Raw Chemical Materials and Chemical Products Manufacturing industry as 101800752670.91 CNY.",
+    "Calculate share = (operating revenue of highest market cap enterprise / total industry operating revenue) × 100%, rounded to 2 decimal places: 21.92%"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Operating revenue of highest market cap enterprise (CNY)": 22316919995.89,
+    "Total operating revenue of Guangdong Province Raw Chemical Materials and Chemical Products Manufacturing industry (CNY)": 101800752670.91,
+    "Share": 21.92
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium041.json b/assets/qa_gold/comprehensive_decision/medium041.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d1f9c65a6553cc8a41056ff3cf7dd841f2325cd
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium041.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium041",
+  "question": "In 2022, nationwide, is the province with the highest mean market capitalization also the province with the highest total net profit?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, do not add any other explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Extract province information for all enterprises from the province field in company_profile.csv, totaling 6895 enterprise records.",
+    "Extract company market capitalization and net profit amount fields for all enterprises from company_operation_status.csv, totaling 6895 enterprise records. Join company_profile.csv and company_operation_status.csv by enterprise name and bmCode fields to merge the two data sources, resulting in 6895 records.",
+    "Filter enterprise records where company market capitalization is not null and greater than 0, totaling 6885 records. Group by province and calculate the mean market capitalization for each province. Sort by mean market capitalization in descending order. The province with the highest mean market capitalization is Taiwan Province, with a mean market cap of 358.689 billion CNY and 12 enterprises.",
+    "Filter enterprise records where net profit amount is not null, totaling 6895 records. Group by province and calculate the total net profit and mean net profit for each province. Sort by total net profit in descending order. The province with the highest total net profit is Beijing, with a total net profit of 4998709315999.03 CNY and 774 enterprises.",
+    "Compare the province with the highest mean market capitalization (Taiwan Province) and the province with the highest total net profit (Beijing). They are not the same, so the conclusion is: No."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province with highest mean market capitalization": "Taiwan Province",
+    "Province with highest total net profit": "Beijing",
+    "Comparison result": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium042.json b/assets/qa_gold/comprehensive_decision/medium042.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a594c7620c85f67acaf39eda31070a06279e9f8
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium042.json
@@ -0,0 +1,25 @@
+{
+  "id": "medium042",
+  "question": "In 2022, among enterprises in regions applicable to the \"Notice on Organizing Applications for First Home Purchase Subsidies for Outstanding Young Talents in the Biomedicine Industry\" policy, what is the net profit of the enterprise with the best operating revenue performance?",
+  "guidelines": "The answer must be an exact number, in CNY, rounded to 2 decimal places. Output only the number, do not add units, commas, or any textual explanation. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 4253373290.96,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter from policy_release_status.csv for policy name=\"Notice on Organizing Applications for First Home Purchase Subsidies for Outstanding Young Talents in the Biomedicine Industry\", extract province field, and determine that the applicable region is \"Guangdong Province\" and the applicable industry is \"Pharmaceutical Manufacturing\".",
+    "Filter from company_profile.csv all enterprise records with province=\"Guangdong Province\" and industry=\"Pharmaceutical Manufacturing\", extract enterprise name and bmCode fields, finding 51 enterprises in Guangdong Province's Pharmaceutical Manufacturing industry.",
+    "Filter from company_operation_status.csv the 2022 data of the above enterprises by bmCode field, extract operating revenue amount and net profit amount fields, obtaining 51 records of enterprise 2022 operating data.",
+    "Sort all enterprises by operating revenue amount in descending order. The enterprise with the highest operating revenue is \"Yishan Zeyuan Pharmaceutical Co., Ltd.\".",
+    "Extract the net profit amount field of this enterprise, obtaining a net profit of 4253373290.96."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Applicable region": "Guangdong Province",
+    "Applicable industry": "Pharmaceutical Manufacturing",
+    "Enterprise with highest operating revenue": "Yishan Zeyuan Pharmaceutical Co., Ltd.",
+    "Net profit amount of enterprise with highest operating revenue (CNY)": 4253373290.96
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium043.json b/assets/qa_gold/comprehensive_decision/medium043.json
new file mode 100644
index 0000000000000000000000000000000000000000..8a9ccad31a95bc256119c31c824cbf46d4ba872e
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium043.json
@@ -0,0 +1,23 @@
+{
+  "id": "medium043",
+  "question": "In 2022, for Pharmaceutical Manufacturing, is the province with the highest total assets also the province with the highest R&D investment?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only \"Yes\" or \"No\", do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Filter from regional_industry_status.csv all provincial records with industry=\"Pharmaceutical Manufacturing\".",
+    "In the filtered results, sort by total assets in descending order. The province with the maximum total assets is \"Beijing\", with total assets of 813813599815.91.",
+    "In the same filtered results, sort by total R&D investment amount in descending order. The province with the maximum total R&D investment amount is \"Beijing\", with total R&D investment of 57722993168.07.",
+    "Compare whether the provinces with the maximum values for both indicators are the same: Beijing == Beijing, they are consistent, so the answer is \"Yes\"."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province with highest total assets in Pharmaceutical Manufacturing": "Beijing",
+    "Province with highest total R&D investment in Pharmaceutical Manufacturing": "Beijing",
+    "Whether the two top provinces are the same": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium044.json b/assets/qa_gold/comprehensive_decision/medium044.json
new file mode 100644
index 0000000000000000000000000000000000000000..fae33e206147e72db5423a9cf15aa6e6c29f872b
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium044.json
@@ -0,0 +1,25 @@
+{
+  "id": "medium044",
+  "question": "In 2022, in Shandong Province's Pharmaceutical Manufacturing industry, is Haishan Changgong Equipment Company's R&D investment ratio higher than the R&D investment ratio of the 10th-ranked Pharmaceutical Manufacturing enterprise in Hunan Province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Search for records with enterprise name=\"Haishan Changgong Equipment Company\" in company_profile.csv. The enterprise has bmCode=100071, industry=\"General Equipment Manufacturing\", province=\"Shandong Province\".",
+    "Filter from company_operation_status.csv the 2022 data by enterprise name=\"Haishan Changgong Equipment Company\" and bmCode=100071, extract the R&D investment ratio field. Haishan Changgong Equipment Company's R&D investment ratio is 4.33%.",
+    "Filter from company_profile.csv all enterprise records with province=\"Hunan Province\" and industry=\"Pharmaceutical Manufacturing\", extract enterprise name and bmCode fields, finding 11 enterprises in Hunan Province's Pharmaceutical Manufacturing industry.",
+    "Filter from company_operation_status.csv the 2022 data of these enterprises by enterprise name and bmCode fields, extract the R&D investment ratio field. Filter enterprise records where R&D investment ratio is not null, totaling 11 records.",
+    "Sort all enterprises by R&D investment ratio in descending order to determine the R&D investment ranking. The 10th-ranked enterprise is \"Ruiying Taiyuan Medical Equipment Co., Ltd.\" (bmCode=823535), with an R&D investment ratio of 2.87%. Compare Haishan Changgong Equipment Company's R&D investment ratio (4.33%) with the 10th-ranked Pharmaceutical Manufacturing enterprise in Hunan Province's R&D investment ratio (2.87%). Since 4.33% > 2.87%, the conclusion is: Yes."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "R&D investment ratio of Haishan Changgong Equipment Company (%)": 4.33,
+    "10th-ranked enterprise in Hunan Province Pharmaceutical Manufacturing": "Ruiying Taiyuan Medical Equipment Co., Ltd.",
+    "R&D investment ratio of 10th-ranked enterprise in Hunan Province Pharmaceutical Manufacturing (%)": 2.87,
+    "Comparison result": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/comprehensive_decision/medium045.json b/assets/qa_gold/comprehensive_decision/medium045.json
new file mode 100644
index 0000000000000000000000000000000000000000..7a27b282872dc815d32c2b7a54427c369162774c
--- /dev/null
+++ b/assets/qa_gold/comprehensive_decision/medium045.json
@@ -0,0 +1,25 @@
+{
+  "id": "medium045",
+  "question": "In 2022, in Sichuan Province, is Zhongbai Jinmao Chain Company's R&D investment higher than the R&D investment of the 15th-ranked enterprise nationwide in its industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  },
+  "steps": [
+    "Search for records with enterprise name=\"Zhongbai Jinmao Chain Company\" in company_profile.csv. The enterprise has bmCode=100120, industry=\"Wholesale and Retail Trade\", province=\"Sichuan Province\".",
+    "Filter from company_operation_status.csv the 2022 data by enterprise name=\"Zhongbai Jinmao Chain Company\" and bmCode=100120, extract the R&D investment amount field. Zhongbai Jinmao Chain Company's R&D investment amount is 11270987.0 CNY.",
+    "Filter from company_profile.csv all enterprise records with industry=\"Wholesale and Retail Trade\", extract enterprise name and bmCode fields, finding 273 enterprises in the Wholesale and Retail Trade industry.",
+    "Filter from company_operation_status.csv the 2022 data of these enterprises by enterprise name and bmCode fields, extract the R&D investment amount field. Filter enterprise records where R&D investment amount is not null, totaling 143 records.",
+    "Sort all enterprises by R&D investment amount in descending order to determine the R&D investment ranking. The 15th-ranked enterprise is \"Lianhua Tongze Trading Company\", with an R&D investment amount of 265616054.7 CNY. Compare Zhongbai Jinmao Chain Company's R&D investment amount (11270987.0 CNY) with the 15th-ranked nationwide enterprise in Wholesale and Retail Trade (265616054.7 CNY). Since 11270987.0 < 265616054.7, the conclusion is: No."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "R&D investment amount of Zhongbai Jinmao Chain Company (CNY)": 11270987.0,
+    "15th-ranked enterprise nationwide in Wholesale and Retail Trade": "Lianhua Tongze Trading Company",
+    "R&D investment amount of 15th-ranked enterprise nationwide in Wholesale and Retail Trade (CNY)": 265616054.7,
+    "Comparison result": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy001.json b/assets/qa_gold/enterprise_industry_analysis/easy001.json
new file mode 100644
index 0000000000000000000000000000000000000000..c9586ca75795956ab9040823254ee045091e293b
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy001.json
@@ -0,0 +1,23 @@
+{
+  "id": "easy001",
+  "question": "In 2022, which is higher: the total number of citations of all patents of Zhong Ji Da Chang Tong Ye Co., Ltd. or the industry median benchmark?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Zhong Ji Da Chang Tong Ye Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Zhong Ji Da Chang Tong Ye Co., Ltd. total citations of all patents in 2022 = 558",
+    "Extracted from company_profile.csv: the company belongs to Metal Products; extracted from national_industry_status.csv: the industry benchmark (median) is 484",
+    "Compared 558 and 484; since 558 > 484, the company is higher, so output \"Zhong Ji Da Chang Tong Ye Co., Ltd.\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Zhong Ji Da Chang Tong Ye Co., Ltd. total citations of all patents in 2022": 558,
+    "Industry of Zhong Ji Da Chang Tong Ye Co., Ltd.": "Metal Products",
+    "Industry benchmark (median) for total citations of all patents in Metal Products": 484,
+    "Comparison result (whether the company is higher than the industry median)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy002.json b/assets/qa_gold/enterprise_industry_analysis/easy002.json
new file mode 100644
index 0000000000000000000000000000000000000000..1a37468244e67a70c1756f1921b27ab4e4fdef17
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy002.json
@@ -0,0 +1,23 @@
+{
+  "id": "easy002",
+  "question": "In 2022, what is the difference between the total liabilities of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. and the median total liabilities of its industry?",
+  "guidelines": "The answer must be a number with three decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 493355813.605,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. total liabilities in 2022 = 5309105105.07 CNY",
+    "Extracted from company_profile.csv: the company belongs to Construction; extracted from national_industry_status.csv: industry median total liabilities = 4815749291.465 CNY",
+    "Calculated by requirement: difference (company - industry median) = 5309105105.07 - 4815749291.465 = 493355813.605, output with three decimal places."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. total liabilities in 2022 (CNY)": 5309105105.07,
+    "Industry of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd.": "Construction",
+    "Median total liabilities in the Construction industry (CNY)": 4815749291.465,
+    "Difference (company - industry median)": 493355813.605
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy003.json b/assets/qa_gold/enterprise_industry_analysis/easy003.json
new file mode 100644
index 0000000000000000000000000000000000000000..9fd9a8b4b5383b67eb73c42dcf1deab3b148232d
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy003.json
@@ -0,0 +1,23 @@
+{
+  "id": "easy003",
+  "question": "In 2022, which is lower: the year-over-year change rate of operating profit of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. or the average of this indicator in its industry?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. year-over-year operating profit change rate in 2022 = -263.46 %",
+    "Extracted from company_profile.csv: the company belongs to Construction; extracted from national_industry_status.csv: industry average = -80.1477027027027 %",
+    "Compared -263.46 and -80.1477027027027; since -263.46 < -80.1477027027027, the company is lower, so the answer is \"Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd.\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. year-over-year operating profit change rate in 2022": "-263.46 %",
+    "Industry of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd.": "Construction",
+    "Average year-over-year operating profit change rate in the Construction industry": "-80.1477027027027 %",
+    "Comparison result (whether the company is lower than the industry average)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy004.json b/assets/qa_gold/enterprise_industry_analysis/easy004.json
new file mode 100644
index 0000000000000000000000000000000000000000..53f4a3ab2f2ac962f960dca66e4c7c3509ca89f7
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy004.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy004",
+  "question": "In 2022, is the year-over-year net profit growth rate of Zhong Tong Jie Tong Yun Shu Co., Ltd. lower than that of Yun Da Hang Chang Kuai Di Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: year-over-year net profit growth rate of Zhong Tong Jie Tong Yun Shu Co., Ltd. in 2022 = -4.46 %",
+    "Extracted from company_operation_status.csv: year-over-year net profit growth rate of Yun Da Hang Chang Kuai Di Co., Ltd. in 2022 = 377.60 %",
+    "Compared -4.46 and 377.60; since -4.46 < 377.60, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Year-over-year net profit growth rate of Zhong Tong Jie Tong Yun Shu Co., Ltd. in 2022": "-4.46 %",
+    "Year-over-year net profit growth rate of Yun Da Hang Chang Kuai Di Co., Ltd. in 2022": "377.60 %",
+    "Comparison result (whether Zhong Tong Jie Tong Yun Shu Co., Ltd. is lower than Yun Da Hang Chang Kuai Di Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy005.json b/assets/qa_gold/enterprise_industry_analysis/easy005.json
new file mode 100644
index 0000000000000000000000000000000000000000..41134799fb6b59353c13f2b529a97cf9c439d59a
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy005.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy005",
+  "question": "In 2022, is the market capitalization of Zhong Tong Jie Tong Yun Shu Co., Ltd. lower than that of Yun Da Hang Chang Kuai Di Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: market capitalization of Zhong Tong Jie Tong Yun Shu Co., Ltd. in 2022 = 33.0 hundred million CNY",
+    "Extracted from company_operation_status.csv: market capitalization of Yun Da Hang Chang Kuai Di Co., Ltd. in 2022 = 136.0 hundred million CNY",
+    "Compared 33.0 and 136.0; since 33.0 < 136.0, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Market capitalization of Zhong Tong Jie Tong Yun Shu Co., Ltd. in 2022 (hundred million CNY)": 33.0,
+    "Market capitalization of Yun Da Hang Chang Kuai Di Co., Ltd. in 2022 (hundred million CNY)": 136.0,
+    "Comparison result (whether Zhong Tong Jie Tong Yun Shu Co., Ltd. is lower than Yun Da Hang Chang Kuai Di Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy006.json b/assets/qa_gold/enterprise_industry_analysis/easy006.json
new file mode 100644
index 0000000000000000000000000000000000000000..daad3814b7763ce5f72c3d00ae45c940ff74d18e
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy006.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy006",
+  "question": "In 2022, is the annual number of authorized Chinese invention patents of Huan Xing Jin Ya Shi Shang Co., Ltd. lower than that of Li Ding Sheng Shang Fang Zhi Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: annual number of authorized Chinese invention patents of Huan Xing Jin Ya Shi Shang Co., Ltd. in 2022 = 8",
+    "Extracted from company_operation_status.csv: annual number of authorized Chinese invention patents of Li Ding Sheng Shang Fang Zhi Co., Ltd. in 2022 = 16",
+    "Compared 8 and 16; since 8 < 16, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Annual number of authorized Chinese invention patents of Huan Xing Jin Ya Shi Shang Co., Ltd. in 2022": 8,
+    "Annual number of authorized Chinese invention patents of Li Ding Sheng Shang Fang Zhi Co., Ltd. in 2022": 16,
+    "Comparison result (whether Huan Xing Jin Ya Shi Shang Co., Ltd. is lower than Li Ding Sheng Shang Fang Zhi Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy007.json b/assets/qa_gold/enterprise_industry_analysis/easy007.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b55b7cf769421cbea22190a620fc7cb5c60ae67
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy007.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy007",
+  "question": "In 2022, is the number of national standards participated in drafting by Bao Xin Ke Hui Ruan Jian Co., Ltd. the same as that of Zhong Ke Chuang Xin Ruan Jian Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: number of national standards participated in drafting by Bao Xin Ke Hui Ruan Jian Co., Ltd. in 2022 = 1",
+    "Extracted from company_operation_status.csv: number of national standards participated in drafting by Zhong Ke Chuang Xin Ruan Jian Co., Ltd. in 2022 = 1",
+    "Compared 1 and 1; since they are equal, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Number of national standards participated in drafting by Bao Xin Ke Hui Ruan Jian Co., Ltd. in 2022": 1,
+    "Number of national standards participated in drafting by Zhong Ke Chuang Xin Ruan Jian Co., Ltd. in 2022": 1,
+    "Comparison result (whether they are the same)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy008.json b/assets/qa_gold/enterprise_industry_analysis/easy008.json
new file mode 100644
index 0000000000000000000000000000000000000000..6fd3c3cf4970d6b36ce8e1b257d92ed475f666e4
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy008.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy008",
+  "question": "In 2022, compared with Zhong Ke Chuang Xin Ruan Jian Co., Ltd., what is the difference in cumulative PCT patent applications of Bao Xin Ke Hui Ruan Jian Co., Ltd.?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number without units, commas, or any text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -483.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: cumulative PCT patent applications of Bao Xin Ke Hui Ruan Jian Co., Ltd. in 2022 = 7",
+    "Extracted from company_operation_status.csv: cumulative PCT patent applications of Zhong Ke Chuang Xin Ruan Jian Co., Ltd. in 2022 = 490",
+    "Calculated the difference (Bao Xin Ke Hui Ruan Jian Co., Ltd. - Zhong Ke Chuang Xin Ruan Jian Co., Ltd.): 7 - 490 = -483.0, and output with one decimal place as required"
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Cumulative PCT patent applications of Bao Xin Ke Hui Ruan Jian Co., Ltd. in 2022": 7,
+    "Cumulative PCT patent applications of Zhong Ke Chuang Xin Ruan Jian Co., Ltd. in 2022": 490,
+    "Difference (Bao Xin Ke Hui Ruan Jian Co., Ltd. - Zhong Ke Chuang Xin Ruan Jian Co., Ltd.)": -483.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy009.json b/assets/qa_gold/enterprise_industry_analysis/easy009.json
new file mode 100644
index 0000000000000000000000000000000000000000..861c5a111f4d874354c1ffb22816ba902e523bb6
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy009.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy009",
+  "question": "In 2022, is the total number of employees of Can Xin Hui Xin Semiconductor Co., Ltd. higher than that of Rui Xin Xin Yao Materials Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Can Xin Hui Xin Semiconductor Co., Ltd. total employees in 2022 = 7277.0",
+    "Extracted from company_operation_status.csv: Rui Xin Xin Yao Materials Co., Ltd. total employees in 2022 = 2989.0",
+    "Compared 7277.0 and 2989.0; since 7277.0 > 2989.0, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Can Xin Hui Xin Semiconductor Co., Ltd. total employees in 2022": 7277.0,
+    "Rui Xin Xin Yao Materials Co., Ltd. total employees in 2022": 2989.0,
+    "Comparison result (whether Can Xin Hui Xin Semiconductor Co., Ltd. is higher than Rui Xin Xin Yao Materials Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy010.json b/assets/qa_gold/enterprise_industry_analysis/easy010.json
new file mode 100644
index 0000000000000000000000000000000000000000..e2164a942ad377ebec14c07f39b3f845d1e125e0
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy010.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy010",
+  "question": "In 2022, is the cumulative number of granted Chinese invention patents of Chuang Wei Yao Yao Dian Qi Co., Ltd. lower than that of Mei Neng Dian Jin Technology Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Chuang Wei Yao Yao Dian Qi Co., Ltd. cumulative granted Chinese invention patents in 2022 = 59",
+    "Extracted from company_operation_status.csv: Mei Neng Dian Jin Technology Co., Ltd. cumulative granted Chinese invention patents in 2022 = 140",
+    "Compared 59 and 140; since 59 < 140, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Chuang Wei Yao Yao Dian Qi Co., Ltd. cumulative granted Chinese invention patents in 2022": 59,
+    "Mei Neng Dian Jin Technology Co., Ltd. cumulative granted Chinese invention patents in 2022": 140,
+    "Comparison result (whether Chuang Wei Yao Yao Dian Qi Co., Ltd. is lower than Mei Neng Dian Jin Technology Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy011.json b/assets/qa_gold/enterprise_industry_analysis/easy011.json
new file mode 100644
index 0000000000000000000000000000000000000000..db2a57aefca9f9efbfd634334e059cb55d837151
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy011.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy011",
+  "question": "In 2022, is the R&D investment ratio of Yong Feng Xin Ruan Network Co., Ltd. higher than that of Jin Fei Shu Ruan Data Services Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Yong Feng Xin Ruan Network Co., Ltd. R&D investment ratio in 2022 = 11.7 %",
+    "Extracted from company_operation_status.csv: Jin Fei Shu Ruan Data Services Co., Ltd. R&D investment ratio in 2022 = 5.85 %",
+    "Compared 11.7 and 5.85; since 11.7 > 5.85, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Yong Feng Xin Ruan Network Co., Ltd. R&D investment ratio in 2022": "11.7 %",
+    "Jin Fei Shu Ruan Data Services Co., Ltd. R&D investment ratio in 2022": "5.85 %",
+    "Comparison result (whether Yong Feng Xin Ruan Network Co., Ltd. is higher than Jin Fei Shu Ruan Data Services Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy012.json b/assets/qa_gold/enterprise_industry_analysis/easy012.json
new file mode 100644
index 0000000000000000000000000000000000000000..88c25582260345930b81ce9f45bcf7b23e25837f
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy012.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy012",
+  "question": "In 2022, is the total assets of Yong Feng Xin Ruan Network Co., Ltd. lower than that of Jin Fei Shu Ruan Data Services Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Yong Feng Xin Ruan Network Co., Ltd. total assets in 2022 = 957348449.13 CNY",
+    "Extracted from company_operation_status.csv: Jin Fei Shu Ruan Data Services Co., Ltd. total assets in 2022 = 674647621.10 CNY",
+    "Compared 957348449.13 and 674647621.10; since 957348449.13 > 674647621.10, the condition \"lower than\" is false, so the judgment is \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Yong Feng Xin Ruan Network Co., Ltd. total assets in 2022 (CNY)": 957348449.13,
+    "Jin Fei Shu Ruan Data Services Co., Ltd. total assets in 2022 (CNY)": 674647621.1,
+    "Comparison result (whether Yong Feng Xin Ruan Network Co., Ltd. is lower than Jin Fei Shu Ruan Data Services Co., Ltd.)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy013.json b/assets/qa_gold/enterprise_industry_analysis/easy013.json
new file mode 100644
index 0000000000000000000000000000000000000000..25aa573b49915639973e9721917b60fdcd0b2ac4
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy013.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy013",
+  "question": "In 2022, is the company market value of Wu Li Chang Yuan Wholesale Co., Ltd. higher than that of Wu Li Hui Jin Retail Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Wu Li Chang Yuan Wholesale Co., Ltd. company market value in 2022 = 220.0 (100 million CNY)",
+    "Extracted from company_operation_status.csv: Wu Li Hui Jin Retail Co., Ltd. company market value in 2022 = 0.58 (100 million CNY)",
+    "Compared 220.0 and 0.58; since 220.0 > 0.58, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Wu Li Chang Yuan Wholesale Co., Ltd. company market value in 2022 (100 million CNY)": 220.0,
+    "Wu Li Hui Jin Retail Co., Ltd. company market value in 2022 (100 million CNY)": 0.58,
+    "Comparison result (whether Wu Li Chang Yuan Wholesale Co., Ltd. is higher than Wu Li Hui Jin Retail Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy014.json b/assets/qa_gold/enterprise_industry_analysis/easy014.json
new file mode 100644
index 0000000000000000000000000000000000000000..8487e97d68cdf7c0779b99673d2c54491ce92643
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy014.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy014",
+  "question": "In 2022, is the operating revenue of Wu Li Chang Yuan Pi Fa Co., Ltd. lower than that of Wu Li Hui Jin Ling Shou Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: operating revenue of Wu Li Chang Yuan Pi Fa Co., Ltd. in 2022 = 73443148744.17 CNY",
+    "Extracted from company_operation_status.csv: operating revenue of Wu Li Hui Jin Ling Shou Co., Ltd. in 2022 = 1144710941.0 CNY",
+    "Compared 73443148744.17 and 1144710941.0; since 73443148744.17 > 1144710941.0, the statement \"is lower\" is false, so the judgment is \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Operating revenue of Wu Li Chang Yuan Pi Fa Co., Ltd. in 2022 (CNY)": 73443148744.17,
+    "Operating revenue of Wu Li Hui Jin Ling Shou Co., Ltd. in 2022 (CNY)": 1144710941.0,
+    "Comparison result (whether Wu Li Chang Yuan Pi Fa Co., Ltd. is lower than Wu Li Hui Jin Ling Shou Co., Ltd.)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy015.json b/assets/qa_gold/enterprise_industry_analysis/easy015.json
new file mode 100644
index 0000000000000000000000000000000000000000..db08ae50c28affc9777bf873343aea3b46956f02
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy015.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy015",
+  "question": "In 2022, is the annual number of Chinese patent applications of Mei Neng Xuan Jin Dian Qi Co., Ltd. higher than that of Li Xin Yao Yue Dian Qi Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: annual number of Chinese patent applications of Mei Neng Xuan Jin Dian Qi Co., Ltd. in 2022 = 4552",
+    "Extracted from company_operation_status.csv: annual number of Chinese patent applications of Li Xin Yao Yue Dian Qi Co., Ltd. in 2022 = 1203",
+    "Compared 4552 and 1203; since 4552 > 1203, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Annual number of Chinese patent applications of Mei Neng Xuan Jin Dian Qi Co., Ltd. in 2022": 4552,
+    "Annual number of Chinese patent applications of Li Xin Yao Yue Dian Qi Co., Ltd. in 2022": 1203,
+    "Comparison result (whether Mei Neng Xuan Jin Dian Qi Co., Ltd. is higher than Li Xin Yao Yue Dian Qi Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy016.json b/assets/qa_gold/enterprise_industry_analysis/easy016.json
new file mode 100644
index 0000000000000000000000000000000000000000..7aed1690912cb0ebeff1edd3d22060377bcb5384
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy016.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy016",
+  "question": "In 2022, is the annual number of authorized Chinese invention patents of Mei Neng Xuan Jin Dian Qi Co., Ltd. higher than the annual number of Chinese invention patent applications of Hai Li Chuang Yao Jia Dian Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: annual number of authorized Chinese invention patents of Mei Neng Xuan Jin Dian Qi Co., Ltd. in 2022 = 3304",
+    "Extracted from company_operation_status.csv: annual number of Chinese invention patent applications of Hai Li Chuang Yao Jia Dian Co., Ltd. in 2022 = 8",
+    "Compared 3304 and 8; since 3304 > 8, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Annual number of authorized Chinese invention patents of Mei Neng Xuan Jin Dian Qi Co., Ltd. in 2022": 3304,
+    "Annual number of Chinese invention patent applications of Hai Li Chuang Yao Jia Dian Co., Ltd. in 2022": 8,
+    "Comparison result (whether Mei Neng Xuan Jin Dian Qi Co., Ltd. is higher than Hai Li Chuang Yao Jia Dian Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy017.json b/assets/qa_gold/enterprise_industry_analysis/easy017.json
new file mode 100644
index 0000000000000000000000000000000000000000..253d2e2de079e2a9793f94e6898bef8d639b12bf
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy017.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy017",
+  "question": "In 2022, is the number of provincial or ministerial science and technology progress awards of Mei Neng Xuan Jin Dian Qi Co., Ltd. lower than that of Hai Li Chuang Yao Jia Dian Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: number of provincial or ministerial science and technology progress awards of Mei Neng Xuan Jin Dian Qi Co., Ltd. in 2022 = 2",
+    "Extracted from company_operation_status.csv: number of provincial or ministerial science and technology progress awards of Hai Li Chuang Yao Jia Dian Co., Ltd. in 2022 = 0",
+    "Compared 2 and 0; since 2 < 0 is false, the judgment is \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Number of provincial or ministerial science and technology progress awards of Mei Neng Xuan Jin Dian Qi Co., Ltd. in 2022": 2,
+    "Number of provincial or ministerial science and technology progress awards of Hai Li Chuang Yao Jia Dian Co., Ltd. in 2022": 0,
+    "Comparison result (whether Mei Neng Xuan Jin Dian Qi Co., Ltd. is lower than Hai Li Chuang Yao Jia Dian Co., Ltd.)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy018.json b/assets/qa_gold/enterprise_industry_analysis/easy018.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e9e9e2063d598d62c033e7a38b680b101630ad8
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy018.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy018",
+  "question": "In 2022, compared with Jing Neng Dian Re Ran Qi Co., Ltd., what is the difference in total liabilities of San Xia Ze Neng Dian Li Co., Ltd.?",
+  "guidelines": "The answer must be a number with two decimal places. Output only the number without units, commas, or any text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 2070319837.92,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: total liabilities of San Xia Ze Neng Dian Li Co., Ltd. in 2022 = 10344552545.65 CNY",
+    "Extracted from company_operation_status.csv: total liabilities of Jing Neng Dian Re Ran Qi Co., Ltd. in 2022 = 8274232707.73 CNY",
+    "Calculated the difference (San Xia Ze Neng Dian Li Co., Ltd. - Jing Neng Dian Re Ran Qi Co., Ltd.): 10344552545.65 - 8274232707.73 = 2070319837.92, and output with two decimal places as required"
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Total liabilities of San Xia Ze Neng Dian Li Co., Ltd. in 2022 (CNY)": 10344552545.65,
+    "Total liabilities of Jing Neng Dian Re Ran Qi Co., Ltd. in 2022 (CNY)": 8274232707.73,
+    "Difference (San Xia Ze Neng Dian Li Co., Ltd. - Jing Neng Dian Re Ran Qi Co., Ltd.)": 2070319837.92
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy019.json b/assets/qa_gold/enterprise_industry_analysis/easy019.json
new file mode 100644
index 0000000000000000000000000000000000000000..72fd0fa06f7e2860e8948e0e41e8de26a6eff79b
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy019.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy019",
+  "question": "In 2022, is the R&D investment ratio of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. lower than that of Yi Shan Tai Tai Medical Devices Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. R&D investment ratio in 2022 = 3.96 %",
+    "Extracted from company_operation_status.csv: Yi Shan Tai Tai Medical Devices Co., Ltd. R&D investment ratio in 2022 = 11.1 %",
+    "Compared 3.96 and 11.1; since 3.96 < 11.1, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. R&D investment ratio in 2022": "3.96 %",
+    "Yi Shan Tai Tai Medical Devices Co., Ltd. R&D investment ratio in 2022": "11.1 %",
+    "Comparison result (whether Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. is lower than Yi Shan Tai Tai Medical Devices Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy020.json b/assets/qa_gold/enterprise_industry_analysis/easy020.json
new file mode 100644
index 0000000000000000000000000000000000000000..74138a43fe169f00e765a541007d63bf71d127cf
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy020.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy020",
+  "question": "In 2022, is the debt-to-asset ratio of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. higher than that of Yi Shan Tai Tai Medical Devices Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. debt-to-asset ratio in 2022 = 63.23 %",
+    "Extracted from company_operation_status.csv: Yi Shan Tai Tai Medical Devices Co., Ltd. debt-to-asset ratio in 2022 = 39.97 %",
+    "Compared 63.23 and 39.97; since 63.23 > 39.97, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. debt-to-asset ratio in 2022": "63.23 %",
+    "Yi Shan Tai Tai Medical Devices Co., Ltd. debt-to-asset ratio in 2022": "39.97 %",
+    "Comparison result (whether Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. is higher than Yi Shan Tai Tai Medical Devices Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy021.json b/assets/qa_gold/enterprise_industry_analysis/easy021.json
new file mode 100644
index 0000000000000000000000000000000000000000..d2b5c7fdf429d98359a6a9c1fab6fb0444827dab
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy021.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy021",
+  "question": "In 2022, is the total citations of all patents of Guang Sheng Chang Ze Group Co., Ltd. higher than the corresponding indicator of Lang Ji Yun Hui Technology Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Guang Sheng Chang Ze Group Co., Ltd. total citations of all patents in 2022 = 199",
+    "Extracted from company_operation_status.csv: Lang Ji Yun Hui Technology Co., Ltd. total citations of all patents in 2022 = 2107",
+    "Compared 199 and 2107; since 199 < 2107, the judgment is \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Guang Sheng Chang Ze Group Co., Ltd. total citations of all patents in 2022": 199,
+    "Lang Ji Yun Hui Technology Co., Ltd. total citations of all patents in 2022": 2107,
+    "Comparison result (whether Guang Sheng Chang Ze Group Co., Ltd. is higher than Lang Ji Yun Hui Technology Co., Ltd.)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy022.json b/assets/qa_gold/enterprise_industry_analysis/easy022.json
new file mode 100644
index 0000000000000000000000000000000000000000..52df318a2b535a917955883e2c1e7c044fcc1118
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy022.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy022",
+  "question": "In 2022, is the annual number of Chinese patent applications of Pu Ge Jian Chen Pharmaceutical Co., Ltd. higher than that of Jin Hu Real Estate Construction Development Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Pu Ge Jian Chen Pharmaceutical Co., Ltd. annual Chinese patent applications in 2022 = 11",
+    "Extracted from company_operation_status.csv: Jin Hu Real Estate Construction Development Co., Ltd. annual Chinese patent applications in 2022 = 6",
+    "Compared 11 and 6; since 11 > 6, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Pu Ge Jian Chen Pharmaceutical Co., Ltd. annual Chinese patent applications in 2022": 11,
+    "Jin Hu Real Estate Construction Development Co., Ltd. annual Chinese patent applications in 2022": 6,
+    "Comparison result (whether Pu Ge Jian Chen Pharmaceutical Co., Ltd. is higher than Jin Hu Real Estate Construction Development Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy023.json b/assets/qa_gold/enterprise_industry_analysis/easy023.json
new file mode 100644
index 0000000000000000000000000000000000000000..6a0e0346be56ed79ebfeac590db21b5258415242
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy023.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy023",
+  "question": "In 2022, which is larger: the R&D investment ratio of Hua Cheng Sheng Yuan Integrated Development Co., Ltd. or the year-over-year net profit change rate of Long He Zhi Jin Real Estate Co., Ltd.?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Hua Cheng Sheng Yuan Integrated Development Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Hua Cheng Sheng Yuan Integrated Development Co., Ltd. R&D investment ratio in 2022 = 4.79 %",
+    "Extracted from company_operation_status.csv: Long He Zhi Jin Real Estate Co., Ltd. year-over-year net profit change rate in 2022 = -32.50 %",
+    "Compared 4.79 and -32.50; since 4.79 > -32.50, the larger one is \"Hua Cheng Sheng Yuan Integrated Development Co., Ltd.\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Hua Cheng Sheng Yuan Integrated Development Co., Ltd. R&D investment ratio in 2022": "4.79 %",
+    "Long He Zhi Jin Real Estate Co., Ltd. year-over-year net profit change rate in 2022": "-32.50 %",
+    "Comparison result (which is larger)": "Hua Cheng Sheng Yuan Integrated Development Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy024.json b/assets/qa_gold/enterprise_industry_analysis/easy024.json
new file mode 100644
index 0000000000000000000000000000000000000000..4d59c2ccef2ef10cd49592401ad199c142b13354
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy024.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy024",
+  "question": "In 2022, is the cumulative number of PCT patent applications of Shi Yang Zhi Guang Dian Qi Co., Ltd. lower than the corresponding value of Xu Ye Zhi Gong Technology Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: cumulative number of PCT patent applications of Shi Yang Zhi Guang Dian Qi Co., Ltd. in 2022 = 93",
+    "Extracted from company_operation_status.csv: cumulative number of PCT patent applications of Xu Ye Zhi Gong Technology Co., Ltd. in 2022 = 10",
+    "Compared 93 and 10; since 93 < 10 is false, the judgment is \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Cumulative number of PCT patent applications of Shi Yang Zhi Guang Dian Qi Co., Ltd. in 2022": 93,
+    "Cumulative number of PCT patent applications of Xu Ye Zhi Gong Technology Co., Ltd. in 2022": 10,
+    "Comparison result (whether Shi Yang Zhi Guang Dian Qi Co., Ltd. is lower than Xu Ye Zhi Gong Technology Co., Ltd.)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy025.json b/assets/qa_gold/enterprise_industry_analysis/easy025.json
new file mode 100644
index 0000000000000000000000000000000000000000..eda9d313a1f0a5a0d21ae3c9119abf9b0c0e4573
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy025.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy025",
+  "question": "In 2022, is the government award funding or subsidy of Shi Yang Zhi Guang Dian Qi Co., Ltd. lower than that of Xu Ye Zhi Gong Technology Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: government award funding or subsidy of Shi Yang Zhi Guang Dian Qi Co., Ltd. in 2022 = 14071208.03 CNY",
+    "Extracted from company_operation_status.csv: government award funding or subsidy of Xu Ye Zhi Gong Technology Co., Ltd. in 2022 = 100476261.0 CNY",
+    "Compared 14071208.03 and 100476261.0; since 14071208.03 < 100476261.0, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Government award funding or subsidy of Shi Yang Zhi Guang Dian Qi Co., Ltd. in 2022 (CNY)": 14071208.03,
+    "Government award funding or subsidy of Xu Ye Zhi Gong Technology Co., Ltd. in 2022 (CNY)": 100476261.0,
+    "Comparison result (whether Shi Yang Zhi Guang Dian Qi Co., Ltd. is lower than Xu Ye Zhi Gong Technology Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy026.json b/assets/qa_gold/enterprise_industry_analysis/easy026.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a48f389514a03f48151724247e8f88d981657a5
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy026.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy026",
+  "question": "In 2022, is the total number of employees of Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. higher than that of Shen Zhou Wu Jin Zi Xun Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: total number of employees of Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. in 2022 = 1066.0",
+    "Extracted from company_operation_status.csv: total number of employees of Shen Zhou Wu Jin Zi Xun Co., Ltd. in 2022 = 133.0",
+    "Compared 1066.0 and 133.0; since 1066.0 > 133.0, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Total number of employees of Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. in 2022": 1066.0,
+    "Total number of employees of Shen Zhou Wu Jin Zi Xun Co., Ltd. in 2022": 133.0,
+    "Comparison result (whether Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. is higher than Shen Zhou Wu Jin Zi Xun Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy027.json b/assets/qa_gold/enterprise_industry_analysis/easy027.json
new file mode 100644
index 0000000000000000000000000000000000000000..286096b32660fb6e5979e34ba4c4624e481e1e53
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy027.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy027",
+  "question": "In 2022, is the cumulative number of invalid Chinese invention patents of Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. lower than the same indicator of Shen Zhou Wu Jin Zi Xun Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: cumulative number of invalid Chinese invention patents of Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. in 2022 = 1",
+    "Extracted from company_operation_status.csv: cumulative number of invalid Chinese invention patents of Shen Zhou Wu Jin Zi Xun Co., Ltd. in 2022 = 8",
+    "Compared 1 and 8; since 1 < 8, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Cumulative number of invalid Chinese invention patents of Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. in 2022": 1,
+    "Cumulative number of invalid Chinese invention patents of Shen Zhou Wu Jin Zi Xun Co., Ltd. in 2022": 8,
+    "Comparison result (whether Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. is lower than Shen Zhou Wu Jin Zi Xun Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy028.json b/assets/qa_gold/enterprise_industry_analysis/easy028.json
new file mode 100644
index 0000000000000000000000000000000000000000..25f7c8ff8175ec9e4c7f3b19c6042b0c81c40902
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy028.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy028",
+  "question": "In 2022, is the government reward fund or subsidy of Guang Sheng Chang Ze Group Co., Ltd. higher than that of Zhong You Zheng Yun Yun Port Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Guang Sheng Chang Ze Group Co., Ltd. government reward fund or subsidy in 2022 = 8527155.34 CNY",
+    "Extracted from company_operation_status.csv: Zhong You Zheng Yun Yun Port Co., Ltd. government reward fund or subsidy in 2022 = 36491108.19 CNY",
+    "Compared 8527155.34 and 36491108.19; since 8527155.34 < 36491108.19, the judgment is \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Guang Sheng Chang Ze Group Co., Ltd. government reward fund or subsidy in 2022 (CNY)": 8527155.34,
+    "Zhong You Zheng Yun Yun Port Co., Ltd. government reward fund or subsidy in 2022 (CNY)": 36491108.19,
+    "Comparison result (whether Guang Sheng Chang Ze Group Co., Ltd. is higher than Zhong You Zheng Yun Yun Port Co., Ltd.)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy029.json b/assets/qa_gold/enterprise_industry_analysis/easy029.json
new file mode 100644
index 0000000000000000000000000000000000000000..1bc02b35cfbb2c8aae4f8f1d25d7e8222a98605d
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy029.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy029",
+  "question": "In 2022, is the year-over-year change rate of R&D investment of Yong Feng Ke Lian Software Co., Ltd. higher than that of Hai Li Xuan Yue Electric Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Yong Feng Ke Lian Software Co., Ltd. year-over-year change rate of R&D investment in 2022 = -27.60 %",
+    "Extracted from company_operation_status.csv: Hai Li Xuan Yue Electric Co., Ltd. year-over-year change rate of R&D investment in 2022 = -32.38 %",
+    "Compared -27.60 and -32.38; since -27.60 > -32.38, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Yong Feng Ke Lian Software Co., Ltd. year-over-year change rate of R&D investment in 2022": "-27.60 %",
+    "Hai Li Xuan Yue Electric Co., Ltd. year-over-year change rate of R&D investment in 2022": "-32.38 %",
+    "Comparison result (whether Yong Feng Ke Lian Software Co., Ltd. is higher than Hai Li Xuan Yue Electric Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy030.json b/assets/qa_gold/enterprise_industry_analysis/easy030.json
new file mode 100644
index 0000000000000000000000000000000000000000..c31fecbf17d5eed05807d250e4a6e00cae29c29c
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy030.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy030",
+  "question": "In 2022, is the total assets of Yong Feng Ke Lian Software Co., Ltd. higher than that of Hai Li Xuan Yue Electric Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Yong Feng Ke Lian Software Co., Ltd. total assets in 2022 = 17094624520.40 CNY",
+    "Extracted from company_operation_status.csv: Hai Li Xuan Yue Electric Co., Ltd. total assets in 2022 = 2678955419.08 CNY",
+    "Compared 17094624520.40 and 2678955419.08; since 17094624520.40 > 2678955419.08, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Yong Feng Ke Lian Software Co., Ltd. total assets in 2022 (CNY)": 17094624520.4,
+    "Hai Li Xuan Yue Electric Co., Ltd. total assets in 2022 (CNY)": 2678955419.08,
+    "Comparison result (whether Yong Feng Ke Lian Software Co., Ltd. is higher than Hai Li Xuan Yue Electric Co., Ltd.)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy031.json b/assets/qa_gold/enterprise_industry_analysis/easy031.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c7c24fa8b7f899fbd5b40e6928727631f703878
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy031.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy031",
+  "question": "In 2022, which is higher: the operating revenue amount of Heng Yi Run Heng Technology Co., Ltd. or the total assets of Lian Ji Ji Jin Ji Chuang Co., Ltd.?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Lian Ji Ji Jin Ji Chuang Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Heng Yi Run Heng Technology Co., Ltd. operating revenue amount in 2022 = 4259181531.38 CNY",
+    "Extracted from company_operation_status.csv: Lian Ji Ji Jin Ji Chuang Co., Ltd. total assets in 2022 = 6102473889.29 CNY",
+    "Compared 4259181531.38 and 6102473889.29; since 6102473889.29 > 4259181531.38, the higher one is \"Lian Ji Ji Jin Ji Chuang Co., Ltd.\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Heng Yi Run Heng Technology Co., Ltd. operating revenue amount in 2022 (CNY)": 4259181531.38,
+    "Lian Ji Ji Jin Ji Chuang Co., Ltd. total assets in 2022 (CNY)": 6102473889.29,
+    "Comparison result (which is higher)": "Lian Ji Ji Jin Ji Chuang Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy032.json b/assets/qa_gold/enterprise_industry_analysis/easy032.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e69d7ec70fd7d3d790e0692a6a88d20cad903d5
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy032.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy032",
+  "question": "In 2022, is the number of R&D personnel of Heng Yi Run Heng Technology Co., Ltd. lower than the annual number of Chinese patent applications of Lian Ji Ji Jin Ji Chuang Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Heng Yi Run Heng Technology Co., Ltd. number of R&D personnel in 2022 = 257.0",
+    "Extracted from company_operation_status.csv: Lian Ji Ji Jin Ji Chuang Co., Ltd. annual Chinese patent applications in 2022 = 75",
+    "Compared 257.0 and 75; since 257.0 < 75 is false, the judgment is \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Heng Yi Run Heng Technology Co., Ltd. number of R&D personnel in 2022": 257.0,
+    "Lian Ji Ji Jin Ji Chuang Co., Ltd. annual Chinese patent applications in 2022": 75,
+    "Comparison result (whether Heng Yi Run Heng Technology Co., Ltd. is lower than Lian Ji Ji Jin Ji Chuang Co., Ltd.)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy033.json b/assets/qa_gold/enterprise_industry_analysis/easy033.json
new file mode 100644
index 0000000000000000000000000000000000000000..da894aaf7f1cead313fa3607465026c65dd78819
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy033.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy033",
+  "question": "In 2022, compared with the total liabilities of Wei Xing Run Jin Ke Ji Co., Ltd., what is the difference in total assets of Ping Ru Gang Tong Yun Wu Liu Co., Ltd.?",
+  "guidelines": "The answer must be a number with two decimal places. Output only the number without units, commas, or any text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 37950292402.87,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: total assets of Ping Ru Gang Tong Yun Wu Liu Co., Ltd. in 2022 = 78458358656.04 CNY",
+    "Extracted from company_operation_status.csv: total liabilities of Wei Xing Run Jin Ke Ji Co., Ltd. in 2022 = 40508066253.17 CNY",
+    "Calculated the difference: 78458358656.04 - 40508066253.17 = 37950292402.87"
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Total assets of Ping Ru Gang Tong Yun Wu Liu Co., Ltd. in 2022 (CNY)": 78458358656.04,
+    "Total liabilities of Wei Xing Run Jin Ke Ji Co., Ltd. in 2022 (CNY)": 40508066253.17,
+    "Difference (Ping Ru Gang Tong Yun Wu Liu Co., Ltd. - Wei Xing Run Jin Ke Ji Co., Ltd.)": 37950292402.87
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy034.json b/assets/qa_gold/enterprise_industry_analysis/easy034.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb6b6f69dc2b2a198c0cee151ef194871dad8ecd
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy034.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy034",
+  "question": "In 2022, between the year-over-year net profit growth rate of Ping Ru Gang Tong Yun Wu Liu Co., Ltd. and the net profit amount of Wei Xing Run Jin Ke Ji Co., Ltd., which value is larger?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Wei Xing Run Jin Ke Ji Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: year-over-year net profit growth rate of Ping Ru Gang Tong Yun Wu Liu Co., Ltd. in 2022 = -9.38 %",
+    "Extracted from company_operation_status.csv: net profit amount of Wei Xing Run Jin Ke Ji Co., Ltd. in 2022 = 388147978.44 CNY",
+    "Compared -9.38 and 388147978.44 as numeric values; since 388147978.44 > -9.38, the larger one is \"Wei Xing Run Jin Ke Ji Co., Ltd.\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Year-over-year net profit growth rate of Ping Ru Gang Tong Yun Wu Liu Co., Ltd. in 2022": "-9.38 %",
+    "Net profit amount of Wei Xing Run Jin Ke Ji Co., Ltd. in 2022 (CNY)": 388147978.44,
+    "Comparison result (which value is larger)": "Wei Xing Run Jin Ke Ji Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy035.json b/assets/qa_gold/enterprise_industry_analysis/easy035.json
new file mode 100644
index 0000000000000000000000000000000000000000..579beaa26dbc7d927e6bfde773fcd9110a888bad
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy035.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy035",
+  "question": "In 2022, are the industries of Gao Yin Ze Tong Pi Fa Co., Ltd. and Yong Hui Ze Hui Pi Fa Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Gao Yin Ze Tong Pi Fa Co., Ltd. = Wholesale and Retail",
+    "Extracted from company_profile.csv: industry of Yong Hui Ze Hui Pi Fa Co., Ltd. = Wholesale and Retail",
+    "Compared the industry text of both companies; they are the same, so the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Gao Yin Ze Tong Pi Fa Co., Ltd.": "Wholesale and Retail",
+    "Industry of Yong Hui Ze Hui Pi Fa Co., Ltd.": "Wholesale and Retail",
+    "Comparison result (whether the industries are the same)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy036.json b/assets/qa_gold/enterprise_industry_analysis/easy036.json
new file mode 100644
index 0000000000000000000000000000000000000000..91cb07884aaed827d4de845d942db44a15e597e8
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy036.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy036",
+  "question": "Compared with the listing date of Yong Hui Ze Hui Pi Fa Co., Ltd., which listing date is earlier for Gao Yin Ze Tong Pi Fa Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Gao Yin Ze Tong Pi Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: listing date of Gao Yin Ze Tong Pi Fa Co., Ltd. = 1992-08-17",
+    "Extracted from company_profile.csv: listing date of Yong Hui Ze Hui Pi Fa Co., Ltd. = 1993-02-02",
+    "Compared the dates; 1992-08-17 is earlier than 1993-02-02, so the earlier one is \"Gao Yin Ze Tong Pi Fa Co., Ltd.\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Listing date of Gao Yin Ze Tong Pi Fa Co., Ltd.": "1992-08-17",
+    "Listing date of Yong Hui Ze Hui Pi Fa Co., Ltd.": "1993-02-02",
+    "Comparison result (which one is earlier)": "Gao Yin Ze Tong Pi Fa Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy037.json b/assets/qa_gold/enterprise_industry_analysis/easy037.json
new file mode 100644
index 0000000000000000000000000000000000000000..73aa42f067fdc5aef2402e3e4d8b5921915f1fd0
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy037.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy037",
+  "question": "In 2022, are the industries of Ma Gang Tai Jin Cai Liao Co., Ltd. and Ma Gang Gang Sheng Bu Xiu Gang Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Ma Gang Tai Jin Cai Liao Co., Ltd. = Metal Products",
+    "Extracted from company_profile.csv: industry of Ma Gang Gang Sheng Bu Xiu Gang Co., Ltd. = Metal Products",
+    "Compared the industry text of both companies; they are the same, so the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Ma Gang Tai Jin Cai Liao Co., Ltd.": "Metal Products",
+    "Industry of Ma Gang Gang Sheng Bu Xiu Gang Co., Ltd.": "Metal Products",
+    "Comparison result (whether the industries are the same)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy038.json b/assets/qa_gold/enterprise_industry_analysis/easy038.json
new file mode 100644
index 0000000000000000000000000000000000000000..215bec6f970cc4654f965b96d0c7948b102395f2
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy038.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy038",
+  "question": "In 2022, is the enterprise type of Magang Taijin Materials Co., Ltd. the same as that of Magang Gangsheng Stainless Steel Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: enterprise type of Magang Taijin Materials Co., Ltd. = Shanghai and Shenzhen",
+    "Extracted from company_profile.csv: enterprise type of Magang Gangsheng Stainless Steel Co., Ltd. = Shanghai and Shenzhen",
+    "Compared the two enterprise type texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Enterprise type of Magang Taijin Materials Co., Ltd.": "Shanghai and Shenzhen",
+    "Enterprise type of Magang Gangsheng Stainless Steel Co., Ltd.": "Shanghai and Shenzhen",
+    "Comparison result (whether the enterprise types are the same)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy039.json b/assets/qa_gold/enterprise_industry_analysis/easy039.json
new file mode 100644
index 0000000000000000000000000000000000000000..771619b834b92b2d66bfdcffc78dee454ee8c248
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy039.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy039",
+  "question": "In 2022, is the industry of Chuangwei Yaosheng Electric Co., Ltd. the same as that of Lixin Zhichuang Home Appliances Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Chuangwei Yaosheng Electric Co., Ltd. = Consumer Electronics and Electrical Industry",
+    "Extracted from company_profile.csv: industry of Lixin Zhichuang Home Appliances Co., Ltd. = Consumer Electronics and Electrical Industry",
+    "Compared the two industry texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Chuangwei Yaosheng Electric Co., Ltd.": "Consumer Electronics and Electrical Industry",
+    "Industry of Lixin Zhichuang Home Appliances Co., Ltd.": "Consumer Electronics and Electrical Industry",
+    "Comparison result (whether the industries are the same)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy040.json b/assets/qa_gold/enterprise_industry_analysis/easy040.json
new file mode 100644
index 0000000000000000000000000000000000000000..4b209ba5a17cf82e6adffd3c0466d8ea5eaddd03
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy040.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy040",
+  "question": "In 2022, is the stock exchange of Chuangwei Yaosheng Electric Co., Ltd. the same as that of Lixin Zhichuang Home Appliances Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: stock exchange of Chuangwei Yaosheng Electric Co., Ltd. = Shenzhen Stock Exchange",
+    "Extracted from company_profile.csv: stock exchange of Lixin Zhichuang Home Appliances Co., Ltd. = Shenzhen Stock Exchange",
+    "Compared the two stock exchange texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Stock exchange of Chuangwei Yaosheng Electric Co., Ltd.": "Shenzhen Stock Exchange",
+    "Stock exchange of Lixin Zhichuang Home Appliances Co., Ltd.": "Shenzhen Stock Exchange",
+    "Comparison result (whether the stock exchanges are the same)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy041.json b/assets/qa_gold/enterprise_industry_analysis/easy041.json
new file mode 100644
index 0000000000000000000000000000000000000000..b3cadee8ed4bf332742f2cf722ff3b54a82c34c6
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy041.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy041",
+  "question": "Comparing the incorporation dates of Biyuan Chanhua Real Estate Holdings Co., Ltd. and Huarun Zhijin Construction Development Co., Ltd., which one was established earlier?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Biyuan Chanhua Real Estate Holdings Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: incorporation date of Biyuan Chanhua Real Estate Holdings Co., Ltd. = 1989-11-22",
+    "Extracted from company_profile.csv: incorporation date of Huarun Zhijin Construction Development Co., Ltd. = 1991-03-30",
+    "Compared dates 1989-11-22 and 1991-03-30; since 1989-11-22 is earlier, the judgment is \"Biyuan Chanhua Real Estate Holdings Co., Ltd.\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Incorporation date of Biyuan Chanhua Real Estate Holdings Co., Ltd.": "1989-11-22",
+    "Incorporation date of Huarun Zhijin Construction Development Co., Ltd.": "1991-03-30",
+    "Comparison result (which one is earlier)": "Biyuan Chanhua Real Estate Holdings Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy042.json b/assets/qa_gold/enterprise_industry_analysis/easy042.json
new file mode 100644
index 0000000000000000000000000000000000000000..ef4f2fbc17120b8099e6a57f4b823a6f481c9170
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy042.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy042",
+  "question": "In 2022, are the industries of Biyuan Chanhua Real Estate Holdings Co., Ltd. and Huarun Zhijin Construction Development Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Biyuan Chanhua Real Estate Holdings Co., Ltd. = Real Estate",
+    "Extracted from company_profile.csv: industry of Huarun Zhijin Construction Development Co., Ltd. = Real Estate",
+    "Compared the two industry texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Biyuan Chanhua Real Estate Holdings Co., Ltd.": "Real Estate",
+    "Industry of Huarun Zhijin Construction Development Co., Ltd.": "Real Estate",
+    "Comparison result (whether the industries are the same)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy043.json b/assets/qa_gold/enterprise_industry_analysis/easy043.json
new file mode 100644
index 0000000000000000000000000000000000000000..07bb949087126160c4389136a9586c292274ceaf
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy043.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy043",
+  "question": "Comparing the incorporation dates of Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd. and Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd., which one was established earlier?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: incorporation date of Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd. = 1994-04-20",
+    "Extracted from company_profile.csv: incorporation date of Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd. = 2004-10-09",
+    "Compared dates 1994-04-20 and 2004-10-09; since 1994-04-20 is earlier, the judgment is \"Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd.\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Incorporation date of Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd.": "1994-04-20",
+    "Incorporation date of Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd.": "2004-10-09",
+    "Comparison result (which one is earlier)": "Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy044.json b/assets/qa_gold/enterprise_industry_analysis/easy044.json
new file mode 100644
index 0000000000000000000000000000000000000000..ade9a736dd34c510913c50120fc0ad66a3eed85a
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy044.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy044",
+  "question": "In 2022, are the stock exchanges of Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd. and Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: stock exchange of Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd. = Shenzhen Stock Exchange",
+    "Extracted from company_profile.csv: stock exchange of Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd. = Shenzhen Stock Exchange",
+    "Compared the two stock exchange texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Stock exchange of Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd.": "Shenzhen Stock Exchange",
+    "Stock exchange of Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd.": "Shenzhen Stock Exchange",
+    "Comparison result (whether the stock exchanges are the same)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy045.json b/assets/qa_gold/enterprise_industry_analysis/easy045.json
new file mode 100644
index 0000000000000000000000000000000000000000..41c7399b86d74633781c45a120a8572b9bc4322d
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy045.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy045",
+  "question": "In 2022, is the industry of Chuang Wei Jin Zhi Electrical Appliances Co., Ltd. the same as that of Long He Chan Zhi Di Chan Holdings Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Chuang Wei Jin Zhi Electrical Appliances Co., Ltd. = Consumer Electronics and Electrical Industry",
+    "Extracted from company_profile.csv: industry of Long He Chan Zhi Di Chan Holdings Co., Ltd. = Real Estate",
+    "Compared the two industry texts and judged \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Chuang Wei Jin Zhi Electrical Appliances Co., Ltd.": "Consumer Electronics and Electrical Industry",
+    "Industry of Long He Chan Zhi Di Chan Holdings Co., Ltd.": "Real Estate",
+    "Comparison result (whether the industries are the same)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy046.json b/assets/qa_gold/enterprise_industry_analysis/easy046.json
new file mode 100644
index 0000000000000000000000000000000000000000..48ea84d0f57a549c65624a65dcb75817c8f9333a
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy046.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy046",
+  "question": "In 2022, are the listing boards of Chuang Wei Jin Zhi Electrical Appliances Co., Ltd. and Long He Chan Zhi Di Chan Holdings Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: listing board of Chuang Wei Jin Zhi Electrical Appliances Co., Ltd. = Main Board",
+    "Extracted from company_profile.csv: listing board of Long He Chan Zhi Di Chan Holdings Co., Ltd. = Main Board",
+    "Compared the two listing board texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Listing board of Chuang Wei Jin Zhi Electrical Appliances Co., Ltd.": "Main Board",
+    "Listing board of Long He Chan Zhi Di Chan Holdings Co., Ltd.": "Main Board",
+    "Comparison result (whether the listing boards are the same)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy047.json b/assets/qa_gold/enterprise_industry_analysis/easy047.json
new file mode 100644
index 0000000000000000000000000000000000000000..f80d824af9a4f68a36ae8b9f45d39ee45a49ad8e
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy047.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy047",
+  "question": "In 2022, are the industries of Hua Ying Tai Sheng Wealth Management Co., Ltd. and Yong Feng Lian Chuang Xi Tong Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Hua Ying Tai Sheng Wealth Management Co., Ltd. = Financial Industry",
+    "Extracted from company_profile.csv: industry of Yong Feng Lian Chuang Xi Tong Co., Ltd. = Information Transmission, Software and IT Services",
+    "Compared the two industry texts and judged \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Hua Ying Tai Sheng Wealth Management Co., Ltd.": "Financial Industry",
+    "Industry of Yong Feng Lian Chuang Xi Tong Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Comparison result (whether the industries are the same)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy048.json b/assets/qa_gold/enterprise_industry_analysis/easy048.json
new file mode 100644
index 0000000000000000000000000000000000000000..c6eb8ca362d15bd437606c16f612898adf492193
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy048.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy048",
+  "question": "In 2022, is the stock exchange of Huaying Taisheng Wealth Management Co., Ltd. the same as that of Yongfeng Lianchuang Systems Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: stock exchange of Huaying Taisheng Wealth Management Co., Ltd. = Shenzhen Stock Exchange",
+    "Extracted from company_profile.csv: stock exchange of Yongfeng Lianchuang Systems Co., Ltd. = Shenzhen Stock Exchange",
+    "Compared the two stock exchange texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Stock exchange of Huaying Taisheng Wealth Management Co., Ltd.": "Shenzhen Stock Exchange",
+    "Stock exchange of Yongfeng Lianchuang Systems Co., Ltd.": "Shenzhen Stock Exchange",
+    "Comparison result (whether the stock exchanges are the same)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy049.json b/assets/qa_gold/enterprise_industry_analysis/easy049.json
new file mode 100644
index 0000000000000000000000000000000000000000..ee2f4ec9039aef6af5119b11dbd5af11273b3152
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy049.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy049",
+  "question": "In 2022, is the board segment of Zhongbai Damao Wholesale Co., Ltd. the same as that of Luxi Runheng Chemical Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: board segment of Zhongbai Damao Wholesale Co., Ltd. = Main Board",
+    "Extracted from company_profile.csv: board segment of Luxi Runheng Chemical Co., Ltd. = Main Board",
+    "Compared the two board segment texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Board segment of Zhongbai Damao Wholesale Co., Ltd.": "Main Board",
+    "Board segment of Luxi Runheng Chemical Co., Ltd.": "Main Board",
+    "Comparison result (whether the board segments are the same)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy050.json b/assets/qa_gold/enterprise_industry_analysis/easy050.json
new file mode 100644
index 0000000000000000000000000000000000000000..dbd8e4220897a35dd94693221f5022e9fb17769d
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy050.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy050",
+  "question": "In 2022, is the ownership type of Zhongbai Damao Wholesale Co., Ltd. the same as that of Luxi Runheng Chemical Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: ownership type of Zhongbai Damao Wholesale Co., Ltd. = Private Enterprise",
+    "Extracted from company_profile.csv: ownership type of Luxi Runheng Chemical Co., Ltd. = Private Enterprise",
+    "Compared the two ownership type texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Ownership type of Zhongbai Damao Wholesale Co., Ltd.": "Private Enterprise",
+    "Ownership type of Luxi Runheng Chemical Co., Ltd.": "Private Enterprise",
+    "Comparison result (whether the ownership types are the same)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy051.json b/assets/qa_gold/enterprise_industry_analysis/easy051.json
new file mode 100644
index 0000000000000000000000000000000000000000..e7a2a5e171daa02461662b02b58aabd88f618aeb
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy051.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy051",
+  "question": "In 2022, is the enterprise type of Meineng Xuanyue Electric Co., Ltd. the same as that of Baotie Yuanchang Metal Products Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: enterprise type of Meineng Xuanyue Electric Co., Ltd. = Shanghai and Shenzhen",
+    "Extracted from company_profile.csv: enterprise type of Baotie Yuanchang Metal Products Co., Ltd. = Hong Kong Stocks",
+    "Compared the two enterprise type texts and judged \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Enterprise type of Meineng Xuanyue Electric Co., Ltd.": "Shanghai and Shenzhen",
+    "Enterprise type of Baotie Yuanchang Metal Products Co., Ltd.": "Hong Kong Stocks",
+    "Comparison result (whether the enterprise types are the same)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy052.json b/assets/qa_gold/enterprise_industry_analysis/easy052.json
new file mode 100644
index 0000000000000000000000000000000000000000..ad94790b92299dd3720faca74565ab1366b4e0d8
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy052.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy052",
+  "question": "In 2022, are the stock exchanges of Mei Neng Xuan Yue Dian Qi Co., Ltd. and Bao Tie Yuan Chang Jin Shu Zhi Pin Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: stock exchange of Mei Neng Xuan Yue Dian Qi Co., Ltd. = Shenzhen Stock Exchange",
+    "Extracted from company_profile.csv: stock exchange of Bao Tie Yuan Chang Jin Shu Zhi Pin Co., Ltd. = Gang Jiao Suo",
+    "Compared the two stock exchange texts and judged \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Stock exchange of Mei Neng Xuan Yue Dian Qi Co., Ltd.": "Shenzhen Stock Exchange",
+    "Stock exchange of Bao Tie Yuan Chang Jin Shu Zhi Pin Co., Ltd.": "Gang Jiao Suo",
+    "Comparison result (whether the stock exchanges are the same)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy053.json b/assets/qa_gold/enterprise_industry_analysis/easy053.json
new file mode 100644
index 0000000000000000000000000000000000000000..819ce9fbb08c2012c6b21b15317186e9c2b52a66
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy053.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy053",
+  "question": "In 2022, does the secondary industry of Bao He Hua Chang Jian She Kai Fa Co., Ltd. belong to the same industry as that of Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: secondary industry of Bao He Hua Chang Jian She Kai Fa Co., Ltd. = Residential Real Estate",
+    "Extracted from company_profile.csv: secondary industry of Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd. = Retail",
+    "Compared the two secondary industry texts and judged \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Secondary industry of Bao He Hua Chang Jian She Kai Fa Co., Ltd.": "Residential Real Estate",
+    "Secondary industry of Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd.": "Retail",
+    "Comparison result (whether they belong to the same industry)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy054.json b/assets/qa_gold/enterprise_industry_analysis/easy054.json
new file mode 100644
index 0000000000000000000000000000000000000000..3abdc119c913fd140cc35e818049fdf6eb3de79f
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy054.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy054",
+  "question": "In 2022, are the industries of Bao He Hua Chang Jian She Kai Fa Co., Ltd. and Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Bao He Hua Chang Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from company_profile.csv: industry of Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd. = Wholesale and Retail",
+    "Compared the two industry texts and judged \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Bao He Hua Chang Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "Industry of Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd.": "Wholesale and Retail",
+    "Comparison result (whether the industries are the same)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy055.json b/assets/qa_gold/enterprise_industry_analysis/easy055.json
new file mode 100644
index 0000000000000000000000000000000000000000..02c17bbb4fe19794e1b953a10c00957b5b4c1012
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy055.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy055",
+  "question": "In 2022, compared with Chuang Xin Yao Rui Integrated Circuit Co., Ltd., did the core competitiveness of Ya Wei Ze Zhi Technology Co., Ltd. also emphasize technological innovation and high-quality customer resources?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted the core competitiveness description of Ya Wei Ze Zhi Technology Co., Ltd. from company_core.csv and identified statements about \"technological innovation capability\" and \"high-quality customer resources\"",
+    "Extracted the core competitiveness description of Chuang Xin Yao Rui Integrated Circuit Co., Ltd. from company_core.csv and identified statements about \"R&D innovation advantages\" and \"high-quality customer and brand advantages (customer resources)\"",
+    "Determined that both companies emphasize technological innovation and high-quality customer resources, and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Core competitiveness keywords of Ya Wei Ze Zhi Technology Co., Ltd.": "Technological innovation capability, high-quality customer resources",
+    "Core competitiveness keywords of Chuang Xin Yao Rui Integrated Circuit Co., Ltd.": "R&D innovation advantages, high-quality customer and brand advantages",
+    "Comparison result (whether both emphasize technological innovation and high-quality customer resources)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy056.json b/assets/qa_gold/enterprise_industry_analysis/easy056.json
new file mode 100644
index 0000000000000000000000000000000000000000..1c300c7a35af60153b148093fb9214218cbba073
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy056.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy056",
+  "question": "In 2022, did the core competitiveness of Zhong Ke Sheng Ke Ji Shu Yan Jiu Yuan Co., Ltd. and that of Lian Ji Zhi Sheng Ji Xie Co., Ltd. in the same province show a competitive relationship in terms of technical capabilities?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted the core competitiveness description of Zhong Ke Sheng Ke Ji Shu Yan Jiu Yuan Co., Ltd. from company_core.csv and identified that its technical capability focuses on biopharmaceutical R&D and production technology (such as recombinant proteins and mRNA raw-material enzymes)",
+    "Extracted the core competitiveness description of Lian Ji Zhi Sheng Ji Xie Co., Ltd. from company_core.csv and identified that its technical capability focuses on manufacturing R&D for fastening tools (such as gas nail guns) and construction hardware",
+    "Determined that the two companies have different technical directions and the descriptions provide no direct evidence of competition, and judged \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Technical focus of Zhong Ke Sheng Ke Ji Shu Yan Jiu Yuan Co., Ltd.": "Biopharmaceutical-related R&D and production technology platform",
+    "Technical focus of Lian Ji Zhi Sheng Ji Xie Co., Ltd.": "R&D and manufacturing of fastening tools and construction hardware",
+    "Comparison result (whether there is a competitive relationship in technical capabilities)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy057.json b/assets/qa_gold/enterprise_industry_analysis/easy057.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea67c47fbecb8cb2ddca4889119e30ac7b10e034
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy057.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy057",
+  "question": "In 2022, did the products of He Lian Chuang Hang She Bei Co., Ltd. and Shan La Da Chuang Zhi Neng Zhuang Bei Co., Ltd. have a competitive relationship?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted the core competitiveness description of He Lian Chuang Hang She Bei Co., Ltd. from company_core.csv and identified its main product focus as medical rehabilitation devices such as pelvic-floor and obstetric rehabilitation",
+    "Extracted the core competitiveness description of Shan La Da Chuang Zhi Neng Zhuang Bei Co., Ltd. from company_core.csv and identified its main product/business focus as precision mold manufacturing such as lithium-battery equipment cutting dies (with other business segments)",
+    "Determined that the two companies have different product directions and the descriptions provide no direct evidence of direct competition in similar products, and judged \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Product focus of He Lian Chuang Hang She Bei Co., Ltd.": "Medical rehabilitation devices and care ecosystem",
+    "Product focus of Shan La Da Chuang Zhi Neng Zhuang Bei Co., Ltd.": "Precision mold manufacturing such as lithium-battery equipment cutting dies",
+    "Comparison result (whether there is a competitive relationship between products)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy058.json b/assets/qa_gold/enterprise_industry_analysis/easy058.json
new file mode 100644
index 0000000000000000000000000000000000000000..f49b7f6659f2176f093695b4e80e07ea6e1471f1
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy058.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy058",
+  "question": "In 2022, compared with the products of Jiejie Dahang Equipment Co., Ltd., are the products of Sansan Gongzhi Technology Co., Ltd. in a competitive relationship?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_core.csv the core competitiveness description of Sansan Gongzhi Technology Co., Ltd. and identified its product focus as medical diagnostic and monitoring devices (blood oxygen, ECG, ultrasound, etc.).",
+    "Extracted from company_core.csv the core competitiveness description of Jiejie Dahang Equipment Co., Ltd. and identified its product focus as precision automation equipment for the electronics industry (printing, dispensing, die-bonding, etc.).",
+    "Judged that the two product directions are different, and no direct evidence of head-to-head competition in similar products is provided in the descriptions, so the result is \"No\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Product focus of Sansan Gongzhi Technology Co., Ltd.": "Medical diagnostic and monitoring devices",
+    "Product focus of Jiejie Dahang Equipment Co., Ltd.": "Precision automation equipment for the electronics industry",
+    "Comparison result (whether there is a competitive relationship between the products)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy059.json b/assets/qa_gold/enterprise_industry_analysis/easy059.json
new file mode 100644
index 0000000000000000000000000000000000000000..e3da98b2a4058a1f911213ad36ef3a7ef8ececca
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy059.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy059",
+  "question": "In 2022, is Haomei Company's core competitiveness in R&D and technology more focused on innovation and diversification than that of Zhongjin Yeye Resources Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_core.csv the core competitiveness description of Haomei Company and identified statements on R&D platform construction, technological innovation, and multi-field application layout (diversification).",
+    "Extracted from company_core.csv the core competitiveness description of Zhongjin Yeye Resources Co., Ltd. and checked whether it takes R&D innovation and diversification as its core expression.",
+    "Compared the two companies' focus on \"innovation and diversification\" and judged that Haomei Company is more focused, so the output is \"Yes\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "R&D and technology focus of Haomei Company": "Innovation and diversified application layout",
+    "R&D and technology focus of Zhongjin Yeye Resources Co., Ltd.": "Region/quality/workforce and management systems (not centered on innovation and diversification)",
+    "Comparison result (whether Haomei Company is more focused on innovation and diversification)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy060.json b/assets/qa_gold/enterprise_industry_analysis/easy060.json
new file mode 100644
index 0000000000000000000000000000000000000000..4cb541fafc2d02720cc9b22483de018f6d8b1d77
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy060.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy060",
+  "question": "In 2022, compared with Sansong Shijin Condiment Co., Ltd.'s strengths in technology development and independent innovation, whose strengths are more focused on market influence: Haishan Weixiang Catering Management Co., Ltd. or Sansong Shijin Condiment Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Haishan Weixiang Catering Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_core.csv the core competitiveness description of Haishan Weixiang Catering Management Co., Ltd. and identified market influence expressions such as \"brand influence/honors/awareness\".",
+    "Extracted from company_core.csv the core competitiveness description of Sansong Shijin Condiment Co., Ltd. and identified that it mainly emphasizes \"technology development and independent innovation/R&D investment/technology platform\".",
+    "Compared the focus of both companies and judged that the one more focused on market influence is \"Haishan Weixiang Catering Management Co., Ltd.\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Strength focus of Haishan Weixiang Catering Management Co., Ltd.": "Brand influence/awareness/honors (market influence)",
+    "Strength focus of Sansong Shijin Condiment Co., Ltd.": "Technology development and independent innovation (technical capability)",
+    "Comparison result (which one is more focused on market influence)": "Haishan Weixiang Catering Management Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy061.json b/assets/qa_gold/enterprise_industry_analysis/easy061.json
new file mode 100644
index 0000000000000000000000000000000000000000..5526eb225582101c2cff74ad7f88cf120b907f0e
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy061.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy061",
+  "question": "In 2022, comparing the R&D and technology of Meineng Dianguang Home Appliances Co., Ltd. with the technological innovation of Lixin Shengyue Intelligent Technology Co., Ltd., which company is more competitive in the connector field?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Meineng Dianguang Home Appliances Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_core.csv the core competitiveness description of Meineng Dianguang Home Appliances Co., Ltd. and confirmed its expressions on R&D, mold, and product capabilities in the connector field.",
+    "Extracted from company_core.csv the core competitiveness description of Lixin Shengyue Intelligent Technology Co., Ltd. and confirmed that its main technology and product focus is on piezoelectric quartz crystal components rather than connectors.",
+    "Compared relevance and competitive focus around the \"connector field\" and judged that \"Meineng Dianguang Home Appliances Co., Ltd.\" is more competitive."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Technology/product focus of Meineng Dianguang Home Appliances Co., Ltd.": "Connector R&D, precision molds, and connector customization",
+    "Technology/product focus of Lixin Shengyue Intelligent Technology Co., Ltd.": "Piezoelectric quartz crystal components and process mass-production capability",
+    "Comparison result (more competitive in the connector field)": "Meineng Dianguang Home Appliances Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy062.json b/assets/qa_gold/enterprise_industry_analysis/easy062.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f703a63af49c38e5469648df676ab33c43a0e2c
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy062.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy062",
+  "question": "In 2022, what is the difference between the number of SSE-listed enterprises in the education industry in Beijing and the number of SSE-listed central state-owned enterprises in the national consumer electronics and electrical industry?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number without units, commas, or any explanatory text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the number of SSE-listed enterprises for Beijing in the education industry is 2.",
+    "Extracted from national_industry_status.csv that the number of SSE-listed central state-owned enterprises in the consumer electronics and electrical industry is 4.",
+    "Calculated the difference: 2 - 4 = -2.0."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Number of SSE-listed enterprises in Beijing's education industry": 2,
+    "Number of SSE-listed central state-owned enterprises in the national consumer electronics and electrical industry": 4,
+    "Difference (Beijing education SSE-listed count - national consumer electronics central SOE SSE-listed count)": -2.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy063.json b/assets/qa_gold/enterprise_industry_analysis/easy063.json
new file mode 100644
index 0000000000000000000000000000000000000000..4d3152e9c0319312f5f64da9de4abb3eb7a3ab3a
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy063.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy063",
+  "question": "In 2022, which is lower: the average year-on-year net profit growth rate of Beijing's furniture manufacturing industry or that of the nationwide chemical fiber manufacturing industry?",
+  "guidelines": "The answer must be either \"Beijing Furniture Manufacturing\" or \"Nationwide Chemical Fiber Manufacturing\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Beijing Furniture Manufacturing",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: average YoY net profit growth rate of Beijing furniture manufacturing = -80.49%",
+    "Extracted from national_industry_status.csv: average YoY net profit growth rate of nationwide chemical fiber manufacturing = -12.5773529411765%",
+    "Compared -80.49 and -12.5773529411765; since -80.49 is lower, output \"Beijing Furniture Manufacturing\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Average YoY net profit growth rate of Beijing furniture manufacturing": "-80.49%",
+    "Average YoY net profit growth rate of nationwide chemical fiber manufacturing": "-12.5773529411765%",
+    "Comparison result (which is lower)": "Beijing Furniture Manufacturing"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy064.json b/assets/qa_gold/enterprise_industry_analysis/easy064.json
new file mode 100644
index 0000000000000000000000000000000000000000..e62eef18e044b700e21a156670cdaebcb153cc20
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy064.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy064",
+  "question": "In 2022, which is higher: the average year-on-year operating profit growth rate of Beijing's real estate industry or that of the nationwide information transmission, software, and IT services industry?",
+  "guidelines": "The answer must be either \"Beijing Real Estate\" or \"Nationwide Information Transmission, Software and IT Services\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Beijing Real Estate",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: average YoY operating profit growth rate of Beijing real estate = -45.385%",
+    "Extracted from national_industry_status.csv: average YoY operating profit growth rate of nationwide information transmission, software and IT services = -137.175403726708%",
+    "Compared -45.385 and -137.175403726708; since -45.385 is higher, output \"Beijing Real Estate\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Average YoY operating profit growth rate of Beijing real estate": "-45.385%",
+    "Average YoY operating profit growth rate of nationwide information transmission, software and IT services": "-137.175403726708%",
+    "Comparison result (which is higher)": "Beijing Real Estate"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy065.json b/assets/qa_gold/enterprise_industry_analysis/easy065.json
new file mode 100644
index 0000000000000000000000000000000000000000..1b9c8a03dab0593e5517083f0c233dd5d195aeb1
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy065.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy065",
+  "question": "In 2022, what is the difference between the number of Shenzhen Stock Exchange-listed local state-owned enterprises in Beijing's information transmission, software, and IT services industry and the number of Shanghai Stock Exchange-listed foreign-funded enterprises in the nationwide real estate industry?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number without units, commas, or any text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: number of SZSE-listed local state-owned enterprises in Beijing information transmission, software and IT services = 4",
+    "Extracted from national_industry_status.csv: number of SSE-listed foreign-funded enterprises in nationwide real estate = 6",
+    "Calculated difference: 4 - 6 = -2.0"
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Number of SZSE-listed local state-owned enterprises in Beijing information transmission, software and IT services": 4,
+    "Number of SSE-listed foreign-funded enterprises in nationwide real estate": 6,
+    "Difference (Beijing information services local SOE SZSE count - nationwide real estate foreign-funded SSE count)": -2.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy066.json b/assets/qa_gold/enterprise_industry_analysis/easy066.json
new file mode 100644
index 0000000000000000000000000000000000000000..986b9b23b73e76f9a0975fba7dca2446647e0551
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy066.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy066",
+  "question": "In 2022, which value is higher: the minimum cumulative number of invalidated PCT invention patents in Beijing's comprehensive industry, or the same metric in the nationwide pharmaceutical manufacturing industry?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: minimum cumulative invalidated PCT invention patents in Beijing comprehensive industry = 0",
+    "Extracted from national_industry_status.csv: minimum cumulative invalidated PCT invention patents in nationwide pharmaceutical manufacturing = 0",
+    "Compared the two values; they are equal, so output \"Equal\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Minimum cumulative invalidated PCT invention patents in Beijing comprehensive industry": 0,
+    "Minimum cumulative invalidated PCT invention patents in nationwide pharmaceutical manufacturing": 0,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy067.json b/assets/qa_gold/enterprise_industry_analysis/easy067.json
new file mode 100644
index 0000000000000000000000000000000000000000..54607046de32179e976094cb1f6b2a60dc50d720
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy067.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy067",
+  "question": "In 2022, which is larger: the number of Shenzhen Stock Exchange-listed foreign-funded enterprises in Guangdong's scientific research and technical services industry, or the number of Shanghai Stock Exchange-listed state-owned institute enterprises in the same industry nationwide?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: number of SZSE-listed foreign-funded enterprises in Guangdong scientific research and technical services = 1",
+    "Extracted from national_industry_status.csv: number of SSE-listed state-owned institute enterprises in nationwide scientific research and technical services = 1",
+    "Compared 1 and 1; they are equal, so output \"Equal\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Number of SZSE-listed foreign-funded enterprises in Guangdong scientific research and technical services": 1,
+    "Number of SSE-listed state-owned institute enterprises in nationwide scientific research and technical services": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy068.json b/assets/qa_gold/enterprise_industry_analysis/easy068.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ba128046a60cdc2c4cc8bf82dac6d0b21cfe7cb
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy068.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy068",
+  "question": "In 2022, which is higher: the average year-over-year change rate of R&D personnel in Guangdong Province's scientific research and technical services industry, or that of the same industry nationwide?",
+  "guidelines": "The answer must be \"National\" or \"Guangdong\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "National",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the average year-over-year change rate of R&D personnel for Guangdong Province in the scientific research and technical services industry is 5.80304347826087 %.",
+    "Extracted from national_industry_status.csv that the average year-over-year change rate of R&D personnel in the scientific research and technical services industry is 18.2512727272727 %.",
+    "Compared 5.80304347826087 and 18.2512727272727; since 18.2512727272727 is higher, output \"National\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Average year-over-year change rate of R&D personnel in Guangdong Province's scientific research and technical services industry": "5.80304347826087 %",
+    "Average year-over-year change rate of R&D personnel in the nationwide scientific research and technical services industry": "18.2512727272727 %",
+    "Comparison result (which is higher)": "National"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy069.json b/assets/qa_gold/enterprise_industry_analysis/easy069.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd29c629f1caeee161387fdebb9a687c94c92a20
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy069.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy069",
+  "question": "In 2022, what percentage does the total number of employees in Jilin Province's comprehensive industry represent relative to the total number of employees in the same industry nationwide?",
+  "guidelines": "The answer must be \"National\" or \"Jilin\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the total number of employees in Jilin Province's comprehensive industry is 0 persons.",
+    "Extracted from national_industry_status.csv that the total number of employees in the comprehensive industry is 439485 persons.",
+    "Calculated Jilin's percentage of the national total: 0 / 439485 * 100 = 0."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Total number of employees in Jilin Province's comprehensive industry (persons)": 0,
+    "Total number of employees in the nationwide comprehensive industry (persons)": 439485,
+    "Percentage (Jilin/National, %)": 0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy070.json b/assets/qa_gold/enterprise_industry_analysis/easy070.json
new file mode 100644
index 0000000000000000000000000000000000000000..3feea29d1dc54c3ec263388b7450afd8d9d0c1a9
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy070.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy070",
+  "question": "In 2022, comparing the median number of participation in drafting industry standards for Jilin Province's comprehensive industry and the same indicator for the nationwide comprehensive industry, which one is lower?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or a company name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the median number of participation in drafting industry standards for Jilin Province's comprehensive industry is 0.",
+    "Extracted from national_industry_status.csv that the median number of participation in drafting industry standards for the comprehensive industry is 0.",
+    "Compared 0 and 0, and determined the result is \"Equal\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Median number of participation in drafting industry standards in Jilin Province's comprehensive industry": 0,
+    "Median number of participation in drafting industry standards in the nationwide comprehensive industry": 0,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy071.json b/assets/qa_gold/enterprise_industry_analysis/easy071.json
new file mode 100644
index 0000000000000000000000000000000000000000..d5b2a9c2a355d3ec1f7a42329145c06151718e01
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy071.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy071",
+  "question": "In 2022, which is higher: the maximum asset-liability ratio in Jilin Province's commercial electrical machinery and equipment manufacturing industry, or the same indicator nationwide?",
+  "guidelines": "The answer must be \"National\" or \"Jilin\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "National",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the maximum asset-liability ratio for Jilin Province in the commercial electrical machinery and equipment manufacturing industry is 65.21 %.",
+    "Extracted from national_industry_status.csv that the maximum asset-liability ratio in the commercial electrical machinery and equipment manufacturing industry is 205.31 %.",
+    "Compared 65.21 and 205.31; since 205.31 is higher, output \"National\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Maximum asset-liability ratio in Jilin Province's commercial electrical machinery and equipment manufacturing industry": "65.21 %",
+    "Maximum asset-liability ratio in the nationwide commercial electrical machinery and equipment manufacturing industry": "205.31 %",
+    "Comparison result (which is higher)": "National"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy072.json b/assets/qa_gold/enterprise_industry_analysis/easy072.json
new file mode 100644
index 0000000000000000000000000000000000000000..8a9de171d5247b4e22101b979b5a754ddca5167f
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy072.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy072",
+  "question": "In 2022, which is larger: the number of Shanghai Stock Exchange-listed Sino-foreign joint venture enterprises in Guangdong's semiconductor industry, or the number of Shenzhen Stock Exchange-listed state-owned institute enterprises in the nationwide semiconductor industry?",
+  "guidelines": "The answer must be \"Equal\" or the comparison conclusion term specified in the question. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: number of SSE-listed Sino-foreign joint venture enterprises in Guangdong semiconductor industry = 1",
+    "Extracted from national_industry_status.csv: number of SZSE-listed state-owned institute enterprises in nationwide semiconductor industry = 1",
+    "Compared 1 and 1, and judged \"Equal\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Number of SSE-listed Sino-foreign joint venture enterprises in Guangdong semiconductor industry": 1,
+    "Number of SZSE-listed state-owned institute enterprises in nationwide semiconductor industry": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy073.json b/assets/qa_gold/enterprise_industry_analysis/easy073.json
new file mode 100644
index 0000000000000000000000000000000000000000..cdfd75024e7df7720e0789b042792da4ecf12712
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy073.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy073",
+  "question": "In 2022, which is larger: the total number of enterprises in Jilin's transportation, warehousing, and postal industry, or the number of Shenzhen Stock Exchange-listed central state-owned enterprises in the same industry nationwide?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Nationwide",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: total enterprises in Jilin transportation, warehousing and postal industry = 1",
+    "Extracted from national_industry_status.csv: number of SZSE-listed central state-owned enterprises in nationwide transportation, warehousing and postal industry = 5",
+    "Compared 1 and 5; since 5 is larger, the nationwide-side metric is larger, so output \"Nationwide\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Total enterprises in Jilin transportation, warehousing and postal industry": 1,
+    "Number of SZSE-listed central state-owned enterprises in nationwide transportation, warehousing and postal industry": 5,
+    "Comparison result (which is larger)": "Nationwide"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy074.json b/assets/qa_gold/enterprise_industry_analysis/easy074.json
new file mode 100644
index 0000000000000000000000000000000000000000..e20e4f09985802eff12814a3eafbe82f15eb3ac5
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy074.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy074",
+  "question": "In 2022, which is higher: the maximum government award funding or subsidy value in Jilin's transportation, warehousing, and postal industry, or the same metric in the nationwide industry?",
+  "guidelines": "The answer must be \"Nationwide\" or \"Jilin\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Nationwide",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: maximum government award funding or subsidy in Jilin transportation, warehousing and postal industry = 1244449.8 yuan",
+    "Extracted from national_industry_status.csv: maximum government award funding or subsidy in nationwide transportation, warehousing and postal industry = 4688008342 yuan",
+    "Compared 1244449.8 and 4688008342; since 4688008342 is higher, output \"Nationwide\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Maximum government award funding or subsidy in Jilin transportation, warehousing and postal industry (yuan)": 1244449.8,
+    "Maximum government award funding or subsidy in nationwide transportation, warehousing and postal industry (yuan)": 4688008342,
+    "Comparison result (which is higher)": "Nationwide"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy075.json b/assets/qa_gold/enterprise_industry_analysis/easy075.json
new file mode 100644
index 0000000000000000000000000000000000000000..6f1caea63cff8223454829d7ec3fe0d8ef2dcdf2
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy075.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy075",
+  "question": "In 2022, which is larger in the Tibet Autonomous Region: the total number of enterprises in the electricity, heat, gas, and water production and supply industry, or in the construction industry?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: total enterprises in Tibet electricity, heat, gas, and water production and supply industry = 0",
+    "Extracted from regional_industry_status.csv: total enterprises in Tibet construction industry = 0",
+    "Compared 0 and 0, and judged \"Equal\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Total enterprises in Tibet electricity, heat, gas, and water production and supply industry": 0,
+    "Total enterprises in Tibet construction industry": 0,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy076.json b/assets/qa_gold/enterprise_industry_analysis/easy076.json
new file mode 100644
index 0000000000000000000000000000000000000000..af702764bbe5945bbbfd8ba5f087d7d4591648ce
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy076.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy076",
+  "question": "In 2022, is there any difference between the total number of enterprises in the Tibet Autonomous Region's electricity, heat, gas, and water production and supply industry and that in the construction industry?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: total enterprises in Tibet electricity, heat, gas, and water production and supply industry = 0",
+    "Extracted from regional_industry_status.csv: total enterprises in Tibet construction industry = 0",
+    "Compared whether the two values are the same, and judged \"Equal\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Total enterprises in Tibet electricity, heat, gas, and water production and supply industry": 0,
+    "Total enterprises in Tibet construction industry": 0,
+    "Comparison result (whether there is a difference)": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy077.json b/assets/qa_gold/enterprise_industry_analysis/easy077.json
new file mode 100644
index 0000000000000000000000000000000000000000..69bf356f1a6feedd6ad2423c10683ccb061c7d02
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy077.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy077",
+  "question": "In 2022, which is larger in the Tibet Autonomous Region: the total number of enterprises in the electricity, heat, gas, and water production and supply industry, or in the leasing and business services industry?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: total enterprises in Tibet electricity, heat, gas, and water production and supply industry = 0",
+    "Extracted from regional_industry_status.csv: total enterprises in Tibet leasing and business services industry = 0",
+    "Compared 0 and 0, and judged \"Equal\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Total enterprises in Tibet electricity, heat, gas, and water production and supply industry": 0,
+    "Total enterprises in Tibet leasing and business services industry": 0,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy078.json b/assets/qa_gold/enterprise_industry_analysis/easy078.json
new file mode 100644
index 0000000000000000000000000000000000000000..a966e303601d9f33506029bb69fa43578e015f33
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy078.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy078",
+  "question": "In 2022, is there a difference between the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry and that in the commercial electrical machinery and equipment manufacturing industry?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or a company name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry is 0.",
+    "Extracted from regional_industry_status.csv that the total number of enterprises in Tibet Autonomous Region's commercial electrical machinery and equipment manufacturing industry is 0.",
+    "Compared whether the two values are consistent, and determined \"Equal\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry": 0,
+    "Total number of enterprises in Tibet Autonomous Region's commercial electrical machinery and equipment manufacturing industry": 0,
+    "Comparison result (whether there is a difference)": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy079.json b/assets/qa_gold/enterprise_industry_analysis/easy079.json
new file mode 100644
index 0000000000000000000000000000000000000000..c684d4bac9e4e449c427f09530a8a0e06fbbb132
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy079.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy079",
+  "question": "In 2022, is the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry the same as that in the metal products industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry is 0.",
+    "Extracted from regional_industry_status.csv that the total number of enterprises in Tibet Autonomous Region's metal products industry is 0.",
+    "Compared the two counts and determined \"Yes\" because they are the same."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry": 0,
+    "Total number of enterprises in Tibet Autonomous Region's metal products industry": 0,
+    "Comparison result (whether the counts are the same)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy080.json b/assets/qa_gold/enterprise_industry_analysis/easy080.json
new file mode 100644
index 0000000000000000000000000000000000000000..d15198003447737b29ee0ddaa05dc8f003915e01
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy080.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy080",
+  "question": "In 2022, comparing the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry with that in the general equipment manufacturing industry, which industry has more enterprises?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or a company name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry is 0.",
+    "Extracted from regional_industry_status.csv that the total number of enterprises in Tibet Autonomous Region's general equipment manufacturing industry is 0.",
+    "Compared 0 and 0, and determined \"Equal\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry": 0,
+    "Total number of enterprises in Tibet Autonomous Region's general equipment manufacturing industry": 0,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy081.json b/assets/qa_gold/enterprise_industry_analysis/easy081.json
new file mode 100644
index 0000000000000000000000000000000000000000..21369e126ecdbff2167d28889f29ee841114b947
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy081.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy081",
+  "question": "In 2022, which is greater: the total number of enterprises in Jilin Province's chemical raw materials and chemical products manufacturing industry, or the number of SSE-listed enterprises in the same industry in Qinghai Province?",
+  "guidelines": "The answer must be \"Equal\", \"Qinghai\", or \"Jilin\" (as specified by the question). If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the total number of enterprises in Jilin Province's chemical raw materials and chemical products manufacturing industry is 1.",
+    "Extracted from regional_industry_status.csv that the number of SSE-listed enterprises in Qinghai Province's chemical raw materials and chemical products manufacturing industry is 1.",
+    "Compared 1 and 1, and determined \"Equal\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Total number of enterprises in Jilin Province's chemical raw materials and chemical products manufacturing industry": 1,
+    "Number of SSE-listed enterprises in Qinghai Province's chemical raw materials and chemical products manufacturing industry": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy082.json b/assets/qa_gold/enterprise_industry_analysis/easy082.json
new file mode 100644
index 0000000000000000000000000000000000000000..8cf83931c70fe4d8904af6f397fa488c68826138
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy082.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy082",
+  "question": "In 2022, which is higher: the number of SZSE-listed private enterprises in Jilin's chemical raw materials and chemical products manufacturing industry, or the number of SSE-listed local state-owned enterprises in the same industry in Qinghai?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or company name without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: SZSE-listed private enterprise count in Jilin chemical raw materials and chemical products manufacturing = 1",
+    "Extracted from regional_industry_status.csv: SSE-listed local state-owned enterprise count in Qinghai chemical raw materials and chemical products manufacturing = 1",
+    "Compared 1 and 1, and judged \"Equal\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Number of SZSE-listed private enterprises in Jilin chemical raw materials and chemical products manufacturing": 1,
+    "Number of SSE-listed local state-owned enterprises in Qinghai chemical raw materials and chemical products manufacturing": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy083.json b/assets/qa_gold/enterprise_industry_analysis/easy083.json
new file mode 100644
index 0000000000000000000000000000000000000000..0064ea0e2bddbfab08a98a8f98523b630b5c312c
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy083.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy083",
+  "question": "In 2022, which is higher: the average capitalized R&D investment in Jilin's petroleum processing, coking, and nuclear fuel processing industry, or the same metric in Xinjiang?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or company name without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: average capitalized R&D investment in Jilin petroleum processing, coking and nuclear fuel processing = 0 yuan",
+    "Extracted from regional_industry_status.csv: average capitalized R&D investment in Xinjiang petroleum processing, coking and nuclear fuel processing = 0 yuan",
+    "Compared 0 and 0, and judged \"Equal\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Average capitalized R&D investment in Jilin petroleum processing, coking and nuclear fuel processing (yuan)": 0,
+    "Average capitalized R&D investment in Xinjiang petroleum processing, coking and nuclear fuel processing (yuan)": 0,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy084.json b/assets/qa_gold/enterprise_industry_analysis/easy084.json
new file mode 100644
index 0000000000000000000000000000000000000000..8fc49e746d06a9235cde1a78829d799e64b90263
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy084.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy084",
+  "question": "In 2022, which is larger: the number of HKEX-listed central state-owned enterprises in China's petroleum processing, coking, and nuclear fuel processing industry, or the total number of enterprises in the textiles, footwear, and apparel industry?",
+  "guidelines": "The answer must be either \"Petroleum Processing, Coking and Nuclear Fuel Processing\" or \"Textiles, Footwear and Apparel\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Textiles, Footwear and Apparel",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from national_industry_status.csv: HKEX-listed central state-owned enterprise count in petroleum processing, coking and nuclear fuel processing = 2",
+    "Extracted from national_industry_status.csv: enterprise count in textiles, footwear and apparel = 177",
+    "Compared 2 and 177; since 177 is larger, output \"Textiles, Footwear and Apparel\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Number of HKEX-listed central state-owned enterprises in nationwide petroleum processing, coking and nuclear fuel processing": 2,
+    "Enterprise count in nationwide textiles, footwear and apparel": 177,
+    "Comparison result (which is larger)": "Textiles, Footwear and Apparel"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy085.json b/assets/qa_gold/enterprise_industry_analysis/easy085.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d6de4644b46f7033f8e0ac001afd365bc9e127f
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy085.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy085",
+  "question": "In 2022, which is higher: the median number of National Technological Invention Awards in China's accommodation and catering industry, or in the real estate industry?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or company name without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from national_industry_status.csv: median National Technological Invention Awards in accommodation and catering industry = 0",
+    "Extracted from national_industry_status.csv: median National Technological Invention Awards in real estate industry = 0",
+    "Compared 0 and 0, and judged \"Equal\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Median National Technological Invention Awards in nationwide accommodation and catering industry": 0,
+    "Median National Technological Invention Awards in nationwide real estate industry": 0,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy086.json b/assets/qa_gold/enterprise_industry_analysis/easy086.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb77f64f902d16718d4e1f6bcf0ffe1a50f7585a
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy086.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy086",
+  "question": "In 2022, which is lower: the minimum cumulative number of Chinese invention patent applications in China's automobile manufacturing industry, or in the metal smelting and rolling processing industry?",
+  "guidelines": "The answer must be an industry name. Output only one term without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Metal Smelting and Rolling Processing",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from national_industry_status.csv: minimum cumulative Chinese invention patent applications in automobile manufacturing = 2",
+    "Extracted from national_industry_status.csv: minimum cumulative Chinese invention patent applications in metal smelting and rolling processing = 1",
+    "Compared 2 and 1; since 1 is lower, output \"Metal Smelting and Rolling Processing\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Minimum cumulative Chinese invention patent applications in nationwide automobile manufacturing": 2,
+    "Minimum cumulative Chinese invention patent applications in nationwide metal smelting and rolling processing": 1,
+    "Comparison result (which is lower)": "Metal Smelting and Rolling Processing"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy088.json b/assets/qa_gold/enterprise_industry_analysis/easy088.json
new file mode 100644
index 0000000000000000000000000000000000000000..4a5776d505e366b3ce18ed5928c87a8c28d0efd8
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy088.json
@@ -0,0 +1,21 @@
+{
+  "id": "easy088",
+  "question": "Were the policy \"Notice on Qualification Recognition Matters for Relevant R&D Institutions in Pudong New Area of Shanghai Applicable to Import Tax Policies\" and the policy \"Notice of the General Office of the Shanghai Municipal People's Government on Issuing the Action Plan for Cultivating the New Metaverse Track\" issued by the same department?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted the issuing authority fields of the two policies from policy_resource.csv.",
+    "Compared the issuing authority texts of the two policies, found them inconsistent, and judged \"No\"."
+  ],
+  "steps_num": 2,
+  "milestone": {
+    "Issuing authorities of Policy 1": "Ministry of Finance, Ministry of Science and Technology, Ministry of Civil Affairs, Ministry of Commerce, General Administration of Customs, State Taxation Administration",
+    "Issuing authority of Policy 2": "General Office of the Shanghai Municipal People's Government",
+    "Comparison result (whether issued by the same department)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy090.json b/assets/qa_gold/enterprise_industry_analysis/easy090.json
new file mode 100644
index 0000000000000000000000000000000000000000..83b38fc7b7dbdb30736d29dd4423438fa491b8c5
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy090.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy090",
+  "question": "Between the minimum change in the R&D expenditure ratio of the Information Transmission, Software and IT Services industry and that of Other Manufacturing in China, which one is smaller?",
+  "guidelines": "The answer must be a single number with two decimal places. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Information Transmission, Software and IT Services",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from national_industry_status.csv that the minimum change in the R&D expenditure ratio for Information Transmission, Software and IT Services is -472 %.",
+    "Extract from national_industry_status.csv that the minimum change in the R&D expenditure ratio for Other Manufacturing is -19.89 %.",
+    "Compare the two minimum values: -472 % < -19.89 %, so Information Transmission, Software and IT Services is smaller."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Information Transmission, Software and IT Services minimum change in R&D expenditure ratio": -472,
+    "Other Manufacturing minimum change in R&D expenditure ratio": -19.89,
+    "Comparison result (smaller one)": "Information Transmission, Software and IT Services"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy091.json b/assets/qa_gold/enterprise_industry_analysis/easy091.json
new file mode 100644
index 0000000000000000000000000000000000000000..d9b50a5cfdb9dfcb743cf759a05ca778c3aeda75
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy091.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy091",
+  "question": "Are Wuli Changyuan Wholesale Company and Xinhua Yuantong Chain Company in a competitive relationship?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Wuli Changyuan Wholesale Company is in the Wholesale and Retail industry.",
+    "Extract from company_profile.csv that Xinhua Yuantong Chain Company is in the Wholesale and Retail industry.",
+    "Since the two companies are in the same industry, they are in a competitive relationship; output \"Yes\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Wuli Changyuan Wholesale Company": "Wholesale and Retail",
+    "Industry of Xinhua Yuantong Chain Company": "Wholesale and Retail",
+    "Competitive relationship": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy092.json b/assets/qa_gold/enterprise_industry_analysis/easy092.json
new file mode 100644
index 0000000000000000000000000000000000000000..e1b0b706ea92e5f26701f130962a6833a651a475
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy092.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy092",
+  "question": "Are Huadianeng Jin Hydropower Company and Huaneng Zeze New Energy Company in a competitive relationship?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Huadianeng Jin Hydropower Company is in the Electricity, Heat, Gas and Water Production and Supply industry.",
+    "Extract from company_profile.csv that Huaneng Zeze New Energy Company is in the Electricity, Heat, Gas and Water Production and Supply industry.",
+    "Since the two companies are in the same industry, a competitive relationship exists; output \"Yes\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Huadianeng Jin Hydropower Company": "Electricity, Heat, Gas and Water Production and Supply",
+    "Industry of Huaneng Zeze New Energy Company": "Electricity, Heat, Gas and Water Production and Supply",
+    "Competitive relationship": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy093.json b/assets/qa_gold/enterprise_industry_analysis/easy093.json
new file mode 100644
index 0000000000000000000000000000000000000000..d481f84d574eeccceed84165efe87da3e1bed93a
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy093.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy093",
+  "question": "Are Run Hui Shu Ke Technology Co., Ltd. and Hang Fa Tie Chuan Hang Kong Technology Co., Ltd. in the same industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Hang Fa Tie Chuan Hang Kong Technology Co., Ltd.'s industry is Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "From company_profile.csv, extract that Run Hui Shu Ke Technology Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "The two companies are in different industries; they are not peers; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Hang Fa Tie Chuan Hang Kong Technology Co., Ltd. industry": "Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "Run Hui Shu Ke Technology Co., Ltd. industry": "Information Transmission, Software and IT Services",
+    "Whether same industry": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy094.json b/assets/qa_gold/enterprise_industry_analysis/easy094.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba2008204a575ef003786a534f2a70453559d9dd
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy094.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy094",
+  "question": "Are Zhong Fang Chang Da Zhong Gong Co., Ltd. and Bao Xin Zhi Zhi Xi Tong Co., Ltd. in a competitive relationship?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong Fang Chang Da Zhong Gong Co., Ltd.'s industry is Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "From company_profile.csv, extract that Bao Xin Zhi Zhi Xi Tong Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "The two companies are in different industries; they do not constitute a competitive relationship; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Zhong Fang Chang Da Zhong Gong Co., Ltd. industry": "Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "Bao Xin Zhi Zhi Xi Tong Co., Ltd. industry": "Information Transmission, Software and IT Services",
+    "Competitive relationship": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy095.json b/assets/qa_gold/enterprise_industry_analysis/easy095.json
new file mode 100644
index 0000000000000000000000000000000000000000..74f724881f45a36b1206046fb0edefb7b2ffc54e
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy095.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy095",
+  "question": "Are Da Zu Jin Jing She Bei Co., Ltd. and Xi Fen Ye Jin Jin Shu Co., Ltd. competitors?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Da Zu Jin Jing She Bei Co., Ltd.'s industry is Other Manufacturing",
+    "From company_profile.csv, extract that Xi Fen Ye Jin Jin Shu Co., Ltd.'s industry is Metal Smelting and Rolling Processing",
+    "The two companies are in different industries; they are not competitors; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Da Zu Jin Jing She Bei Co., Ltd. industry": "Other Manufacturing",
+    "Xi Fen Ye Jin Jin Shu Co., Ltd. industry": "Metal Smelting and Rolling Processing",
+    "Whether competitors": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy096.json b/assets/qa_gold/enterprise_industry_analysis/easy096.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0faa26b55e05dfdb70c99f2b7c6ca3901ecd742
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy096.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy096",
+  "question": "Are Lv Tai Jie Xun Huan Bao Technology Co., Ltd. and Feng Huo Chuang Ze Wang Luo She Bei Co., Ltd. competitors?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Lv Tai Jie Xun Huan Bao Technology Co., Ltd.'s industry is Comprehensive Utilization of Waste Resources",
+    "From company_profile.csv, extract that Feng Huo Chuang Ze Wang Luo She Bei Co., Ltd.'s industry is Communication Transmission Equipment",
+    "The two companies are in different industries; they are not competitors; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Lv Tai Jie Xun Huan Bao Technology Co., Ltd. industry": "Comprehensive Utilization of Waste Resources",
+    "Feng Huo Chuang Ze Wang Luo She Bei Co., Ltd. industry": "Communication Transmission Equipment",
+    "Whether competitors": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy097.json b/assets/qa_gold/enterprise_industry_analysis/easy097.json
new file mode 100644
index 0000000000000000000000000000000000000000..12ad471923d30c16c0b1ab45b797cf1b80ea131e
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy097.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy097",
+  "question": "Are Lv Shan Zhi Jin Real Estate Development Co., Ltd. and Huan Qiu Tai Jin Zhi Neng Dian Qi Co., Ltd. in the same industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Huan Qiu Tai Jin Zhi Neng Dian Qi Co., Ltd.'s industry is Commercial Electrical Machinery and Equipment Manufacturing",
+    "From company_profile.csv, extract that Lv Shan Zhi Jin Real Estate Development Co., Ltd.'s industry is Real Estate",
+    "The two companies are in different industries; they are not peers; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Huan Qiu Tai Jin Zhi Neng Dian Qi Co., Ltd. industry": "Commercial Electrical Machinery and Equipment Manufacturing",
+    "Lv Shan Zhi Jin Real Estate Development Co., Ltd. industry": "Real Estate",
+    "Whether same industry": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy098.json b/assets/qa_gold/enterprise_industry_analysis/easy098.json
new file mode 100644
index 0000000000000000000000000000000000000000..c69c6e2f626a59423d95bc665973182b663df2a1
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy098.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy098",
+  "question": "Are Jingxin Ruihui Microelectronics Company and Ruixin Yaolan Integrated Circuit Company in the same industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Ruixin Yaolan Integrated Circuit Company is in the Semiconductor industry.",
+    "Extract from company_profile.csv that Jingxin Ruihui Microelectronics Company is in the Semiconductor industry.",
+    "Since the two companies are in the same industry, they are peers; output \"Yes\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Ruixin Yaolan Integrated Circuit Company": "Semiconductor industry",
+    "Industry of Jingxin Ruihui Microelectronics Company": "Semiconductor industry",
+    "Whether same industry": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy099.json b/assets/qa_gold/enterprise_industry_analysis/easy099.json
new file mode 100644
index 0000000000000000000000000000000000000000..1b7e4dab6d6fd39d3f74ab79b0f09c823514e096
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy099.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy099",
+  "question": "Will a downturn in the rubber and plastic products industry directly affect the operating conditions of Yao Shi Yuan Ze Sheng Wu Yi Yao Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Yao Shi Yuan Ze Sheng Wu Yi Yao Co., Ltd.'s industry is Pharmaceutical Manufacturing",
+    "Conclude that a downturn in the rubber and plastic products industry is not directly equivalent to the operating performance measure of pharmaceutical manufacturing and does not constitute a direct impact",
+    "Output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Yao Shi Yuan Ze Sheng Wu Yi Yao Co., Ltd. industry": "Pharmaceutical Manufacturing",
+    "Comparison industry": "Rubber and Plastic Products",
+    "Whether direct impact": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy100.json b/assets/qa_gold/enterprise_industry_analysis/easy100.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c65fd00bf19241a5de286c02a255fde6035c9d9
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy100.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy100",
+  "question": "Will a downturn in the Information Transmission, Software and IT Services industry directly affect the operating conditions of Zhong Ji Chang Yuan Gang Tie Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong Ji Chang Yuan Gang Tie Co., Ltd.'s industry is Metal Products",
+    "Conclude that Information Transmission, Software and IT Services and Metal Products do not belong to the same industry category",
+    "Therefore a downturn in the former is not directly equivalent to a direct impact on the company's operating conditions; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Zhong Ji Chang Yuan Gang Tie Co., Ltd. industry": "Metal Products",
+    "Comparison industry": "Information Transmission, Software and IT Services",
+    "Whether direct impact": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy101.json b/assets/qa_gold/enterprise_industry_analysis/easy101.json
new file mode 100644
index 0000000000000000000000000000000000000000..536d17e02da64634ac2bab250a3f2ca471244da6
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy101.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy101",
+  "question": "Does the non-metallic mineral products industry include Lang Ji Lian Chuang Xin Xi Ji Shu Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Lang Ji Lian Chuang Xin Xi Ji Shu Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Compare with the industry named in the question, \"Non-metallic Mineral Products\"; the two are not the same",
+    "Therefore the company does not belong to non-metallic mineral products; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Lang Ji Lian Chuang Xin Xi Ji Shu Co., Ltd. industry": "Information Transmission, Software and IT Services",
+    "Comparison industry": "Non-metallic Mineral Products",
+    "Whether belongs to this industry": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy102.json b/assets/qa_gold/enterprise_industry_analysis/easy102.json
new file mode 100644
index 0000000000000000000000000000000000000000..922e1d11068e418bacacf932c503e4cb6c36f254
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy102.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy102",
+  "question": "Does Lu An Fu Chang Mei Tan Co., Ltd. belong to the Comprehensive Utilization of Waste Resources industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Lu An Fu Chang Mei Tan Co., Ltd.'s industry is Mining",
+    "Compare with the industry named in the question, \"Comprehensive Utilization of Waste Resources\"; the two are not the same",
+    "Therefore the company does not belong to comprehensive utilization of waste resources; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Lu An Fu Chang Mei Tan Co., Ltd. industry": "Mining",
+    "Comparison industry": "Comprehensive Utilization of Waste Resources",
+    "Whether belongs to this industry": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy103.json b/assets/qa_gold/enterprise_industry_analysis/easy103.json
new file mode 100644
index 0000000000000000000000000000000000000000..098fcc684a166f2d683772df00122a8a932813f8
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy103.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy103",
+  "question": "Does Heng Li Yun Chuang Xin Xi Ji Shu Co., Ltd. belong to the Automobile Manufacturing industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Heng Li Yun Chuang Xin Xi Ji Shu Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Compare with the industry named in the question, \"Automobile Manufacturing\"; the two are not the same",
+    "Therefore the company does not belong to automobile manufacturing; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Heng Li Yun Chuang Xin Xi Ji Shu Co., Ltd. industry": "Information Transmission, Software and IT Services",
+    "Comparison industry": "Automobile Manufacturing",
+    "Whether belongs to this industry": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy104.json b/assets/qa_gold/enterprise_industry_analysis/easy104.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae33b2269fd5451c128d1ab4059340de97acd03c
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy104.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy104",
+  "question": "Does Wan Hui Sheng Zhi Construction Development Co., Ltd. belong to the Real Estate industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Wan Hui Sheng Zhi Construction Development Co., Ltd.'s industry is Real Estate",
+    "This matches the industry named in the question, \"Real Estate\"",
+    "Therefore the company belongs to real estate; output \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Wan Hui Sheng Zhi Construction Development Co., Ltd. industry": "Real Estate",
+    "Comparison industry": "Real Estate",
+    "Whether belongs to this industry": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy105.json b/assets/qa_gold/enterprise_industry_analysis/easy105.json
new file mode 100644
index 0000000000000000000000000000000000000000..e0e326aa00654133f38c292348699fe9858e0f9d
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy105.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy105",
+  "question": "Does Hang Fa Yuan Jin Hang Kong Technology Co., Ltd. belong to Education?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Hang Fa Yuan Jin Hang Kong Technology Co., Ltd.'s industry is Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "Compare with the industry named in the question, \"Education\"; the two are not the same",
+    "Therefore the company does not belong to education; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Hang Fa Yuan Jin Hang Kong Technology Co., Ltd. industry": "Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "Comparison industry": "Education",
+    "Whether belongs to this industry": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy106.json b/assets/qa_gold/enterprise_industry_analysis/easy106.json
new file mode 100644
index 0000000000000000000000000000000000000000..46d8864193f0e2182cbf6c95d68b43c92b7054d4
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy106.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy106",
+  "question": "Will a downturn in the Water Conservancy, Environment and Public Facilities Management industry directly affect the operating conditions of Hua Lu Rong Rong Hua Xue Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Hua Lu Rong Rong Hua Xue Co., Ltd.'s industry is Chemical Raw Materials and Chemical Products Manufacturing",
+    "Conclude that Water Conservancy, Environment and Public Facilities Management and Chemical Raw Materials and Chemical Products Manufacturing do not belong to the same industry category",
+    "Therefore a downturn in the former is not directly equivalent to a direct impact on the company's operating conditions; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Hua Lu Rong Rong Hua Xue Co., Ltd. industry": "Chemical Raw Materials and Chemical Products Manufacturing",
+    "Comparison industry": "Water Conservancy, Environment and Public Facilities Management",
+    "Whether direct impact": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy107.json b/assets/qa_gold/enterprise_industry_analysis/easy107.json
new file mode 100644
index 0000000000000000000000000000000000000000..56054c65717c6b0d5c218cdd17d5afb92916620f
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy107.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy107",
+  "question": "Does San San Gong Ji Technology Co., Ltd. belong to the Non-metallic Mineral Products industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that San San Gong Ji Technology Co., Ltd.'s industry is Specialized Equipment Manufacturing",
+    "Compare with the industry named in the question, \"Non-metallic Mineral Products\"; the two are not the same",
+    "Therefore the company does not belong to non-metallic mineral products; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "San San Gong Ji Technology Co., Ltd. industry": "Specialized Equipment Manufacturing",
+    "Comparison industry": "Non-metallic Mineral Products",
+    "Whether belongs to this industry": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy108.json b/assets/qa_gold/enterprise_industry_analysis/easy108.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d864ef1f76c5de06968a71ccc3ee6bcd3e55f72
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy108.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy108",
+  "question": "Is Zhong Ju Yue Yin Shi Pin Co., Ltd. registered in Chongqing Municipality?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong Ju Yue Yin Shi Pin Co., Ltd.'s registration province is Hainan Province",
+    "Compare Hainan Province with the province named in the question, \"Chongqing Municipality\"; the two are not the same",
+    "Therefore the company is not registered in Chongqing Municipality; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Zhong Ju Yue Yin Shi Pin Co., Ltd. registration province": "Hainan Province",
+    "Comparison province": "Chongqing Municipality",
+    "Whether registered in Chongqing Municipality": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy109.json b/assets/qa_gold/enterprise_industry_analysis/easy109.json
new file mode 100644
index 0000000000000000000000000000000000000000..0ec7d7c76cb6af044c604a97236ba104c5685b74
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy109.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy109",
+  "question": "Does Hua Cheng Jin Jin Zong He Kai Fa Co., Ltd. contribute to the development of Zhejiang Province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Hua Cheng Jin Jin Zong He Kai Fa Co., Ltd.'s registration region is Hong Kong SAR",
+    "Compare with the province named in the question, \"Zhejiang Province\"; the two are not the same",
+    "Therefore the condition stated in the question is not met; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Hua Cheng Jin Jin Zong He Kai Fa Co., Ltd. registration region": "Hong Kong SAR",
+    "Comparison province": "Zhejiang Province",
+    "Whether contributes to development of Zhejiang Province": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy110.json b/assets/qa_gold/enterprise_industry_analysis/easy110.json
new file mode 100644
index 0000000000000000000000000000000000000000..8cdee3b1e1f79c87eaef757f4f3f1640bc845424
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy110.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy110",
+  "question": "Does Guangdong Province have the enterprise Lv Shan Chan Jin Zhi Ye Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Lv Shan Chan Jin Zhi Ye Co., Ltd.'s registration province is Guangdong Province",
+    "This matches the province named in the question, \"Guangdong Province\"",
+    "Therefore Guangdong Province has this enterprise; output \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Lv Shan Chan Jin Zhi Ye Co., Ltd. registration province": "Guangdong Province",
+    "Comparison province": "Guangdong Province",
+    "Whether Guangdong Province has this enterprise": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy111.json b/assets/qa_gold/enterprise_industry_analysis/easy111.json
new file mode 100644
index 0000000000000000000000000000000000000000..d83d5171279ad8bc2e597c22f44aef993f2e9eb2
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy111.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy111",
+  "question": "Is Hua Xin Ze Chang Xin Cai Liao Co., Ltd. registered in Hebei Province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Hua Xin Ze Chang Xin Cai Liao Co., Ltd.'s registration province is Guangdong Province",
+    "Compare Guangdong Province with the province named in the question, \"Hebei Province\"; the two are not the same",
+    "Therefore the company is not registered in Hebei Province; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Hua Xin Ze Chang Xin Cai Liao Co., Ltd. registration province": "Guangdong Province",
+    "Comparison province": "Hebei Province",
+    "Whether registered in Hebei Province": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy112.json b/assets/qa_gold/enterprise_industry_analysis/easy112.json
new file mode 100644
index 0000000000000000000000000000000000000000..0acd4e2b44de8c555f1e77829190446d7096975e
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy112.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy112",
+  "question": "Does Wan Hui Jin Sheng Real Estate Development Co., Ltd. contribute to the development of Guangdong Province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Wan Hui Jin Sheng Real Estate Development Co., Ltd.'s registration province is Guangdong Province",
+    "This matches the province named in the question, \"Guangdong Province\"",
+    "Therefore it satisfies the condition stated in the question; output \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Wan Hui Jin Sheng Real Estate Development Co., Ltd. registration province": "Guangdong Province",
+    "Comparison province": "Guangdong Province",
+    "Whether contributes to development of Guangdong Province": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy113.json b/assets/qa_gold/enterprise_industry_analysis/easy113.json
new file mode 100644
index 0000000000000000000000000000000000000000..27a48b97aac0064e1772ad34051670e34a371c7d
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy113.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy113",
+  "question": "Will changes in Anhui Province's economic environment affect Bao Jin Jin Chang Tong Ye Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Bao Jin Jin Chang Tong Ye Co., Ltd.'s registration province is Jiangsu Province",
+    "Compare Jiangsu Province with the province named in the question, \"Anhui Province\"; the two are not the same",
+    "Therefore changes in Anhui Province's economic environment are not directly relevant to this enterprise; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Bao Jin Jin Chang Tong Ye Co., Ltd. registration province": "Jiangsu Province",
+    "Comparison province": "Anhui Province",
+    "Whether affected by Anhui Province economic environment changes": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy114.json b/assets/qa_gold/enterprise_industry_analysis/easy114.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d1f6c6aa2405fce9590f1aceacf70b2a1e65ea0
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy114.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy114",
+  "question": "Will changes in Guizhou Province's economic environment affect Zhong You Zheng Da Jin Yun Shu Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong You Zheng Da Jin Yun Shu Co., Ltd.'s registration province is Shanghai Municipality",
+    "Compare Shanghai Municipality with the province named in the question, \"Guizhou Province\"; the two are not the same",
+    "Therefore changes in Guizhou Province's economic environment are not directly relevant to this enterprise; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Zhong You Zheng Da Jin Yun Shu Co., Ltd. registration province": "Shanghai Municipality",
+    "Comparison province": "Guizhou Province",
+    "Whether affected by Guizhou Province economic environment changes": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy115.json b/assets/qa_gold/enterprise_industry_analysis/easy115.json
new file mode 100644
index 0000000000000000000000000000000000000000..64d1810216e71186ea2b7c134d010b46d9e2c269
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy115.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy115",
+  "question": "Does Huan Xing Jin Ya Apparel Co., Ltd. contribute to the development of Jiangxi Province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: registered province of Huan Xing Jin Ya Apparel Co., Ltd. = Jiangsu Province",
+    "Compared Jiangsu Province with the question's province, Jiangxi Province; they do not match",
+    "Therefore, the condition in the question is not satisfied; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Registered province of Huan Xing Jin Ya Apparel Co., Ltd.": "Jiangsu Province",
+    "Compared province": "Jiangxi Province",
+    "Whether it contributes to Jiangxi Province's development": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy116.json b/assets/qa_gold/enterprise_industry_analysis/easy116.json
new file mode 100644
index 0000000000000000000000000000000000000000..83f8e373c8d3a6cf49717e9a991247b602b21f0b
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy116.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy116",
+  "question": "Would changes in Guangdong Province's economic environment affect Hua Dian Neng Jin Hydropower Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: registered province of Hua Dian Neng Jin Hydropower Co., Ltd. = Guangdong Province",
+    "Compared Guangdong Province with the question's province, Guangdong Province; they match",
+    "Therefore, provincial economic changes described in the question can affect the company's operating environment; output \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Registered province of Hua Dian Neng Jin Hydropower Co., Ltd.": "Guangdong Province",
+    "Compared province": "Guangdong Province",
+    "Whether affected by economic changes in Guangdong Province": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/easy117.json b/assets/qa_gold/enterprise_industry_analysis/easy117.json
new file mode 100644
index 0000000000000000000000000000000000000000..b429c4731cbe33c5175e167c3dbe12f5268625e0
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/easy117.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy117",
+  "question": "Is Zhongke Zhiyun Data Services Co., Ltd. registered in Guangdong Province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: registered province of Zhongke Zhiyun Data Services Co., Ltd. = Guangdong Province",
+    "Compared Guangdong Province with the question's province, Guangdong Province; they match",
+    "Therefore, the company is registered in Guangdong Province; output \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Registered province of Zhongke Zhiyun Data Services Co., Ltd.": "Guangdong Province",
+    "Compared province": "Guangdong Province",
+    "Whether registered in Guangdong Province": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium001.json b/assets/qa_gold/enterprise_industry_analysis/medium001.json
new file mode 100644
index 0000000000000000000000000000000000000000..8aba5879a4eb0f5b0ef5b7a4d1cd1c44203f98e5
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium001.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium001",
+  "question": "In 2022, what is the difference between the year-over-year employee change rate of Kangsheng Kangjian Pharmaceutical Co., Ltd. and the minimum level of its industry?",
+  "guidelines": "The answer must be a number with two decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 94.35,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Kangsheng Kangjian Pharmaceutical Co., Ltd. 2022 year-over-year employee change rate is 28.63 %",
+    "Extracted from company_profile.csv: the company's industry is Pharmaceutical Manufacturing",
+    "Extracted from national_industry_status.csv: for Pharmaceutical Manufacturing, the minimum year-over-year employee change rate is -65.72 %",
+    "Computed difference: 28.63 - (-65.72) = 94.35"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Kangsheng Kangjian Pharmaceutical Co., Ltd. year-over-year employee change rate": "28.63 %",
+    "Industry of Kangsheng Kangjian Pharmaceutical Co., Ltd.": "Pharmaceutical Manufacturing",
+    "Minimum year-over-year employee change rate in Pharmaceutical Manufacturing": "-65.72 %",
+    "Difference between the two": 94.35
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium002.json b/assets/qa_gold/enterprise_industry_analysis/medium002.json
new file mode 100644
index 0000000000000000000000000000000000000000..6ee92966fce5ed6860c3554c0623eba162bebd42
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium002.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium002",
+  "question": "In 2022, what is the difference between the total number of employees of Kangsheng Kangjian Pharmaceutical Co., Ltd. and the industry average?",
+  "guidelines": "The answer must be an exact number and keep all significant decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -958.60986547085,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Kangsheng Kangjian Pharmaceutical Co., Ltd. total employees in 2022 is 2099.0",
+    "Extracted from company_profile.csv: the company's industry is Pharmaceutical Manufacturing",
+    "Extracted from national_industry_status.csv: average total employees in Pharmaceutical Manufacturing is 3057.60986547085",
+    "Computed difference (company - industry average): 2099.0 - 3057.60986547085 = -958.60986547085"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Kangsheng Kangjian Pharmaceutical Co., Ltd. total employees in 2022": 2099.0,
+    "Industry of Kangsheng Kangjian Pharmaceutical Co., Ltd.": "Pharmaceutical Manufacturing",
+    "Average total employees in Pharmaceutical Manufacturing": 3057.60986547085,
+    "Difference (company - industry average)": -958.60986547085
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium003.json b/assets/qa_gold/enterprise_industry_analysis/medium003.json
new file mode 100644
index 0000000000000000000000000000000000000000..e88da7bd6bbe72463823f6ef8fb191aa3eeeba80
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium003.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium003",
+  "question": "In 2022, what is the difference between the total number of employees of Ling You Se Ye Da Zi Yuan Co., Ltd. and the industry maximum?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -65010.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Ling You Se Ye Da Zi Yuan Co., Ltd. total employees in 2022 = 2367.0",
+    "Extracted from company_profile.csv: the company belongs to Metal Smelting and Rolling Processing Industry",
+    "Extracted from national_industry_status.csv: industry maximum total employees = 67377",
+    "Calculated by requirement: difference (company - industry maximum) = 2367.0 - 67377 = -65010.0, output with one decimal place"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Ling You Se Ye Da Zi Yuan Co., Ltd. total employees in 2022": 2367.0,
+    "Industry of Ling You Se Ye Da Zi Yuan Co., Ltd.": "Metal Smelting and Rolling Processing Industry",
+    "Maximum total employees in Metal Smelting and Rolling Processing Industry": 67377,
+    "Difference (company - industry maximum)": -65010.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium004.json b/assets/qa_gold/enterprise_industry_analysis/medium004.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f266827623ab255b46fac4d829103f99beadca7
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium004.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium004",
+  "question": "In 2022, what is the difference between the year-over-year net profit change rate of Ling You Se Ye Da Zi Yuan Co., Ltd. and the minimum value of this indicator in its industry?",
+  "guidelines": "The answer must be a number with two decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 2110.42,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Ling You Se Ye Da Zi Yuan Co., Ltd. year-over-year net profit change rate in 2022 = -38.88 %",
+    "Extracted from company_profile.csv: the company belongs to Metal Smelting and Rolling Processing Industry",
+    "Extracted from national_industry_status.csv: minimum year-over-year net profit change rate in the industry = -2149.3 %",
+    "Calculated difference: -38.88 - (-2149.3) = 2110.42"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Ling You Se Ye Da Zi Yuan Co., Ltd. year-over-year net profit change rate in 2022": "-38.88 %",
+    "Industry of Ling You Se Ye Da Zi Yuan Co., Ltd.": "Metal Smelting and Rolling Processing Industry",
+    "Minimum year-over-year net profit change rate in Metal Smelting and Rolling Processing Industry": "-2149.3 %",
+    "Difference between the two": 2110.42
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium005.json b/assets/qa_gold/enterprise_industry_analysis/medium005.json
new file mode 100644
index 0000000000000000000000000000000000000000..225903d73306b4186696c306a83d5f9be549b46c
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium005.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium005",
+  "question": "In 2022, which is higher: the year-over-year employee change rate of Yong Feng Xin Chuang Ke Ji Co., Ltd. or the maximum value of this indicator in its industry?",
+  "guidelines": "The answer must be either \"industry\" or the company name. Output only one word or the company name, without any explanation, analysis, or descriptive text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Yong Feng Xin Chuang Ke Ji Co., Ltd. year-over-year employee change rate in 2022 = 12.61 %",
+    "Extracted from company_profile.csv: the company belongs to Information Transmission, Software and Information Technology Services",
+    "Extracted from national_industry_status.csv: industry maximum year-over-year employee change rate = 416.95 %",
+    "Compared 12.61 and 416.95; since 416.95 > 12.61, the industry is higher, so output \"industry\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Yong Feng Xin Chuang Ke Ji Co., Ltd. year-over-year employee change rate in 2022": "12.61 %",
+    "Industry of Yong Feng Xin Chuang Ke Ji Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Maximum year-over-year employee change rate in Information Transmission, Software and Information Technology Services": "416.95 %",
+    "Comparison result (whether the industry is higher)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium006.json b/assets/qa_gold/enterprise_industry_analysis/medium006.json
new file mode 100644
index 0000000000000000000000000000000000000000..466c1c5e93645837bbd167db8035bfc735d5d5c7
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium006.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium006",
+  "question": "In 2022, is the year-over-year net profit change rate of Yong Feng Xin Chuang Ke Ji Co., Ltd. higher than the median of this indicator in its industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Yong Feng Xin Chuang Ke Ji Co., Ltd. year-over-year net profit change rate in 2022 = 40.82 %",
+    "Extracted from company_profile.csv: the company belongs to Information Transmission, Software and Information Technology Services",
+    "Extracted from national_industry_status.csv: industry median year-over-year net profit change rate = -15.96 %",
+    "Compared 40.82 and -15.96; since 40.82 > -15.96, the judgment is \"Yes\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Yong Feng Xin Chuang Ke Ji Co., Ltd. year-over-year net profit change rate in 2022": "40.82 %",
+    "Industry of Yong Feng Xin Chuang Ke Ji Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Median year-over-year net profit change rate in Information Transmission, Software and Information Technology Services": "-15.96 %",
+    "Comparison result (whether the company is higher than the industry median)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium007.json b/assets/qa_gold/enterprise_industry_analysis/medium007.json
new file mode 100644
index 0000000000000000000000000000000000000000..a1d756fef3cf0c538167eb139d3bcb55290d848b
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium007.json
@@ -0,0 +1,25 @@
+{
+  "id": "medium007",
+  "question": "In 2022, what is the difference between the operating profit amount of Lian Ji Chuang Ji Ji Chuang Co., Ltd. and the total operating profit amount of the same industry in its province?",
+  "guidelines": "The answer must be a number with two decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -1738758524.92,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Lian Ji Chuang Ji Ji Chuang Co., Ltd. operating profit amount in 2022 = -41697747.51 CNY",
+    "Extracted from company_profile.csv: the company is in Anhui Province and belongs to Water Conservancy, Environment and Public Facilities Management",
+    "Extracted from regional_industry_status.csv: total operating profit amount in Anhui Province for this industry = 1697060777.41 CNY",
+    "Calculated difference (company - same-province same-industry total): -41697747.51 - 1697060777.41 = -1738758524.92"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Lian Ji Chuang Ji Ji Chuang Co., Ltd. operating profit amount in 2022 (CNY)": -41697747.51,
+    "Province of Lian Ji Chuang Ji Ji Chuang Co., Ltd.": "Anhui Province",
+    "Industry of Lian Ji Chuang Ji Ji Chuang Co., Ltd.": "Water Conservancy, Environment and Public Facilities Management",
+    "Total operating profit amount in Anhui Province for Water Conservancy, Environment and Public Facilities Management (CNY)": 1697060777.41,
+    "Difference (company - same-province same-industry total)": -1738758524.92
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium008.json b/assets/qa_gold/enterprise_industry_analysis/medium008.json
new file mode 100644
index 0000000000000000000000000000000000000000..75de5df7047efecc07fe70ba9aa5799d8634611c
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium008.json
@@ -0,0 +1,25 @@
+{
+  "id": "medium008",
+  "question": "In 2022, compared with the average level of the same industry in its province, which is higher: the number of R&D personnel of Lianji Chuangji Machine Tool Company or the industry average?",
+  "guidelines": "The answer must be either the company name or the word \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Lianji Chuangji Machine Tool Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_operation_status.csv that the number of R&D personnel of Lianji Chuangji Machine Tool Company in 2022 is 290.0.",
+    "Extract from company_profile.csv that the company is located in Anhui Province and belongs to the Water Conservancy, Environment and Public Facilities Management industry.",
+    "Extract from regional_industry_status.csv that the average number of R&D personnel in Anhui Province for this industry is 132.111111111111.",
+    "Compare 290.0 with 132.111111111111; since 290.0 > 132.111111111111, the company is higher, so output \"Lianji Chuangji Machine Tool Company\"."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Number of R&D personnel of Lianji Chuangji Machine Tool Company in 2022": 290.0,
+    "Province of Lianji Chuangji Machine Tool Company": "Anhui Province",
+    "Industry of Lianji Chuangji Machine Tool Company": "Water Conservancy, Environment and Public Facilities Management",
+    "Average number of R&D personnel in Anhui Province for the Water Conservancy, Environment and Public Facilities Management industry": 132.111111111111,
+    "Comparison result (whether the company is higher)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium009.json b/assets/qa_gold/enterprise_industry_analysis/medium009.json
new file mode 100644
index 0000000000000000000000000000000000000000..53705be98bf5f371583aeade10ccda0ccf2e5839
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium009.json
@@ -0,0 +1,25 @@
+{
+  "id": "medium009",
+  "question": "In 2022, which is higher: the operating profit amount of Run Hui Shu Zhi Xi Tong Co., Ltd. or the total operating profit amount of the same industry in its province?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Run Hui Shu Zhi Xi Tong Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Run Hui Shu Zhi Xi Tong Co., Ltd. operating profit amount in 2022 = 9041715.19 CNY",
+    "Extracted from company_profile.csv: the company is in Jilin Province and belongs to Scientific Research and Technical Services",
+    "Extracted from regional_industry_status.csv: total operating profit amount in Jilin Province for this industry = 0 CNY",
+    "Compared 9041715.19 and 0; since 9041715.19 > 0, the company is higher, so output \"Run Hui Shu Zhi Xi Tong Co., Ltd.\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Run Hui Shu Zhi Xi Tong Co., Ltd. operating profit amount in 2022 (CNY)": 9041715.19,
+    "Province of Run Hui Shu Zhi Xi Tong Co., Ltd.": "Jilin Province",
+    "Industry of Run Hui Shu Zhi Xi Tong Co., Ltd.": "Scientific Research and Technical Services",
+    "Total operating profit amount in Jilin Province for Scientific Research and Technical Services (CNY)": 0,
+    "Comparison result (whether the company is higher)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium010.json b/assets/qa_gold/enterprise_industry_analysis/medium010.json
new file mode 100644
index 0000000000000000000000000000000000000000..110d2eb80deb56092ea6f1fd6720c6b49c4f464e
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium010.json
@@ -0,0 +1,25 @@
+{
+  "id": "medium010",
+  "question": "In 2022, is the operating revenue amount of Run Hui Shu Zhi Xi Tong Co., Ltd. higher than the total operating revenue amount of the corresponding industry in its province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Run Hui Shu Zhi Xi Tong Co., Ltd. operating revenue amount in 2022 = 157705748.68 CNY",
+    "Extracted from company_profile.csv: the company is in Jilin Province and belongs to Scientific Research and Technical Services",
+    "Extracted from regional_industry_status.csv: total operating revenue amount in Jilin Province for this industry = 0 CNY",
+    "Compared 157705748.68 and 0; since 157705748.68 > 0, the judgment is \"Yes\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Run Hui Shu Zhi Xi Tong Co., Ltd. operating revenue amount in 2022 (CNY)": 157705748.68,
+    "Province of Run Hui Shu Zhi Xi Tong Co., Ltd.": "Jilin Province",
+    "Industry of Run Hui Shu Zhi Xi Tong Co., Ltd.": "Scientific Research and Technical Services",
+    "Total operating revenue amount in Jilin Province for Scientific Research and Technical Services (CNY)": 0,
+    "Comparison result (whether the company is higher than the provincial same-industry total)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium011.json b/assets/qa_gold/enterprise_industry_analysis/medium011.json
new file mode 100644
index 0000000000000000000000000000000000000000..64bbe0b85f36b6b3a1a14edbf29c3959db1a4433
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium011.json
@@ -0,0 +1,25 @@
+{
+  "id": "medium011",
+  "question": "In 2022, which is higher: the cumulative number of PCT invention patent applications of Zhong Ji Da Chang Tong Ye Co., Ltd. or the minimum value of the same indicator in the same industry in its province?",
+  "guidelines": "The answer must be \"equal\", the company name, or \"industry\". Output only one word or the company name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Zhong Ji Da Chang Tong Ye Co., Ltd. cumulative PCT invention patent applications in 2022 = 1",
+    "Extracted from company_profile.csv: the company is in Jiangsu Province and belongs to Agriculture, Forestry, Animal Husbandry and Fishery",
+    "Extracted from regional_industry_status.csv: minimum cumulative PCT invention patent applications in the provincial same industry = 1",
+    "Compared 1 and 1; they are equal, so output \"equal\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Zhong Ji Da Chang Tong Ye Co., Ltd. cumulative PCT invention patent applications in 2022": 1,
+    "Province of Zhong Ji Da Chang Tong Ye Co., Ltd.": "Jiangsu Province",
+    "Industry of Zhong Ji Da Chang Tong Ye Co., Ltd.": "Agriculture, Forestry, Animal Husbandry and Fishery",
+    "Minimum cumulative PCT invention patent applications in Jiangsu Province for Agriculture, Forestry, Animal Husbandry and Fishery": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium012.json b/assets/qa_gold/enterprise_industry_analysis/medium012.json
new file mode 100644
index 0000000000000000000000000000000000000000..4ea6a7dea6d00b240deedb08b67281a707245662
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium012.json
@@ -0,0 +1,25 @@
+{
+  "id": "medium012",
+  "question": "In 2022, is the debt-to-asset ratio of Zhong Ji Da Chang Tong Ye Co., Ltd. higher than the minimum debt-to-asset ratio of the corresponding industry in its province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Zhong Ji Da Chang Tong Ye Co., Ltd. debt-to-asset ratio in 2022 = 35.4 %",
+    "Extracted from company_profile.csv: the company is in Jiangsu Province and belongs to Agriculture, Forestry, Animal Husbandry and Fishery",
+    "Extracted from regional_industry_status.csv: minimum debt-to-asset ratio in the provincial same industry = 23.15 %",
+    "Compared 35.4 and 23.15; since 35.4 > 23.15, the judgment is \"Yes\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Zhong Ji Da Chang Tong Ye Co., Ltd. debt-to-asset ratio in 2022": "35.4 %",
+    "Province of Zhong Ji Da Chang Tong Ye Co., Ltd.": "Jiangsu Province",
+    "Industry of Zhong Ji Da Chang Tong Ye Co., Ltd.": "Agriculture, Forestry, Animal Husbandry and Fishery",
+    "Minimum debt-to-asset ratio in Jiangsu Province for Agriculture, Forestry, Animal Husbandry and Fishery": "23.15 %",
+    "Comparison result (whether the company is higher than the provincial same-industry minimum)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium013.json b/assets/qa_gold/enterprise_industry_analysis/medium013.json
new file mode 100644
index 0000000000000000000000000000000000000000..70b0ae29b7992ce398f4896cecd9a38e6f00940d
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium013.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium013",
+  "question": "In 2022, is the market capitalization of Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. lower than the operating revenue of Long He Zhi Jin Zhi Ye Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: market capitalization of Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. in 2022 = 310.0 hundred million CNY",
+    "Converted 310.0 hundred million CNY to CNY: 310.0 * 100000000 = 31000000000 CNY",
+    "Extracted from company_operation_status.csv: operating revenue of Long He Zhi Jin Zhi Ye Co., Ltd. in 2022 = 1702394443.0 CNY",
+    "Compared 31000000000 and 1702394443.0; since 31000000000 > 1702394443.0, the judgment is \"No\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Market capitalization of Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. in 2022 (hundred million CNY)": 310.0,
+    "Converted market capitalization of Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. in 2022 (CNY)": 31000000000,
+    "Operating revenue of Long He Zhi Jin Zhi Ye Co., Ltd. in 2022 (CNY)": 1702394443.0,
+    "Comparison result (whether the market capitalization of Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. is lower than the operating revenue of Long He Zhi Jin Zhi Ye Co., Ltd.)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium014.json b/assets/qa_gold/enterprise_industry_analysis/medium014.json
new file mode 100644
index 0000000000000000000000000000000000000000..f1b05ed87ff3309ba335dabd0cd8192176073f0d
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium014.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium014",
+  "question": "Comparing the number of SSE-listed state-owned enterprise institutes in the industry of Huijin Jinrui Wealth Management Co., Ltd. with the number of HKEX-listed sino-foreign joint ventures in the industry of Zhongke Zhiyun Data Services Co., Ltd., which value is larger?",
+  "guidelines": "The answer must be a company name, \"industry\", or \"Same\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Same",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Huijin Jinrui Wealth Management Co., Ltd. is Financial Industry.",
+    "Extracted from national_industry_status.csv that the number of SSE-listed state-owned enterprise institutes in Financial Industry is 1.",
+    "Extracted from company_profile.csv that the industry of Zhongke Zhiyun Data Services Co., Ltd. is Information Transmission, Software and Information Technology Services.",
+    "Extracted from national_industry_status.csv that the number of HKEX-listed sino-foreign joint ventures in Information Transmission, Software and Information Technology Services is 1.",
+    "Compared the two counts and output the result corresponding to the predefined answer field: Same."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Huijin Jinrui Wealth Management Co., Ltd.": "Financial Industry",
+    "Number of SSE-listed state-owned enterprise institutes in Financial Industry": 1,
+    "Industry of Zhongke Zhiyun Data Services Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Number of HKEX-listed sino-foreign joint ventures in Information Transmission, Software and Information Technology Services": 1,
+    "Comparison result (answer output)": "Same"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium015.json b/assets/qa_gold/enterprise_industry_analysis/medium015.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac3df44f2721ce0cdcabe161ab7ff93f43a2c454
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium015.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium015",
+  "question": "Is the minimum R&D personnel ratio in the industry of Huijin Jinrui Wealth Management Co., Ltd. lower than the minimum R&D personnel ratio in the industry of Zhongke Zhiyun Data Services Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Huijin Jinrui Wealth Management Co., Ltd. is Financial Industry.",
+    "Extracted from national_industry_status.csv that the minimum R&D personnel ratio in Financial Industry is 0.64 %.",
+    "Extracted from company_profile.csv that the industry of Zhongke Zhiyun Data Services Co., Ltd. is Information Transmission, Software and Information Technology Services.",
+    "Extracted from national_industry_status.csv that the minimum R&D personnel ratio in Information Transmission, Software and Information Technology Services is 0.92 %.",
+    "Judged whether 0.64 % is lower than 0.92 %; the conclusion is \"Yes\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Huijin Jinrui Wealth Management Co., Ltd.": "Financial Industry",
+    "Minimum R&D personnel ratio in Financial Industry": "0.64 %",
+    "Industry of Zhongke Zhiyun Data Services Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Minimum R&D personnel ratio in Information Transmission, Software and Information Technology Services": "0.92 %",
+    "Comparison result (whether the Financial Industry minimum is lower than that of Information Services)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium016.json b/assets/qa_gold/enterprise_industry_analysis/medium016.json
new file mode 100644
index 0000000000000000000000000000000000000000..fa9144ed5ea55e05ac94cf2de5b67ea7eb048c08
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium016.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium016",
+  "question": "What is the difference between the median annual number of Chinese invention patent applications in the industry of Changqiao Jinchuang Technology Co., Ltd. and the corresponding metric for the industry in Zhejiang Province where Wuli Huida Chain Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 16.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Changqiao Jinchuang Technology Co., Ltd. = Consumer Electronics and Electrical Industry",
+    "Extracted from national_industry_status.csv: median annual Chinese invention patent applications in Consumer Electronics and Electrical Industry = 18",
+    "Extracted from company_profile.csv: province of Wuli Huida Chain Co., Ltd. = Zhejiang Province",
+    "Extracted from regional_industry_status.csv: median annual Chinese invention patent applications in Zhejiang Wholesale and Retail Trade = 2",
+    "Calculated the difference: 18 - 2 = 16.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Changqiao Jinchuang Technology Co., Ltd.": "Consumer Electronics and Electrical Industry",
+    "Median annual Chinese invention patent applications in Consumer Electronics and Electrical Industry": 18,
+    "Province of Wuli Huida Chain Co., Ltd.": "Zhejiang Province",
+    "Median annual Chinese invention patent applications in Zhejiang Wholesale and Retail Trade": 2,
+    "Difference (Consumer Electronics and Electrical Industry - Zhejiang Wholesale and Retail Trade)": 16.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium017.json b/assets/qa_gold/enterprise_industry_analysis/medium017.json
new file mode 100644
index 0000000000000000000000000000000000000000..acc1ff980d1ce1f13feb3b0c5d918c4b11ca49a3
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium017.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium017",
+  "question": "What is the difference between the minimum R&D personnel ratio in the industry of Changqiao Jinchuang Technology Co., Ltd. and the corresponding metric for the industry in Zhejiang Province where Wuli Huida Chain Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to two decimal places. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 1.04,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Changqiao Jinchuang Technology Co., Ltd. = Consumer Electronics and Electrical Industry",
+    "Extracted from national_industry_status.csv: minimum R&D personnel ratio in Consumer Electronics and Electrical Industry = 1.87%",
+    "Extracted from company_profile.csv: province of Wuli Huida Chain Co., Ltd. = Zhejiang Province",
+    "Extracted from regional_industry_status.csv: minimum R&D personnel ratio in Zhejiang Wholesale and Retail Trade = 0.83%",
+    "Calculated the difference: 1.87 - 0.83 = 1.04"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Changqiao Jinchuang Technology Co., Ltd.": "Consumer Electronics and Electrical Industry",
+    "Minimum R&D personnel ratio in Consumer Electronics and Electrical Industry": "1.87%",
+    "Province of Wuli Huida Chain Co., Ltd.": "Zhejiang Province",
+    "Minimum R&D personnel ratio in Zhejiang Wholesale and Retail Trade": "0.83%",
+    "Difference (Consumer Electronics and Electrical Industry - Zhejiang Wholesale and Retail Trade)": 1.04
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium018.json b/assets/qa_gold/enterprise_industry_analysis/medium018.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0ddf490703452961e49f04074b9f2bf84d69ccf
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium018.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium018",
+  "question": "Which is greater: the number of SSE-listed local state-owned enterprises in the corresponding industry of the province where Zhong Ke Shu Ruan Software Co., Ltd. is located, or the number of SZSE-listed foreign-funded enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Zhong Ke Shu Ruan Software Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: SSE-listed local state-owned enterprise count in Guangdong Information Transmission, Software and IT Services = 1",
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: SZSE-listed foreign-funded enterprise count in Real Estate = 2",
+    "Compared 1 and 2; since 2 is greater, output \"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Zhong Ke Shu Ruan Software Co., Ltd.": "Guangdong Province",
+    "SSE-listed local state-owned enterprise count in Guangdong Information Transmission, Software and IT Services": 1,
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "SZSE-listed foreign-funded enterprise count in Real Estate": 2,
+    "Comparison result (which value is greater)": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium019.json b/assets/qa_gold/enterprise_industry_analysis/medium019.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f68d11fea2a851423966552a97bf664f76c6c76
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium019.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium019",
+  "question": "Which is greater: the number of HKEX-listed foreign-funded enterprises in the corresponding industry of the province where Zhong Ke Shu Ruan Software Co., Ltd. is located, or the number of HKEX-listed enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Zhong Ke Shu Ruan Software Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: HKEX-listed foreign-funded enterprise count in Guangdong Information Transmission, Software and IT Services = 5",
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: HKEX-listed enterprise count in Real Estate = 185",
+    "Compared 5 and 185; since 185 is greater, output \"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Zhong Ke Shu Ruan Software Co., Ltd.": "Guangdong Province",
+    "HKEX-listed foreign-funded enterprise count in Guangdong Information Transmission, Software and IT Services": 5,
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "HKEX-listed enterprise count in Real Estate": 185,
+    "Comparison result (which value is greater)": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium020.json b/assets/qa_gold/enterprise_industry_analysis/medium020.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff76aa17afb3be74cacb8f64b8ca085f79d76c85
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium020.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium020",
+  "question": "Which is greater: the number of SZSE-listed central state-owned enterprises in the industry of Bi Yuan Zhi Ze Urban Development Co., Ltd., or the number of SSE-listed local state-owned enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Bi Yuan Zhi Ze Urban Development Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: SZSE-listed central state-owned enterprise count in Real Estate = 7",
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: SSE-listed local state-owned enterprise count in Real Estate = 38",
+    "Compared 7 and 38; since 38 is greater, output \"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Bi Yuan Zhi Ze Urban Development Co., Ltd.": "Real Estate",
+    "SZSE-listed central state-owned enterprise count in Real Estate": 7,
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "SSE-listed local state-owned enterprise count in Real Estate": 38,
+    "Comparison result (which is greater)": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium021.json b/assets/qa_gold/enterprise_industry_analysis/medium021.json
new file mode 100644
index 0000000000000000000000000000000000000000..e1c4bbdd95fff2e0bafed5e75147ddf8b43e309e
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium021.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium021",
+  "question": "What is the difference between the median operating profit amount of the industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd. and that of the industry of Tong Tong Ze Hong Securities Co., Ltd.?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -880561639.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: median operating profit amount in Real Estate = 130368786",
+    "Extracted from company_profile.csv: industry of Tong Tong Ze Hong Securities Co., Ltd. = Financial Industry",
+    "Extracted from national_industry_status.csv: median operating profit amount in Financial Industry = 1010930425",
+    "Calculated the difference (Real Estate - Financial Industry): 130368786 - 1010930425 = -880561639.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd.": "Real Estate",
+    "Median operating profit amount in Real Estate (yuan)": 130368786,
+    "Industry of Tong Tong Ze Hong Securities Co., Ltd.": "Financial Industry",
+    "Median operating profit amount in Financial Industry (yuan)": 1010930425,
+    "Difference (Real Estate - Financial Industry)": -880561639.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium022.json b/assets/qa_gold/enterprise_industry_analysis/medium022.json
new file mode 100644
index 0000000000000000000000000000000000000000..78110081d6cc56b1044ee13aeae0eb28b605ba10
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium022.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium022",
+  "question": "Comparing the number of SZSE-listed central state-owned enterprises in the industry of Zhaoye Huachang Real Estate Development Co., Ltd. with the number of SZSE-listed enterprises in the industry of Tongtong Zehong Securities Co., Ltd., which is larger?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Tongtong Zehong Securities Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Zhaoye Huachang Real Estate Development Co., Ltd. is Real Estate.",
+    "Extracted from national_industry_status.csv that the number of SZSE-listed central state-owned enterprises in Real Estate is 7.",
+    "Extracted from company_profile.csv that the industry of Tongtong Zehong Securities Co., Ltd. is Financial Industry.",
+    "Extracted from national_industry_status.csv that the number of SZSE-listed enterprises in Financial Industry is 38.",
+    "Compared 7 and 38; 38 is larger, so output \"Tongtong Zehong Securities Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhaoye Huachang Real Estate Development Co., Ltd.": "Real Estate",
+    "Number of SZSE-listed central state-owned enterprises in Real Estate": 7,
+    "Industry of Tongtong Zehong Securities Co., Ltd.": "Financial Industry",
+    "Number of SZSE-listed enterprises in Financial Industry": 38,
+    "Conclusion (larger value)": "Tongtong Zehong Securities Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium023.json b/assets/qa_gold/enterprise_industry_analysis/medium023.json
new file mode 100644
index 0000000000000000000000000000000000000000..0436e00645aab35c04316561c500ee17a0363db6
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium023.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium023",
+  "question": "Comparing the number of SZSE-listed local state-owned enterprises in the industry of Aijian Yikang Fuzhongxin Co., Ltd. with the number of SZSE-listed sino-foreign joint ventures in the industry of Zhongke Zhiyun Data Services Co., Ltd., which value is larger?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Zhongke Zhiyun Data Services Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Aijian Yikang Fuzhongxin Co., Ltd. is Health and Social Work.",
+    "Extracted from national_industry_status.csv that the number of SZSE-listed local state-owned enterprises in Health and Social Work is 1.",
+    "Extracted from company_profile.csv that the industry of Zhongke Zhiyun Data Services Co., Ltd. is Information Transmission, Software and Information Technology Services.",
+    "Extracted from national_industry_status.csv that the number of SZSE-listed sino-foreign joint ventures in Information Transmission, Software and Information Technology Services is 2.",
+    "Compared 1 and 2; 2 is larger, so output \"Zhongke Zhiyun Data Services Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Aijian Yikang Fuzhongxin Co., Ltd.": "Health and Social Work",
+    "Number of SZSE-listed local state-owned enterprises in Health and Social Work": 1,
+    "Industry of Zhongke Zhiyun Data Services Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Number of SZSE-listed sino-foreign joint ventures in Information Transmission, Software and Information Technology Services": 2,
+    "Conclusion (larger value)": "Zhongke Zhiyun Data Services Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium024.json b/assets/qa_gold/enterprise_industry_analysis/medium024.json
new file mode 100644
index 0000000000000000000000000000000000000000..46dfb718f01b31633aca9fe67ee63d73dad8f62f
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium024.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium024",
+  "question": "What is the difference between the minimum value of provincial enterprise technology innovation awards in the industry of Aijian Yikang Fuzhongxin Co., Ltd. and that in the industry of Zhongke Zhiyun Data Services Co., Ltd.?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Aijian Yikang Fuzhongxin Co., Ltd. is Health and Social Work.",
+    "Extracted from national_industry_status.csv that the minimum value of provincial enterprise technology innovation awards in Health and Social Work is 0.",
+    "Extracted from company_profile.csv that the industry of Zhongke Zhiyun Data Services Co., Ltd. is Information Transmission, Software and Information Technology Services.",
+    "Extracted from national_industry_status.csv that the minimum value of provincial enterprise technology innovation awards in Information Transmission, Software and Information Technology Services is 0.",
+    "Calculated the difference: 0 - 0 = 0.0."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Aijian Yikang Fuzhongxin Co., Ltd.": "Health and Social Work",
+    "Minimum value of provincial enterprise technology innovation awards in Health and Social Work": 0,
+    "Industry of Zhongke Zhiyun Data Services Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Minimum value of provincial enterprise technology innovation awards in Information Transmission, Software and Information Technology Services": 0,
+    "Difference": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium025.json b/assets/qa_gold/enterprise_industry_analysis/medium025.json
new file mode 100644
index 0000000000000000000000000000000000000000..91c4cf47c1b774ea2cf44b38bc935d644f58b342
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium025.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium025",
+  "question": "Is the average number of provincial or ministerial natural science awards in the industry of Biyuan Shenghua Construction Development Co., Ltd. lower than that in the industry of Baoxin Huihui Network Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Biyuan Shenghua Construction Development Co., Ltd. is Real Estate.",
+    "Extracted from national_industry_status.csv that the average number of provincial or ministerial natural science awards in Real Estate is 0.",
+    "Extracted from company_profile.csv that the industry of Baoxin Huihui Network Co., Ltd. is Information Transmission, Software and Information Technology Services.",
+    "Extracted from national_industry_status.csv that the average number of provincial or ministerial natural science awards in Information Transmission, Software and Information Technology Services is 1.",
+    "Judged whether 0 is lower than 1; the conclusion is \"Yes\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Biyuan Shenghua Construction Development Co., Ltd.": "Real Estate",
+    "Average number of provincial or ministerial natural science awards in Real Estate": 0,
+    "Industry of Baoxin Huihui Network Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Average number of provincial or ministerial natural science awards in Information Transmission, Software and Information Technology Services": 1,
+    "Whether lower": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium026.json b/assets/qa_gold/enterprise_industry_analysis/medium026.json
new file mode 100644
index 0000000000000000000000000000000000000000..4202425fde05a65ee3f0b21a97a82fd87ffecab9
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium026.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium026",
+  "question": "Is the median cumulative citation count of all patents in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. lower than the same metric in the industry of Bao Xin Hui Hui Wang Luo Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: median cumulative citation count of all patents in Real Estate = 63",
+    "Extracted from company_profile.csv: industry of Bao Xin Hui Hui Wang Luo Co., Ltd. = Information Transmission, Software and IT Services",
+    "Extracted from national_industry_status.csv: median cumulative citation count of all patents in Information Transmission, Software and IT Services = 340.5",
+    "Compared 63 and 340.5; since 63 is lower, the judgment is \"Yes\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "Median cumulative citation count of all patents in Real Estate": 63,
+    "Industry of Bao Xin Hui Hui Wang Luo Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Median cumulative citation count of all patents in Information Transmission, Software and IT Services": 340.5,
+    "Whether lower": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium027.json b/assets/qa_gold/enterprise_industry_analysis/medium027.json
new file mode 100644
index 0000000000000000000000000000000000000000..91948f2d569787c92e53bddf298ad982fc8a6f21
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium027.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium027",
+  "question": "Is the number of SSE-listed enterprises in the industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. equal to the number of SSE-listed state-owned institute enterprises in the industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. = Health and Social Work",
+    "Extracted from national_industry_status.csv: number of SSE-listed enterprises in Health and Social Work = 2",
+    "Extracted from company_profile.csv: industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd. = Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Extracted from national_industry_status.csv: number of SSE-listed state-owned institute enterprises in that industry = 1",
+    "Compared 2 and 1; since they are not equal, the judgment is \"No\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd.": "Health and Social Work",
+    "Number of SSE-listed enterprises in Health and Social Work": 2,
+    "Industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd.": "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Number of SSE-listed state-owned institute enterprises in that industry": 1,
+    "Whether equal": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium028.json b/assets/qa_gold/enterprise_industry_analysis/medium028.json
new file mode 100644
index 0000000000000000000000000000000000000000..d71f6da5300ebd45295b5c789e3845c02af16b89
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium028.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium028",
+  "question": "Are the maximum values of the State Technological Invention Award metric the same between the industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. and the industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. = Health and Social Work",
+    "Extracted from national_industry_status.csv: maximum State Technological Invention Award value in Health and Social Work = 0",
+    "Extracted from company_profile.csv: industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd. = Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Extracted from national_industry_status.csv: maximum State Technological Invention Award value in that industry = 0",
+    "Compared 0 and 0; since they are the same, the judgment is \"Yes\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd.": "Health and Social Work",
+    "Maximum State Technological Invention Award value in Health and Social Work": 0,
+    "Industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd.": "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Maximum State Technological Invention Award value in that industry": 0,
+    "Whether identical": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium029.json b/assets/qa_gold/enterprise_industry_analysis/medium029.json
new file mode 100644
index 0000000000000000000000000000000000000000..5d785acfcd32afc21818fdb1a4fce42457267052
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium029.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium029",
+  "question": "Which is greater: the number of SSE-listed central state-owned enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd., or the number of SSE-listed enterprises in the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: number of SSE-listed central state-owned enterprises in Real Estate = 3",
+    "Extracted from company_profile.csv: industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. = Education",
+    "Extracted from national_industry_status.csv: number of SSE-listed enterprises in Education = 5",
+    "Compared 3 and 5; since 5 is greater, output \"Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "Number of SSE-listed central state-owned enterprises in Real Estate": 3,
+    "Industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.": "Education",
+    "Number of SSE-listed enterprises in Education": 5,
+    "Comparison result (greater)": "Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium030.json b/assets/qa_gold/enterprise_industry_analysis/medium030.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba3c64762d8ee7b6d9124e8de992e052b7381bd7
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium030.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium030",
+  "question": "Which is lower: the minimum total liabilities value in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd., or the same metric in the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: minimum total liabilities in Real Estate = -16582853780.5 yuan",
+    "Extracted from company_profile.csv: industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. = Education",
+    "Extracted from national_industry_status.csv: minimum total liabilities in Education = 30298132.02 yuan",
+    "Compared -16582853780.5 and 30298132.02; since the former is lower, output \"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "Minimum total liabilities in Real Estate (yuan)": -16582853780.5,
+    "Industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.": "Education",
+    "Minimum total liabilities in Education (yuan)": 30298132.02,
+    "Comparison result (lower)": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium031.json b/assets/qa_gold/enterprise_industry_analysis/medium031.json
new file mode 100644
index 0000000000000000000000000000000000000000..1df976909750fc51adce479cb8fe0e8a7a1a52a2
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium031.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium031",
+  "question": "Which is greater: the number of SSE-listed state-owned institute enterprises in the industry of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd., or the number of SZSE-listed private enterprises in the industry of Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. = Financial Industry",
+    "Extracted from national_industry_status.csv: number of SSE-listed state-owned institute enterprises in Financial Industry = 1",
+    "Extracted from company_profile.csv: industry of Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: number of SZSE-listed private enterprises in Real Estate = 24",
+    "Compared 1 and 24; since 24 is greater, output \"Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.": "Financial Industry",
+    "Number of SSE-listed state-owned institute enterprises in Financial Industry": 1,
+    "Industry of Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.": "Real Estate",
+    "Number of SZSE-listed private enterprises in Real Estate": 24,
+    "Comparison result (greater)": "Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium032.json b/assets/qa_gold/enterprise_industry_analysis/medium032.json
new file mode 100644
index 0000000000000000000000000000000000000000..5efde5a7d6c9754aff7a8d846bc4081606c6502e
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium032.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium032",
+  "question": "Comparing the number of SZSE-listed private enterprises in the industry of Jinzhi Hongsheng Asset Management Co., Ltd. with the number of SSE-listed private enterprises in the industry of Biyuan Zhize Urban Development Co., Ltd., which is larger?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Jinzhi Hongsheng Asset Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Jinzhi Hongsheng Asset Management Co., Ltd. is Financial Industry.",
+    "Extracted from national_industry_status.csv that the number of SZSE-listed private enterprises in Financial Industry is 14.",
+    "Extracted from company_profile.csv that the industry of Biyuan Zhize Urban Development Co., Ltd. is Real Estate.",
+    "Extracted from national_industry_status.csv that the number of SSE-listed private enterprises in Real Estate is 12.",
+    "Compared 14 and 12; 14 is larger, so output \"Jinzhi Hongsheng Asset Management Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Jinzhi Hongsheng Asset Management Co., Ltd.": "Financial Industry",
+    "Number of SZSE-listed private enterprises in Financial Industry": 14,
+    "Industry of Biyuan Zhize Urban Development Co., Ltd.": "Real Estate",
+    "Number of SSE-listed private enterprises in Real Estate": 12,
+    "Conclusion (larger value)": "Jinzhi Hongsheng Asset Management Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium033.json b/assets/qa_gold/enterprise_industry_analysis/medium033.json
new file mode 100644
index 0000000000000000000000000000000000000000..674edb18b6911810393092885a338d65dadc437a
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium033.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium033",
+  "question": "Comparing the number of enterprises in Health and Social Work in the industry of Jianfan Ningze Elderly Care Services Co., Ltd. with the number of SSE-listed central state-owned enterprises in the industry of Zhongche Yuanze Shipbuilding Co., Ltd., which value is larger?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Jianfan Ningze Elderly Care Services Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Jianfan Ningze Elderly Care Services Co., Ltd. is Health and Social Work.",
+    "Extracted from national_industry_status.csv that the number of enterprises in Health and Social Work is 29.",
+    "Extracted from company_profile.csv that the industry of Zhongche Yuanze Shipbuilding Co., Ltd. is Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing.",
+    "Extracted from national_industry_status.csv that the number of SSE-listed central state-owned enterprises in this industry is 25.",
+    "Compared 29 and 25; 29 is larger, so output \"Jianfan Ningze Elderly Care Services Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Jianfan Ningze Elderly Care Services Co., Ltd.": "Health and Social Work",
+    "Number of enterprises in Health and Social Work": 29,
+    "Industry of Zhongche Yuanze Shipbuilding Co., Ltd.": "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Number of SSE-listed central state-owned enterprises in Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing": 25,
+    "Conclusion (larger value)": "Jianfan Ningze Elderly Care Services Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium034.json b/assets/qa_gold/enterprise_industry_analysis/medium034.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae107de4c0f3c5e682c6ef5f972dcef2e96c7424
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium034.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium034",
+  "question": "Is the minimum market capitalization in the industry of Jianfan Ningze Elderly Care Services Co., Ltd. lower than the minimum market capitalization in the industry of Zhongche Yuanze Shipbuilding Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Jianfan Ningze Elderly Care Services Co., Ltd. is Health and Social Work.",
+    "Extracted from national_industry_status.csv that the minimum market capitalization in Health and Social Work is 2.04 hundred million yuan.",
+    "Extracted from company_profile.csv that the industry of Zhongche Yuanze Shipbuilding Co., Ltd. is Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing.",
+    "Extracted from national_industry_status.csv that the minimum market capitalization in this industry is 6.2 hundred million yuan.",
+    "Judged whether 2.04 is lower than 6.2; the conclusion is \"Yes\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Jianfan Ningze Elderly Care Services Co., Ltd.": "Health and Social Work",
+    "Minimum market capitalization in Health and Social Work (hundred million yuan)": 2.04,
+    "Industry of Zhongche Yuanze Shipbuilding Co., Ltd.": "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Minimum market capitalization in Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing (hundred million yuan)": 6.2,
+    "Whether lower": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium035.json b/assets/qa_gold/enterprise_industry_analysis/medium035.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e64a31427b99a3104a6bc9de3bb651d5e4a1552
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium035.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium035",
+  "question": "Comparing the number of HKEX-listed sino-foreign joint ventures in the industry of Yihai Changjin Business Co., Ltd. with the number of SSE-listed local state-owned enterprises in the corresponding industry of Shanghai where Jinzhi Hongsheng Asset Management Co., Ltd. is located, which is larger?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Jinzhi Hongsheng Asset Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Yihai Changjin Business Co., Ltd. is Leasing and Business Services.",
+    "Extracted from national_industry_status.csv that the number of HKEX-listed sino-foreign joint ventures in Leasing and Business Services is 1.",
+    "Extracted from company_profile.csv that Jinzhi Hongsheng Asset Management Co., Ltd. is located in Shanghai Municipality.",
+    "Extracted from regional_industry_status.csv that the number of SSE-listed local state-owned enterprises in Shanghai Financial Industry is 8.",
+    "Compared 1 and 8; 8 is larger, so output \"Jinzhi Hongsheng Asset Management Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Yihai Changjin Business Co., Ltd.": "Leasing and Business Services",
+    "Number of HKEX-listed sino-foreign joint ventures in Leasing and Business Services": 1,
+    "Province/municipality where Jinzhi Hongsheng Asset Management Co., Ltd. is located": "Shanghai Municipality",
+    "Number of SSE-listed local state-owned enterprises in Shanghai Financial Industry": 8,
+    "Conclusion (larger value)": "Jinzhi Hongsheng Asset Management Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium036.json b/assets/qa_gold/enterprise_industry_analysis/medium036.json
new file mode 100644
index 0000000000000000000000000000000000000000..c11ea92225902d33e95ac8560fa6f66c8d53f6e0
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium036.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium036",
+  "question": "Is the mean annual number of Chinese invention patent applications in the industry of Yi Hai Chang Jin Shang Wu Co., Ltd. greater than the corresponding industry metric in Shanghai, where Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. is located?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Yi Hai Chang Jin Shang Wu Co., Ltd. = Leasing and Business Services",
+    "Extracted from national_industry_status.csv: mean annual Chinese invention patent applications in Leasing and Business Services = 9.27586206896552",
+    "Extracted from company_profile.csv: province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: mean annual Chinese invention patent applications in Shanghai Financial Industry = 35.2380952380952",
+    "Compared 9.27586206896552 and 35.2380952380952; since the former is smaller, the answer is \"No\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Mean annual Chinese invention patent applications in Leasing and Business Services": 9.27586206896552,
+    "Province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.": "Shanghai Municipality",
+    "Mean annual Chinese invention patent applications in Shanghai Financial Industry": 35.2380952380952,
+    "Whether greater": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium037.json b/assets/qa_gold/enterprise_industry_analysis/medium037.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca2984c413dfc38316690dcdb708b246bde981a9
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium037.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium037",
+  "question": "Which is greater: the number of SZSE-listed local state-owned enterprises in the industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd., or the number of HKEX-listed enterprises in the corresponding industry of Guangdong Province where Gao Yin Ze Tong Pi Fa Co., Ltd. is located?",
+  "guidelines": "The answer must be either \"Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. industry local state-owned enterprise SZSE-listed count\" or \"HKEX-listed enterprise count\". Output only the answer text without explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. industry local state-owned enterprise SZSE-listed count",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. = Information Transmission, Software and IT Services",
+    "Extracted from national_industry_status.csv: SZSE-listed local state-owned enterprise count in that industry = 26",
+    "Extracted from company_profile.csv: province of Gao Yin Ze Tong Pi Fa Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: HKEX-listed enterprise count in Guangdong Wholesale and Retail = 18",
+    "Compared 26 and 18; since 26 is greater, output the first allowed option"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.": "Information Transmission, Software and IT Services",
+    "SZSE-listed local state-owned enterprise count in that industry": 26,
+    "Province of Gao Yin Ze Tong Pi Fa Co., Ltd.": "Guangdong Province",
+    "HKEX-listed enterprise count in Guangdong Wholesale and Retail": 18,
+    "Comparison result (greater)": "Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. industry local state-owned enterprise SZSE-listed count"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium038.json b/assets/qa_gold/enterprise_industry_analysis/medium038.json
new file mode 100644
index 0000000000000000000000000000000000000000..93049db8758093eec4e2af08e2a3ab51c450fe95
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium038.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium038",
+  "question": "What is the difference between the maximum capitalized R&D expenditure of the industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. and the corresponding industry metric in Guangdong Province where Gao Yin Ze Tong Pi Fa Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to two decimal places. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 4026077173.89,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. = Information Transmission, Software and IT Services",
+    "Extracted from national_industry_status.csv: maximum capitalized R&D expenditure in that industry = 4055287084",
+    "Extracted from company_profile.csv: province of Gao Yin Ze Tong Pi Fa Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: maximum capitalized R&D expenditure in Guangdong Wholesale and Retail = 29209910.11",
+    "Calculated difference (national industry metric - provincial industry metric): 4055287084 - 29209910.11 = 4026077173.89"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Maximum capitalized R&D expenditure in that industry (yuan)": 4055287084,
+    "Province of Gao Yin Ze Tong Pi Fa Co., Ltd.": "Guangdong Province",
+    "Maximum capitalized R&D expenditure in Guangdong Wholesale and Retail (yuan)": 29209910.11,
+    "Difference": 4026077173.89
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium039.json b/assets/qa_gold/enterprise_industry_analysis/medium039.json
new file mode 100644
index 0000000000000000000000000000000000000000..3533f16979c4b9be12ffbb965e912aea450eb13e
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium039.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium039",
+  "question": "Is there any difference between the minimum State Technological Invention Award value in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. and the corresponding industry metric in Shanghai where Lang Ji Hui Ruan Technology Co., Ltd. is located?",
+  "guidelines": "The answer must be \"No difference\" or the other entity mentioned in the question. Output only the entity text without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No difference",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: minimum State Technological Invention Award value in Real Estate = 0",
+    "Extracted from company_profile.csv: province of Lang Ji Hui Ruan Technology Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: minimum State Technological Invention Award value in Shanghai Information Transmission, Software and IT Services = 0",
+    "Both values are 0, so output \"No difference\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "Minimum State Technological Invention Award value in Real Estate": 0,
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.": "Shanghai Municipality",
+    "Minimum State Technological Invention Award value in Shanghai Information Transmission, Software and IT Services": 0,
+    "Whether different": "No difference"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium040.json b/assets/qa_gold/enterprise_industry_analysis/medium040.json
new file mode 100644
index 0000000000000000000000000000000000000000..da222886d7e899a95aa1fd93af7d32b49b63f993
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium040.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium040",
+  "question": "Which is larger: the number of HKEX-listed foreign-funded enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd., or the number of SSE-listed enterprises in the corresponding industry of Shanghai where Lang Ji Hui Ruan Technology Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: HKEX-listed foreign-funded enterprise count in Real Estate = 34",
+    "Extracted from company_profile.csv: province of Lang Ji Hui Ruan Technology Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: SSE-listed enterprise count in Shanghai Information Transmission, Software and IT Services = 27",
+    "Compared 34 and 27; since 34 is larger, output \"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "HKEX-listed foreign-funded enterprise count in Real Estate": 34,
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.": "Shanghai Municipality",
+    "SSE-listed enterprise count in Shanghai Information Transmission, Software and IT Services": 27,
+    "Comparison result (larger)": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium041.json b/assets/qa_gold/enterprise_industry_analysis/medium041.json
new file mode 100644
index 0000000000000000000000000000000000000000..3b79ab54759d7390141fc205769117beb644a92a
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium041.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium041",
+  "question": "Which is larger: the number of SSE-listed local state-owned enterprises in the industry of Hua Xin Yuan Shi New Materials Co., Ltd., or the number of BSE-listed enterprises in the corresponding industry of Guangdong where Zhong Ke Ke Shu Software Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Hua Xin Yuan Shi New Materials Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Hua Xin Yuan Shi New Materials Co., Ltd. = Non-metallic Mineral Products",
+    "Extracted from national_industry_status.csv: SSE-listed local state-owned enterprise count in Non-metallic Mineral Products = 13",
+    "Extracted from company_profile.csv: province of Zhong Ke Ke Shu Software Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: BSE-listed enterprise count in Guangdong Information Transmission, Software and IT Services = 1",
+    "Compared 13 and 1; since 13 is larger, output \"Hua Xin Yuan Shi New Materials Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Hua Xin Yuan Shi New Materials Co., Ltd.": "Non-metallic Mineral Products",
+    "SSE-listed local state-owned enterprise count in Non-metallic Mineral Products": 13,
+    "Province of Zhong Ke Ke Shu Software Co., Ltd.": "Guangdong Province",
+    "BSE-listed enterprise count in Guangdong Information Transmission, Software and IT Services": 1,
+    "Comparison result (larger)": "Hua Xin Yuan Shi New Materials Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium042.json b/assets/qa_gold/enterprise_industry_analysis/medium042.json
new file mode 100644
index 0000000000000000000000000000000000000000..b1ddb0f31c1e01882b14b26edf85083c1750516b
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium042.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium042",
+  "question": "Comparing the minimum cumulative citation count of core patents in the industry of Huijin Jinrui Wealth Management Co., Ltd. with the same indicator in the corresponding industry of the province where the company is located, which value is larger?",
+  "guidelines": "The answer must be either \"minimum cumulative citation count of core patents in the industry of Huijin Jinrui Wealth Management Co., Ltd.\" or \"the same indicator in the corresponding industry of the province where the company is located\". Output only the province or region name without any explanation, analysis, or descriptive text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Huijin Jinrui Wealth Management Co., Ltd. is Financial Industry.",
+    "Extracted from national_industry_status.csv that the minimum cumulative citation count of core patents in Financial Industry is 0.",
+    "Extracted from company_profile.csv that the province where Huijin Jinrui Wealth Management Co., Ltd. is located is Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the minimum cumulative citation count of core patents in Guangdong Financial Industry is 0.",
+    "Both values are 0, so neither side is larger; output \"Equal\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Huijin Jinrui Wealth Management Co., Ltd.": "Financial Industry",
+    "Minimum cumulative citation count of core patents in Financial Industry": 0,
+    "Province where Huijin Jinrui Wealth Management Co., Ltd. is located": "Guangdong Province",
+    "Minimum cumulative citation count of core patents in Guangdong Financial Industry": 0,
+    "Comparison conclusion": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium043.json b/assets/qa_gold/enterprise_industry_analysis/medium043.json
new file mode 100644
index 0000000000000000000000000000000000000000..ddef344203beed0567d9b4d9214467fcb44671d7
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium043.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium043",
+  "question": "What is the difference between the number of SZSE-listed local state-owned enterprises in the industry of Huijin Jinrui Wealth Management Co., Ltd. and the number of SZSE-listed enterprises in the corresponding industry of the province where the company is located?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number without units, commas, or any explanatory text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 10.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Huijin Jinrui Wealth Management Co., Ltd. is Financial Industry.",
+    "Extracted from national_industry_status.csv that the number of SZSE-listed local state-owned enterprises in Financial Industry is 18.",
+    "Extracted from company_profile.csv that the province where Huijin Jinrui Wealth Management Co., Ltd. is located is Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the number of SZSE-listed enterprises in Guangdong Financial Industry is 8.",
+    "Calculated the difference (national-industry local-SOE SZSE count minus provincial-industry SZSE count): 18 - 8 = 10.0."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Huijin Jinrui Wealth Management Co., Ltd.": "Financial Industry",
+    "Number of SZSE-listed local state-owned enterprises in Financial Industry": 18,
+    "Province where Huijin Jinrui Wealth Management Co., Ltd. is located": "Guangdong Province",
+    "Number of SZSE-listed enterprises in Guangdong Financial Industry": 8,
+    "Difference": 10.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium044.json b/assets/qa_gold/enterprise_industry_analysis/medium044.json
new file mode 100644
index 0000000000000000000000000000000000000000..22bb80ee0b833daaaeea01422bf06f80668a64e8
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium044.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium044",
+  "question": "Which is higher: the number of SSE-listed enterprises in the corresponding industry of the province where Biyuan Shenghua Construction Development Co., Ltd. is located, or the number of HKEX-listed local state-owned enterprises in the industry of Huaying Taisheng Wealth Management Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Huaying Taisheng Wealth Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that Biyuan Shenghua Construction Development Co., Ltd. is located in Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the number of SSE-listed enterprises in Guangdong Real Estate is 7.",
+    "Extracted from company_profile.csv that the industry of Huaying Taisheng Wealth Management Co., Ltd. is Financial Industry.",
+    "Extracted from national_industry_status.csv that the number of HKEX-listed local state-owned enterprises in Financial Industry is 40.",
+    "Compared 7 and 40; 40 is higher, so output \"Huaying Taisheng Wealth Management Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province where Biyuan Shenghua Construction Development Co., Ltd. is located": "Guangdong Province",
+    "Number of SSE-listed enterprises in Guangdong Real Estate": 7,
+    "Industry of Huaying Taisheng Wealth Management Co., Ltd.": "Financial Industry",
+    "Number of HKEX-listed local state-owned enterprises in Financial Industry": 40,
+    "Comparison conclusion (higher value)": "Huaying Taisheng Wealth Management Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium045.json b/assets/qa_gold/enterprise_industry_analysis/medium045.json
new file mode 100644
index 0000000000000000000000000000000000000000..33375d72788f68582910ac8c3bd2e5b916d3f689
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium045.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium045",
+  "question": "Comparing the number of HKEX-listed enterprises in the province where Huatu Wenjiao Online Education Co., Ltd. is located with the number of SZSE-listed enterprises in the industry of Yihai Changjin Business Co., Ltd., which value is larger?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yihai Changjin Business Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that Huatu Wenjiao Online Education Co., Ltd. is located in Shanghai Municipality.",
+    "Extracted from regional_industry_status.csv that the number of HKEX-listed enterprises in Shanghai Education is 3.",
+    "Extracted from company_profile.csv that the industry of Yihai Changjin Business Co., Ltd. is Leasing and Business Services.",
+    "Extracted from national_industry_status.csv that the number of SZSE-listed enterprises in Leasing and Business Services is 44.",
+    "Compared 3 and 44; 44 is larger, so output \"Yihai Changjin Business Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province where Huatu Wenjiao Online Education Co., Ltd. is located": "Shanghai Municipality",
+    "Number of HKEX-listed enterprises in Shanghai Education": 3,
+    "Industry of Yihai Changjin Business Co., Ltd.": "Leasing and Business Services",
+    "Number of SZSE-listed enterprises in Leasing and Business Services": 44,
+    "Comparison conclusion (larger value)": "Yihai Changjin Business Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium046.json b/assets/qa_gold/enterprise_industry_analysis/medium046.json
new file mode 100644
index 0000000000000000000000000000000000000000..505c2512cd9d759286140fbe0c78b9709ec401ea
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium046.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium046",
+  "question": "Which is higher: the median operating profit of the corresponding industry in Shanghai, where Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. is located, or the median operating profit of the industry of Yi Hai Chang Jin Shang Wu Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: median operating profit in Shanghai Education = 129544365.28 Yuan",
+    "Extracted from company_profile.csv: industry of Yi Hai Chang Jin Shang Wu Co., Ltd. = Leasing and Business Services",
+    "Extracted from national_industry_status.csv: median operating profit in Leasing and Business Services = 11445832 Yuan",
+    "Compared the two values; since 129544365.28 is higher, output \"Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Education Operating profit (median)（Yuan）": 129544365.28,
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Median operating profit in Leasing and Business Services (Yuan)": 11445832,
+    "Comparison result (higher)": "Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium047.json b/assets/qa_gold/enterprise_industry_analysis/medium047.json
new file mode 100644
index 0000000000000000000000000000000000000000..81bc7dde017268b79640b7b00b7faa9dcbfc8b15
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium047.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium047",
+  "question": "Which is higher: the maximum cumulative citations of all patents in the corresponding industry in Shanghai, where Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. is located, or the same metric in the industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: maximum cumulative citations of all patents in Shanghai Financial Industry = 11161",
+    "Extracted from company_profile.csv: industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: maximum cumulative citations of all patents in Real Estate = 6946",
+    "Compared 11161 and 6946; since 11161 is higher, output \"Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Financial Industry Cumulative citations of all patents (maximum)": 11161,
+    "Industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.": "Real Estate",
+    "Maximum cumulative citations of all patents in Real Estate": 6946,
+    "Comparison result (higher)": "Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium048.json b/assets/qa_gold/enterprise_industry_analysis/medium048.json
new file mode 100644
index 0000000000000000000000000000000000000000..a775b8a7626e4c1931c6c3c4086837932166dc21
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium048.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium048",
+  "question": "Which is larger: the number of HKEX-listed central state-owned enterprises in the corresponding industry in Shanghai where Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. is located, or the number of SSE-listed local state-owned enterprises in the industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: HKEX-listed central state-owned enterprise count in Shanghai Financial Industry = 3",
+    "Extracted from company_profile.csv: industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: SSE-listed local state-owned enterprise count in Real Estate = 38",
+    "Compared 3 and 38; since 38 is larger, output \"Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Financial Industry Central state-owned enterprise_Number of HKEX-listed enterprises": 3,
+    "Industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.": "Real Estate",
+    "SSE-listed local state-owned enterprise count in Real Estate": 38,
+    "Comparison result (larger)": "Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium049.json b/assets/qa_gold/enterprise_industry_analysis/medium049.json
new file mode 100644
index 0000000000000000000000000000000000000000..3773befb082d009f81e07aeb31fdc0316a3329a2
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium049.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium049",
+  "question": "What is the difference between the total number of enterprises in the corresponding industry of Tianjin, where Bi Yuan Chan Jin Bu Dong Chan Co., Ltd. is located, and the number of SZSE-listed state-owned institute enterprises in the industry of Zhang Qiao Jin Chuang Technology Co., Ltd.?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number, without units, commas, or any text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 6.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Bi Yuan Chan Jin Bu Dong Chan Co., Ltd. = Tianjin Municipality",
+    "Extracted from regional_industry_status.csv: total enterprise count in Tianjin Real Estate = 7",
+    "Extracted from company_profile.csv: industry of Zhang Qiao Jin Chuang Technology Co., Ltd. = Consumer Electronics and Electrical Industry",
+    "Extracted from national_industry_status.csv: SZSE-listed state-owned institute enterprise count in Consumer Electronics and Electrical Industry = 1",
+    "Calculated difference: 7 - 1 = 6.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Bi Yuan Chan Jin Bu Dong Chan Co., Ltd.": "Tianjin Municipality",
+    "Total enterprise count in Tianjin Real Estate": 7,
+    "Industry of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Consumer Electronics and Electrical Industry",
+    "SZSE-listed state-owned institute enterprise count in Consumer Electronics and Electrical Industry": 1,
+    "difference": 6.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium050.json b/assets/qa_gold/enterprise_industry_analysis/medium050.json
new file mode 100644
index 0000000000000000000000000000000000000000..4dffafbfeb6e2939477c695281db5f5852e870e6
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium050.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium050",
+  "question": "What is the difference between the number of SZSE-listed local state-owned enterprises in the corresponding industry of Tianjin, where Bi Yuan Chan Jin Bu Dong Chan Co., Ltd. is located, and the number of HKEX-listed private enterprises in the industry of Zhang Qiao Jin Chuang Technology Co., Ltd.?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number, without units, commas, or any text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -43.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Bi Yuan Chan Jin Bu Dong Chan Co., Ltd. = Tianjin Municipality",
+    "Extracted from regional_industry_status.csv: SZSE-listed local state-owned enterprise count in Tianjin Real Estate = 2",
+    "Extracted from company_profile.csv: industry of Zhang Qiao Jin Chuang Technology Co., Ltd. = Consumer Electronics and Electrical Industry",
+    "Extracted from national_industry_status.csv: HKEX-listed private enterprise count in Consumer Electronics and Electrical Industry = 45",
+    "Calculated difference: 2 - 45 = -43.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Bi Yuan Chan Jin Bu Dong Chan Co., Ltd.": "Tianjin Municipality",
+    "SZSE-listed local state-owned enterprise count in Tianjin Real Estate": 2,
+    "Industry of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Consumer Electronics and Electrical Industry",
+    "HKEX-listed private enterprise count in Consumer Electronics and Electrical Industry": 45,
+    "difference": -43.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium051.json b/assets/qa_gold/enterprise_industry_analysis/medium051.json
new file mode 100644
index 0000000000000000000000000000000000000000..e2d1c29224bcb011464ab8cd94a8c37ec8c0bc70
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium051.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium051",
+  "question": "What is the gap between the number of SZSE-listed foreign-funded enterprises in the corresponding industry of Guangdong, where Gao Yin Ze Tong Pi Fa Co., Ltd. is located, and the number of SSE-listed enterprises in the industry of Lang Ji Hui Ruan Technology Co., Ltd.?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number, without units, commas, or any text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -139.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Gao Yin Ze Tong Pi Fa Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: SZSE-listed foreign-funded enterprise count in Guangdong Wholesale and Retail = 2",
+    "Extracted from company_profile.csv: industry of Lang Ji Hui Ruan Technology Co., Ltd. = Information Transmission, Software and IT Services",
+    "Extracted from national_industry_status.csv: SSE-listed enterprise count in Information Transmission, Software and IT Services = 141",
+    "Calculated difference: 2 - 141 = -139.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Gao Yin Ze Tong Pi Fa Co., Ltd.": "Guangdong Province",
+    "SZSE-listed foreign-funded enterprise count in Guangdong Wholesale and Retail": 2,
+    "Industry of Lang Ji Hui Ruan Technology Co., Ltd.": "Information Transmission, Software and IT Services",
+    "SSE-listed enterprise count in Information Transmission, Software and IT Services": 141,
+    "difference": -139.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium052.json b/assets/qa_gold/enterprise_industry_analysis/medium052.json
new file mode 100644
index 0000000000000000000000000000000000000000..054107be745eb4377a40e17738f066c9d2ecc2be
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium052.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium052",
+  "question": "Comparing the minimum value of State Natural Science Awards in the corresponding industry of the province where Gaoyin Zetong Wholesale Co., Ltd. is located with the minimum value of State Natural Science Awards in the industry of Langji Huiruan Technology Co., Ltd., which value is larger?",
+  "guidelines": "The answer must be a company name, \"industry\", or \"Equal\". Output only the name or \"Equal\" without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that Gaoyin Zetong Wholesale Co., Ltd. is located in Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the minimum value of State Natural Science Awards in Guangdong wholesale and retail industry is 0.",
+    "Extracted from company_profile.csv that the industry of Langji Huiruan Technology Co., Ltd. is Information Transmission, Software and Information Technology Services.",
+    "Extracted from national_industry_status.csv that the minimum value of State Natural Science Awards in this industry is 0.",
+    "Both values are 0, so neither side is larger; output \"Equal\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province where Gaoyin Zetong Wholesale Co., Ltd. is located": "Guangdong Province",
+    "Minimum value of State Natural Science Awards in Guangdong wholesale and retail industry": 0,
+    "Industry of Langji Huiruan Technology Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Minimum value of State Natural Science Awards in Information Transmission, Software and Information Technology Services": 0,
+    "Comparison conclusion": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium053.json b/assets/qa_gold/enterprise_industry_analysis/medium053.json
new file mode 100644
index 0000000000000000000000000000000000000000..25bb60b0f97b1f30ac4c642a900c8047cffdfa1e
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium053.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium053",
+  "question": "What is the difference between the number of HKEX-listed private enterprises in the corresponding industry of the province where Zhangqiao Jinchuang Technology Co., Ltd. is located and the number of HKEX-listed enterprises in the corresponding industry of the province where Zhaoye Huachang Real Estate Development Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -11.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that Zhangqiao Jinchuang Technology Co., Ltd. is located in Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the number of HKEX-listed private enterprises in Guangdong consumer electronics and electrical industry is 27.",
+    "Extracted from company_profile.csv that Zhaoye Huachang Real Estate Development Co., Ltd. is located in Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the number of HKEX-listed enterprises in Guangdong real estate industry is 38.",
+    "Calculated the difference (former minus latter): 27 - 38 = -11.0."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province where Zhangqiao Jinchuang Technology Co., Ltd. is located": "Guangdong Province",
+    "Number of HKEX-listed private enterprises in Guangdong consumer electronics and electrical industry": 27,
+    "Province where Zhaoye Huachang Real Estate Development Co., Ltd. is located": "Guangdong Province",
+    "Number of HKEX-listed enterprises in Guangdong real estate industry": 38,
+    "Difference": -11.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium054.json b/assets/qa_gold/enterprise_industry_analysis/medium054.json
new file mode 100644
index 0000000000000000000000000000000000000000..56edfc7cece1a252a245792ffcdb1d9eb3973f47
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium054.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium054",
+  "question": "Which is higher: the minimum R&D expenditure ratio in the corresponding industry of the province where Zhangqiao Jinchuang Technology Co., Ltd. is located, or the minimum R&D expenditure ratio in the corresponding industry of the province where Zhaoye Huachang Real Estate Development Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Zhangqiao Jinchuang Technology Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that Zhangqiao Jinchuang Technology Co., Ltd. is located in Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the minimum R&D expenditure ratio in Guangdong consumer electronics and electrical industry is 1.3 %.",
+    "Extracted from company_profile.csv that Zhaoye Huachang Real Estate Development Co., Ltd. is located in Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the minimum R&D expenditure ratio in Guangdong real estate industry is 0 %.",
+    "Compared 1.3 % and 0 %; 1.3 % is higher, so output \"Zhangqiao Jinchuang Technology Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province where Zhangqiao Jinchuang Technology Co., Ltd. is located": "Guangdong Province",
+    "Minimum R&D expenditure ratio in Guangdong consumer electronics and electrical industry": 1.3,
+    "Province where Zhaoye Huachang Real Estate Development Co., Ltd. is located": "Guangdong Province",
+    "Minimum R&D expenditure ratio in Guangdong real estate industry": 0,
+    "Comparison conclusion (higher value)": "Zhangqiao Jinchuang Technology Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium055.json b/assets/qa_gold/enterprise_industry_analysis/medium055.json
new file mode 100644
index 0000000000000000000000000000000000000000..e45fd63f5db5adb1c750ab8f35db178f888c6c18
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium055.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium055",
+  "question": "Between the median annual number of PCT invention patent applications for the corresponding industry in the province where Jinzhi Hongsheng Asset Management Company is located and that in the province where Zhonghai Gongchangjin Architectural Design Company is located, which is higher?",
+  "guidelines": "The answer must be either a company name or the word \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Jinzhi Hongsheng Asset Management Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Jinzhi Hongsheng Asset Management Company is located in Shanghai.",
+    "Extract from regional_industry_status.csv that the median annual number of PCT invention patent applications in Shanghai for the financial industry is 2.",
+    "Extract from company_profile.csv that Zhonghai Gongchangjin Architectural Design Company is located in Guangdong Province.",
+    "Extract from regional_industry_status.csv that the median annual number of PCT invention patent applications in Guangdong Province for the construction industry is 0.",
+    "Compare 2 and 0; 2 is higher, so output \"Jinzhi Hongsheng Asset Management Company\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Jinzhi Hongsheng Asset Management Company": "Shanghai",
+    "Median annual number of PCT invention patent applications in Shanghai for the financial industry": 2,
+    "Province of Zhonghai Gongchangjin Architectural Design Company": "Guangdong Province",
+    "Median annual number of PCT invention patent applications in Guangdong Province for the construction industry": 0,
+    "Comparison result (higher one)": "Jinzhi Hongsheng Asset Management Company"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium056.json b/assets/qa_gold/enterprise_industry_analysis/medium056.json
new file mode 100644
index 0000000000000000000000000000000000000000..49dd8ebe23fc5017e22367417d6353f722fae243
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium056.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium056",
+  "question": "Is the total liabilities (total) for the industry corresponding to the province where Jin Zhi Hong Sheng Zi Chan Management Co., Ltd. is located higher than the total liabilities (total) for the industry corresponding to the province where Zhong Hai Gong Chang Jin Jian Zhu She Ji Co., Ltd. is located?",
+  "guidelines": "The answer must be \"Yes\" or \"No\", output only one word, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that the province of Jin Zhi Hong Sheng Zi Chan Management Co., Ltd. is Shanghai Municipality",
+    "Extract from regional_industry_status.csv that Shanghai Municipality Financial Services total liabilities (total) is 44405835055827.7 Yuan",
+    "Extract from company_profile.csv that the province of Zhong Hai Gong Chang Jin Jian Zhu She Ji Co., Ltd. is Guangdong Province",
+    "Extract from regional_industry_status.csv that Guangdong Province Construction total liabilities (total) is 247639719094.03 Yuan",
+    "Determine whether the former is higher than the latter; conclusion is \"Yes\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Jin Zhi Hong Sheng Zi Chan Management Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Financial Services Total liabilities (total) (Yuan)": 44405835055827.7,
+    "Province of Zhong Hai Gong Chang Jin Jian Zhu She Ji Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Construction Total liabilities (total) (Yuan)": 247639719094.03,
+    "Whether higher than": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium057.json b/assets/qa_gold/enterprise_industry_analysis/medium057.json
new file mode 100644
index 0000000000000000000000000000000000000000..c55ceb9a294c33f127c97d61b19c28d437760147
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium057.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium057",
+  "question": "Is the maximum State Science and Technology Progress Award value in the corresponding industry of Shanghai, where Lang Ji Hui Ruan Technology Co., Ltd. is located, the same as the corresponding value in Beijing, where Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. is located?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Lang Ji Hui Ruan Technology Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: maximum State Science and Technology Progress Award in Shanghai Information Transmission, Software and IT Services = 0",
+    "Extracted from company_profile.csv: province of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. = Beijing Municipality",
+    "Extracted from regional_industry_status.csv: maximum State Science and Technology Progress Award in Beijing Health and Social Work = 0",
+    "Both values are 0, so the answer is \"Yes\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Information Transmission, Software and IT Services State Science and Technology Progress Award (maximum)": 0,
+    "Province of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd.": "Beijing Municipality",
+    "Beijing Municipality Health and Social Work State Science and Technology Progress Award (maximum)": 0,
+    "Whether identical": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium058.json b/assets/qa_gold/enterprise_industry_analysis/medium058.json
new file mode 100644
index 0000000000000000000000000000000000000000..8baa775560189e82d0ddf1aeac7c89e5ee0ceacd
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium058.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium058",
+  "question": "By how much is the mean capitalized R&D expenditure in the corresponding industry of Shanghai, where Lang Ji Hui Ruan Technology Co., Ltd. is located, higher than the corresponding value in Beijing, where Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. is located?",
+  "guidelines": "The answer must be an exact number, preserving all significant decimal places. Output only the number without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 85678136.92375,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Lang Ji Hui Ruan Technology Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: mean capitalized R&D expenditure in Shanghai Information Transmission, Software and IT Services = 85678136.92375 Yuan",
+    "Extracted from company_profile.csv: province of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. = Beijing Municipality",
+    "Extracted from regional_industry_status.csv: mean capitalized R&D expenditure in Beijing Health and Social Work = 0 Yuan",
+    "Calculated the amount higher: 85678136.92375 - 0 = 85678136.92375"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Information Transmission, Software and IT Services Capitalized R&D expenditure (mean) (Yuan)": 85678136.92375,
+    "Province of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd.": "Beijing Municipality",
+    "Beijing Municipality Health and Social Work Capitalized R&D expenditure (mean)（Yuan）": 0,
+    "Amount higher": 85678136.92375
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium059.json b/assets/qa_gold/enterprise_industry_analysis/medium059.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e07de8c431ea0b7dfa1090ad20b11271a94e282
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium059.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium059",
+  "question": "Which is larger: the HKEX-listed private enterprise count in the corresponding industry of Guangdong, where Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. is located, or the total HKEX-listed enterprise count in the corresponding industry of Guangdong, where Zhao Ye Ze Jin Di Chan Holdings Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: HKEX-listed private enterprise count in Guangdong Real Estate = 26",
+    "Extracted from company_profile.csv: province of Zhao Ye Ze Jin Di Chan Holdings Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: total HKEX-listed enterprise count in Guangdong Real Estate = 38",
+    "Compared 26 and 38; since 38 is larger, output \"Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Real Estate private enterprise HKEX-listed count": 26,
+    "Province of Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Real Estate HKEX-listed enterprise count": 38,
+    "Comparison result (larger)": "Zhao Ye Ze Jin Di Chan Holdings Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium060.json b/assets/qa_gold/enterprise_industry_analysis/medium060.json
new file mode 100644
index 0000000000000000000000000000000000000000..04df66cf1bf770f715b652c14660e8a9c87cccd7
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium060.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium060",
+  "question": "Are the minimum values of the Provincial or Ministerial Science and Technology Progress Award metric the same between the corresponding industries in Guangdong for Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. and Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: minimum Provincial or Ministerial Science and Technology Progress Award value in Guangdong Real Estate = 0",
+    "Extracted from company_profile.csv: province of Zhao Ye Ze Jin Di Chan Holdings Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: minimum Provincial or Ministerial Science and Technology Progress Award value in Guangdong Real Estate = 0",
+    "Both values are 0, so the answer is \"Yes\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.": "Guangdong Province",
+    "Minimum Provincial or Ministerial Science and Technology Progress Award value in Guangdong Real Estate": 0,
+    "Province of Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.": "Guangdong Province",
+    "Minimum Provincial or Ministerial Science and Technology Progress Award value in Guangdong Real Estate (comparison side)": 0,
+    "Whether identical": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium061.json b/assets/qa_gold/enterprise_industry_analysis/medium061.json
new file mode 100644
index 0000000000000000000000000000000000000000..a8900f796b61bb76e814be177a113ffd4cdb42e0
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium061.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium061",
+  "question": "Are the total State Natural Science Award values the same between the corresponding industry in Hong Kong, where Rui Xing Jian Kang Zhi Yao Co., Ltd. is located, and the corresponding industry in Zhejiang, where Wu Li Hui Da Chain Co., Ltd. is located?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Rui Xing Jian Kang Zhi Yao Co., Ltd. = Hong Kong SAR",
+    "Extracted from regional_industry_status.csv: total State Natural Science Award value in Hong Kong Pharmaceutical Manufacturing = 0",
+    "Extracted from company_profile.csv: province of Wu Li Hui Da Chain Co., Ltd. = Zhejiang Province",
+    "Extracted from regional_industry_status.csv: total State Natural Science Award value in Zhejiang Wholesale and Retail = 0",
+    "Both values are 0, so the answer is \"Yes\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province of Rui Xing Jian Kang Zhi Yao Co., Ltd.": "Hong Kong SAR",
+    "Total State Natural Science Award value in Hong Kong Pharmaceutical Manufacturing": 0,
+    "Province of Wu Li Hui Da Chain Co., Ltd.": "Zhejiang Province",
+    "Total State Natural Science Award value in Zhejiang Wholesale and Retail": 0,
+    "Whether identical": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium062.json b/assets/qa_gold/enterprise_industry_analysis/medium062.json
new file mode 100644
index 0000000000000000000000000000000000000000..046c749d5c6510e4394d920d4c70ef4b63da571a
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium062.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium062",
+  "question": "Rui Xing Jian Kang Zhi Yao Co., Ltd.industry in its province R&D headcount YoY change (minimum) and Wu Li Hui Da Chain Co., Ltd.industry in its province R&D headcount YoY change (minimum)compared with difference how much？",
+  "guidelines": "The answer must units, .Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "answer": -19.19,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Rui Xing Jian Kang Zhi Yao Co., Ltd.province = Hong Kong SAR, industry=",
+    "regional_industry_status.csv in extractHong Kong SAR- R&D headcount YoY change (minimum)= -33.3 %",
+    "Extracted from company_profile.csv: Wu Li Hui Da Chain Co., Ltd.province = Zhejiang Province, industry=Retail",
+    "regional_industry_status.csv in extractZhejiang Province-Retail R&D headcount YoY change (minimum)= -14.11 %",
+    "Calculate difference: -33.3 - (-14.11) = -19.19"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Rui Xing Jian Kang Zhi Yao Co., Ltd.province": "Hong Kong SAR",
+    "Hong Kong SAR  R&D headcount YoY change (minimum)": -33.3,
+    "Wu Li Hui Da Chain Co., Ltd.province": "Zhejiang Province",
+    "Zhejiang Province Retail R&D headcount YoY change (minimum)": -14.11,
+    "difference": -19.19
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium063.json b/assets/qa_gold/enterprise_industry_analysis/medium063.json
new file mode 100644
index 0000000000000000000000000000000000000000..5eed2e44b0c66fbe3b510b1722831d500e888f37
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium063.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium063",
+  "question": "What is the difference between the number of central state-owned enterprises listed on the Shenzhen Stock Exchange in the corresponding industry of the province where Baoxin Huihui Network Company is located and the number of central state-owned enterprises in the metal smelting and rolling industry listed on the Hong Kong Stock Exchange?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": -2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Baoxin Huihui Network Company is located in Beijing Municipality and its corresponding industry is Information Transmission, Software and IT Services.",
+    "Extract from regional_industry_status.csv that the number of central state-owned enterprises in Beijing Municipality – Information Transmission, Software and IT Services that are listed on the Shenzhen Stock Exchange is 4.",
+    "Extract from national_industry_status.csv that the number of central state-owned enterprises in the metal smelting and rolling processing industry that are listed on the Hong Kong Stock Exchange is 6.",
+    "Calculate the difference: 4 - 6 = -2.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Baoxin Huihui Network Company": "Beijing Municipality",
+    "Number of central state-owned enterprises listed on the Shenzhen Stock Exchange in Beijing Municipality – Information Transmission, Software and IT Services": 4,
+    "Number of central state-owned enterprises listed on the Hong Kong Stock Exchange in the metal smelting and rolling processing industry": 6,
+    "Difference": -2.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium064.json b/assets/qa_gold/enterprise_industry_analysis/medium064.json
new file mode 100644
index 0000000000000000000000000000000000000000..44930fd6fa85c8356f5688590781e4fd920d16ab
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium064.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium064",
+  "question": "Between the median R&D expenditure ratio of the corresponding industry in the province where Baoxin Huihui Network Company is located and that of the metal smelting and rolling processing industry, which is higher?",
+  "guidelines": "The answer must be either a company name or the word \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Baoxin Huihui Network Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Baoxin Huihui Network Company is located in Beijing Municipality and its corresponding industry is Information Transmission, Software and IT Services.",
+    "Extract from regional_industry_status.csv that the median R&D expenditure ratio for Information Transmission, Software and IT Services in Beijing Municipality is 12.16 %.",
+    "Extract from national_industry_status.csv that the median R&D expenditure ratio for the metal smelting and rolling processing industry is 3.03 %.",
+    "Compare the two medians: 12.16 % > 3.03 %, so the corresponding industry in the province where Baoxin Huihui Network Company is located is higher."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Baoxin Huihui Network Company": "Beijing Municipality",
+    "Median R&D expenditure ratio for Information Transmission, Software and IT Services in Beijing Municipality": 12.16,
+    "Median R&D expenditure ratio for the metal smelting and rolling processing industry": 3.03,
+    "Comparison result (higher one)": "Baoxin Huihui Network Company"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium065.json b/assets/qa_gold/enterprise_industry_analysis/medium065.json
new file mode 100644
index 0000000000000000000000000000000000000000..9af6d417ec3f80ca401ef0652f9146acc2823a91
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium065.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium065",
+  "question": "Between the number of Shanghai Stock Exchange-listed private enterprises in the corresponding industry of the province where Beikong Zejing Water Company is located and the number of Beijing Stock Exchange-listed enterprises in the Information Transmission, Software and IT Services industry, which is larger?",
+  "guidelines": "The answer must be either \"the number of SSE-listed private enterprises in the corresponding industry of the province where Beikong Zejing Water Company is located\" or \"the number of BSE-listed enterprises in the Information Transmission, Software and IT Services industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "the number of BSE-listed enterprises in the Information Transmission, Software and IT Services industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Beikong Zejing Water Company is located in Guangdong Province and its corresponding industry is Water Conservancy, Environment and Public Facilities Management.",
+    "Extract from regional_industry_status.csv that the number of SSE-listed private enterprises in Guangdong Province – Water Conservancy, Environment and Public Facilities Management is 1.",
+    "Extract from national_industry_status.csv that the number of BSE-listed enterprises in the Information Transmission, Software and IT Services industry is 17.",
+    "Compare the counts: 17 > 1, so the number of BSE-listed enterprises in the Information Transmission, Software and IT Services industry is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Beikong Zejing Water Company": "Guangdong Province",
+    "Number of SSE-listed private enterprises in Guangdong Province – Water Conservancy, Environment and Public Facilities Management": 1,
+    "Number of BSE-listed enterprises in the Information Transmission, Software and IT Services industry": 17,
+    "Comparison result (larger one)": "the number of BSE-listed enterprises in the Information Transmission, Software and IT Services industry"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium066.json b/assets/qa_gold/enterprise_industry_analysis/medium066.json
new file mode 100644
index 0000000000000000000000000000000000000000..11c5244f8345643645fbd6ecf049298d59d2f705
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium066.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium066",
+  "question": "Which is larger: the total assets of the corresponding industry in the province where Beikong Zejing Water Company is located, or the total assets of the Information Transmission, Software and IT Services industry?",
+  "guidelines": "The answer must be either \"the total assets of the corresponding industry in the province where Beikong Zejing Water Company is located\" or \"the Information Transmission, Software and IT Services industry\". Output only the name, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "the Information Transmission, Software and IT Services industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Beikong Zejing Water Company is located in Guangdong Province and its corresponding industry is Water Conservancy, Environment and Public Facilities Management.",
+    "Extract from regional_industry_status.csv that the total assets of Water Conservancy, Environment and Public Facilities Management in Guangdong Province are 194502203663.44 Yuan.",
+    "Extract from national_industry_status.csv that the total assets of the Information Transmission, Software and IT Services industry are 18848109950318 Yuan.",
+    "Compare the total assets: 18848109950318 > 194502203663.44, so the Information Transmission, Software and IT Services industry is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Beikong Zejing Water Company": "Guangdong Province",
+    "Total assets of Water Conservancy, Environment and Public Facilities Management in Guangdong Province (Yuan)": 194502203663.44,
+    "Total assets of the Information Transmission, Software and IT Services industry (Yuan)": 18848109950318,
+    "Comparison result (larger one)": "the Information Transmission, Software and IT Services industry"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium067.json b/assets/qa_gold/enterprise_industry_analysis/medium067.json
new file mode 100644
index 0000000000000000000000000000000000000000..684d2ba2ec6a312fad30bfe41f76d121bbdb0363
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium067.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium067",
+  "question": "Between the number of HKEX-listed foreign-funded enterprises in the corresponding industry of the province where Aijian Yikang Fuzhongxin Company is located and the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry, which value is larger?",
+  "guidelines": "The answer must be either \"the number of HKEX-listed foreign-funded enterprises in the corresponding industry of the province where Aijian Yikang Fuzhongxin Company is located\" or \"the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Aijian Yikang Fuzhongxin Company is located in Beijing Municipality and its corresponding industry is Health and Social Work.",
+    "Extract from regional_industry_status.csv that the number of HKEX-listed foreign-funded enterprises in Beijing Municipality – Health and Social Work is 2.",
+    "Extract from national_industry_status.csv that the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry is 7.",
+    "Compare the counts: 7 > 2, so the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Aijian Yikang Fuzhongxin Company": "Beijing Municipality",
+    "Number of HKEX-listed foreign-funded enterprises in Beijing Municipality – Health and Social Work": 2,
+    "Number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry": 7,
+    "Comparison result (larger one)": "the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium068.json b/assets/qa_gold/enterprise_industry_analysis/medium068.json
new file mode 100644
index 0000000000000000000000000000000000000000..f6c80ee638f7ad579c45650776e13d2cefb3cbd4
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium068.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium068",
+  "question": "What is the difference between the maximum annual number of China patent applications for the corresponding industry in the province where Aijian Yikang Fuzhongxin Company is located and the maximum annual number of China patent applications for the pharmaceutical manufacturing industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": -329.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Aijian Yikang Fuzhongxin Company is located in Beijing Municipality and its corresponding industry is Health and Social Work.",
+    "Extract from regional_industry_status.csv that the maximum annual number of China patent applications in Beijing Municipality - Health and Social Work is 0.",
+    "Extract from national_industry_status.csv that the maximum annual number of China patent applications in the pharmaceutical manufacturing industry is 329.",
+    "Calculate the difference: 0 - 329 = -329.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Aijian Yikang Fuzhongxin Company": "Beijing Municipality",
+    "Maximum annual number of China patent applications in Beijing Municipality - Health and Social Work": 0,
+    "Maximum annual number of China patent applications in the pharmaceutical manufacturing industry": 329,
+    "Difference": -329.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium069.json b/assets/qa_gold/enterprise_industry_analysis/medium069.json
new file mode 100644
index 0000000000000000000000000000000000000000..57886502a1347a05da1b0097c2f20cfc02d46808
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium069.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium069",
+  "question": "Between the maximum capitalized R&D expenditure of the corresponding industry in the province where Zhongche Yuanze Shipbuilding Company is located and the maximum capitalized R&D expenditure of the Electricity, Heat, Gas and Water Production and Supply industry, which is higher?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhongche Yuanze Shipbuilding Company is located in Beijing Municipality and its corresponding industry is Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing.",
+    "Extract from regional_industry_status.csv that the maximum capitalized R&D expenditure in Beijing Municipality - Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing is 353771542.85 Yuan.",
+    "Extract from national_industry_status.csv that the maximum capitalized R&D expenditure in the Electricity, Heat, Gas and Water Production and Supply industry is 1982740987.15 Yuan.",
+    "Compare the two maximum values: 1982740987.15 Yuan > 353771542.85 Yuan, so the industry is higher."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Zhongche Yuanze Shipbuilding Company": "Beijing Municipality",
+    "Maximum capitalized R&D expenditure in Beijing Municipality - Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing (Yuan)": 353771542.85,
+    "Maximum capitalized R&D expenditure in the Electricity, Heat, Gas and Water Production and Supply industry (Yuan)": 1982740987.15,
+    "Comparison result (higher one)": "industry"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium070.json b/assets/qa_gold/enterprise_industry_analysis/medium070.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ccb63ce83db89cf5baf2ce4b66698577f2aeda7
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium070.json
@@ -0,0 +1,23 @@
+{
+  "id": "medium070",
+  "question": "What is the difference between the number of local state-owned enterprises listed on the Shanghai Stock Exchange in the corresponding industry in Beijing Municipality and the number of private enterprises listed on the Shanghai Stock Exchange in the Electricity, Heat, Gas and Water Production and Supply industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": -18.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhongche Yuanze Shipbuilding Company is located in Beijing Municipality and its corresponding industry is Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing.",
+    "Extract from regional_industry_status.csv that the number of SSE-listed local state-owned enterprises in Beijing Municipality - Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing is 1.",
+    "Extract from national_industry_status.csv that the number of SSE-listed private enterprises in the Electricity, Heat, Gas and Water Production and Supply industry is 19.",
+    "Calculate the difference: 1 - 19 = -18.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Number of SSE-listed local state-owned enterprises in Beijing Municipality - Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing": 1,
+    "Number of SSE-listed private enterprises in the Electricity, Heat, Gas and Water Production and Supply industry": 19,
+    "Difference": -18.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium071.json b/assets/qa_gold/enterprise_industry_analysis/medium071.json
new file mode 100644
index 0000000000000000000000000000000000000000..551063364e6aaaa212861e6d4a8252f3c0978b33
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium071.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium071",
+  "question": "Which is larger: the number of SSE-listed enterprises in the province where Biyuan Chanjin Real Estate Company is located, or the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry?",
+  "guidelines": "The answer must be either \"the number of SSE-listed enterprises in the province where Biyuan Chanjin Real Estate Company is located\" or \"the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Biyuan Chanjin Real Estate Company is located in Tianjin Municipality and its corresponding industry is Real Estate.",
+    "Extract from regional_industry_status.csv that the number of SSE-listed enterprises in Tianjin Municipality - Real Estate is 3.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry is 15.",
+    "Compare the counts: 15 > 3, so the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Biyuan Chanjin Real Estate Company": "Tianjin Municipality",
+    "Number of SSE-listed enterprises in Tianjin Municipality - Real Estate": 3,
+    "Number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry": 15,
+    "Comparison result (larger one)": "the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium072.json b/assets/qa_gold/enterprise_industry_analysis/medium072.json
new file mode 100644
index 0000000000000000000000000000000000000000..be4f5b1413556734cc4304749b6674f45fdc2027
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium072.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium072",
+  "question": "Which value is larger: the total number of real estate enterprises in the province where Biyuan Chanjin Real Estate Company is located, or the number of SZSE-listed central state-owned enterprises in the Transportation, Storage and Postal Services industry?",
+  "guidelines": "The answer must be either \"the total number of real estate enterprises in the province where Biyuan Chanjin Real Estate Company is located\" or \"the number of SZSE-listed central state-owned enterprises in the Transportation, Storage and Postal Services industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "the total number of real estate enterprises in the province where Biyuan Chanjin Real Estate Company is located",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Biyuan Chanjin Real Estate Company is located in Tianjin Municipality and its corresponding industry is Real Estate.",
+    "Extract from regional_industry_status.csv that the total number of enterprises in Tianjin Municipality - Real Estate is 7.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed central state-owned enterprises in the Transportation, Storage and Postal Services industry is 5.",
+    "Compare the values: 7 > 5, so the total number of real estate enterprises in the province where Biyuan Chanjin Real Estate Company is located is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Biyuan Chanjin Real Estate Company": "Tianjin Municipality",
+    "Total number of enterprises in Tianjin Municipality - Real Estate": 7,
+    "Number of SZSE-listed central state-owned enterprises in the Transportation, Storage and Postal Services industry": 5,
+    "Comparison result (larger one)": "the total number of real estate enterprises in the province where Biyuan Chanjin Real Estate Company is located"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium073.json b/assets/qa_gold/enterprise_industry_analysis/medium073.json
new file mode 100644
index 0000000000000000000000000000000000000000..33f62b717ec0e88c700ce1a00fb32da693ef2579
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium073.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium073",
+  "question": "What is the difference between the number of SZSE-listed foreign-funded enterprises in the industry where Zhongke Keshu Software Company operates and the number of BSE-listed private enterprises in the construction industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": 2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhongke Keshu Software Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed foreign-funded enterprises in Information Transmission, Software and IT Services is 4.",
+    "Extract from national_industry_status.csv that the number of BSE-listed private enterprises in the construction industry is 2.",
+    "Calculate the difference: 4 - 2 = 2.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Zhongke Keshu Software Company": "Information Transmission, Software and IT Services",
+    "Number of SZSE-listed foreign-funded enterprises in Information Transmission, Software and IT Services": 4,
+    "Number of BSE-listed private enterprises in the construction industry": 2,
+    "Difference": 2.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium074.json b/assets/qa_gold/enterprise_industry_analysis/medium074.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0448d0e5f5f3e34f447d849b99875cedf1228a4
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium074.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium074",
+  "question": "What is the difference between the number of HKEX-listed local state-owned enterprises in the industry where Zhongke Keshu Software Company operates and the number of HKEX-listed foreign-funded enterprises in the construction industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": 3.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhongke Keshu Software Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed local state-owned enterprises in Information Transmission, Software and IT Services is 9.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed foreign-funded enterprises in the construction industry is 6.",
+    "Calculate the difference: 9 - 6 = 3.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Zhongke Keshu Software Company": "Information Transmission, Software and IT Services",
+    "Number of HKEX-listed local state-owned enterprises in Information Transmission, Software and IT Services": 9,
+    "Number of HKEX-listed foreign-funded enterprises in the construction industry": 6,
+    "Difference": 3.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium075.json b/assets/qa_gold/enterprise_industry_analysis/medium075.json
new file mode 100644
index 0000000000000000000000000000000000000000..d2b0d44b94fb0afccc05eac696728fe1db0b7bec
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium075.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium075",
+  "question": "What is the difference between the number of SSE-listed foreign-funded enterprises in the industry where Hengli Kezhi Software Company operates and the number of SSE-listed private enterprises in the conglomerates industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": 2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hengli Kezhi Software Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the number of SSE-listed foreign-funded enterprises in Information Transmission, Software and IT Services is 7.",
+    "Extract from national_industry_status.csv that the number of SSE-listed private enterprises in the conglomerates industry is 5.",
+    "Calculate the difference: 7 - 5 = 2.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Hengli Kezhi Software Company": "Information Transmission, Software and IT Services",
+    "Number of SSE-listed foreign-funded enterprises in Information Transmission, Software and IT Services": 7,
+    "Number of SSE-listed private enterprises in the conglomerates industry": 5,
+    "Difference": 2.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium076.json b/assets/qa_gold/enterprise_industry_analysis/medium076.json
new file mode 100644
index 0000000000000000000000000000000000000000..9f9700c77ac843e74c7214206dabed975dd1509b
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium076.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium076",
+  "question": "Between the median number of industry standards participated in drafting in the industry where Hengli Kezhi Software Company operates and that of the conglomerates industry, which value is larger?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Hengli Kezhi Software Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hengli Kezhi Software Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the median number of industry standards participated in drafting for Information Transmission, Software and IT Services is 2.",
+    "Extract from national_industry_status.csv that the median number of industry standards participated in drafting for the conglomerates industry is 0.",
+    "Compare the two medians: 2 > 0, so the industry where Hengli Kezhi Software Company operates is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Hengli Kezhi Software Company": "Information Transmission, Software and IT Services",
+    "Median number of industry standards participated in drafting for Information Transmission, Software and IT Services": 2,
+    "Median number of industry standards participated in drafting for the conglomerates industry": 0,
+    "Comparison result (larger one)": "Hengli Kezhi Software Company"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium077.json b/assets/qa_gold/enterprise_industry_analysis/medium077.json
new file mode 100644
index 0000000000000000000000000000000000000000..d213a48e85ba07308c024523ed4e80cceb9c7959
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium077.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium077",
+  "question": "Which is greater: the number of BSE-listed enterprises in the industry of Lang Ji Hui Ruan Technology Co., Ltd., or the number of enterprises in chemical raw materials and chemical products manufacturing?",
+  "guidelines": "The answer must be either \"Number of BSE-listed enterprises in the industry of Lang Ji Hui Ruan Technology Co., Ltd.\" or \"Number of enterprises in chemical raw materials and chemical products manufacturing\"; output only the designated answer text, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Chemical Raw Materials and Chemical Products Manufacturing",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Lang Ji Hui Ruan Technology Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "From national_industry_status.csv, extract that the number of BSE-listed enterprises for Information Transmission, Software and IT Services is 17",
+    "From national_industry_status.csv, extract that the number of enterprises in chemical raw materials and chemical products manufacturing is 364",
+    "Compare counts: 364 > 17; therefore the number of enterprises in chemical raw materials and chemical products manufacturing is greater"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Lang Ji Hui Ruan Technology Co., Ltd. industry": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services Number of BSE-listed enterprises": 17,
+    "Chemical Raw Materials and Chemical Products Manufacturing Number of enterprises": 364,
+    "Comparison conclusion (greater)": "Chemical Raw Materials and Chemical Products Manufacturing"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium078.json b/assets/qa_gold/enterprise_industry_analysis/medium078.json
new file mode 100644
index 0000000000000000000000000000000000000000..954fce5ceea8bfe00dbdbb6ab39d9f04340183ea
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium078.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium078",
+  "question": "Between the mean change in R&D expenditure ratio for the industry where Langji Huiruan Technology Company operates and the same indicator for the Chemical Raw Materials and Chemical Products Manufacturing industry, which is higher?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Langji Huiruan Technology Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Langji Huiruan Technology Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the mean change in R&D expenditure ratio for Information Transmission, Software and IT Services is 1.98752122241087 %.",
+    "Extract from national_industry_status.csv that the mean change in R&D expenditure ratio for Chemical Raw Materials and Chemical Products Manufacturing is -0.0206470588235294 %.",
+    "Compare the means: 1.98752122241087 % > -0.0206470588235294 %, so the industry where Langji Huiruan Technology Company operates is higher."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Langji Huiruan Technology Company": "Information Transmission, Software and IT Services",
+    "Mean change in R&D expenditure ratio for Information Transmission, Software and IT Services": 1.98752122241087,
+    "Mean change in R&D expenditure ratio for Chemical Raw Materials and Chemical Products Manufacturing": -0.0206470588235294,
+    "Comparison result (higher one)": "Langji Huiruan Technology Company"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium079.json b/assets/qa_gold/enterprise_industry_analysis/medium079.json
new file mode 100644
index 0000000000000000000000000000000000000000..e86504f57e5b9744145162a97ccb9f6a81796d9b
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium079.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium079",
+  "question": "What is the difference between the number of SZSE-listed central state-owned enterprises in the industry where Huijin Jinrui Wealth Management Company operates and the number of HKEX-listed central state-owned enterprises in the Electrical Machinery and Equipment Manufacturing industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": 2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Huijin Jinrui Wealth Management Company belongs to the financial industry.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed central state-owned enterprises in the financial industry is 5.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed central state-owned enterprises in the Electrical Machinery and Equipment Manufacturing industry is 3.",
+    "Calculate the difference: 5 - 3 = 2.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Huijin Jinrui Wealth Management Company": "Financial industry",
+    "Number of SZSE-listed central state-owned enterprises in the financial industry": 5,
+    "Number of HKEX-listed central state-owned enterprises in the Electrical Machinery and Equipment Manufacturing industry": 3,
+    "Difference": 2.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium080.json b/assets/qa_gold/enterprise_industry_analysis/medium080.json
new file mode 100644
index 0000000000000000000000000000000000000000..95d2dd215ca82233e082ba374c9614bcc98b69b3
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium080.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium080",
+  "question": "Between the median net profit amount of Huijin Jinrui Wealth Management Company and that of the Electrical Machinery and Equipment Manufacturing industry, which is higher?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Huijin Jinrui Wealth Management Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Huijin Jinrui Wealth Management Company belongs to the financial industry.",
+    "Extract from national_industry_status.csv that the median net profit amount of the financial industry is 837989445.74 Yuan.",
+    "Extract from national_industry_status.csv that the median net profit amount of the Electrical Machinery and Equipment Manufacturing industry is 117943061.895 Yuan.",
+    "Compare the two medians: 837989445.74 Yuan > 117943061.895 Yuan, so Huijin Jinrui Wealth Management Company is higher."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Huijin Jinrui Wealth Management Company": "Financial industry",
+    "Median net profit amount of the financial industry (Yuan)": 837989445.74,
+    "Median net profit amount of the Electrical Machinery and Equipment Manufacturing industry (Yuan)": 117943061.895,
+    "Comparison result (higher one)": "Huijin Jinrui Wealth Management Company"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium081.json b/assets/qa_gold/enterprise_industry_analysis/medium081.json
new file mode 100644
index 0000000000000000000000000000000000000000..1291c0798e499ce3c4098d3e7230f60cd8edf3e2
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium081.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium081",
+  "question": "What is the difference between the number of SZSE-listed enterprises in the industry where Zhaoye Huachang Real Estate Development Company operates and the number of SZSE-listed private enterprises in the General Equipment Manufacturing industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": -35.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhaoye Huachang Real Estate Development Company belongs to the Real Estate industry.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed enterprises in the Real Estate industry is 51.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed private enterprises in the General Equipment Manufacturing industry is 86.",
+    "Calculate the difference: 51 - 86 = -35.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Zhaoye Huachang Real Estate Development Company": "Real Estate",
+    "Number of SZSE-listed enterprises in the Real Estate industry": 51,
+    "Number of SZSE-listed private enterprises in the General Equipment Manufacturing industry": 86,
+    "Difference": -35.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium082.json b/assets/qa_gold/enterprise_industry_analysis/medium082.json
new file mode 100644
index 0000000000000000000000000000000000000000..e96f40647636aeb279ba45c6ca0e8eb53f33d7ad
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium082.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium082",
+  "question": "Between the maximum operating revenue amount of the industry where Zhaoye Huachang Real Estate Development Company operates and that of the General Equipment Manufacturing industry, which is higher?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Zhaoye Huachang Real Estate Development Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhaoye Huachang Real Estate Development Company belongs to the Real Estate industry.",
+    "Extract from national_industry_status.csv that the maximum operating revenue amount of the Real Estate industry is 503838390573.74 Yuan.",
+    "Extract from national_industry_status.csv that the maximum operating revenue amount of the General Equipment Manufacturing industry is 117623139663 Yuan.",
+    "Compare the two maximum values: 503838390573.74 Yuan > 117623139663 Yuan, so Zhaoye Huachang Real Estate Development Company is higher."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Zhaoye Huachang Real Estate Development Company": "Real Estate",
+    "Maximum operating revenue amount of the Real Estate industry (Yuan)": 503838390573.74,
+    "Maximum operating revenue amount of the General Equipment Manufacturing industry (Yuan)": 117623139663,
+    "Comparison result (higher one)": "Zhaoye Huachang Real Estate Development Company"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium083.json b/assets/qa_gold/enterprise_industry_analysis/medium083.json
new file mode 100644
index 0000000000000000000000000000000000000000..ade631200a139f540a6d6a3e85c17c014f4f7661
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium083.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium083",
+  "question": "What is the difference between the number of SZSE-listed enterprises in the industry where Yihai Changjin Business Company operates and the number of SZSE-listed central state-owned enterprises in the Communication Transmission Equipment industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": 41.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Yihai Changjin Business Company belongs to the Leasing and Business Services industry.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed enterprises in the Leasing and Business Services industry is 44.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed central state-owned enterprises in the Communication Transmission Equipment industry is 3.",
+    "Calculate the difference: 44 - 3 = 41.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Yihai Changjin Business Company": "Leasing and Business Services",
+    "Number of SZSE-listed enterprises in the Leasing and Business Services industry": 44,
+    "Number of SZSE-listed central state-owned enterprises in the Communication Transmission Equipment industry": 3,
+    "Difference": 41.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium084.json b/assets/qa_gold/enterprise_industry_analysis/medium084.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac46237a8707557ef5bc4e134d7dc24502523f97
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium084.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium084",
+  "question": "Which is larger: the number of SSE-listed central state-owned enterprises in the industry where Yihai Changjin Business Company operates, or the total number of enterprises in the Communication Transmission Equipment industry?",
+  "guidelines": "The answer must be either \"the number of SSE-listed central state-owned enterprises in the industry where Yihai Changjin Business Company operates\" or \"the total number of enterprises in the Communication Transmission Equipment industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "the total number of enterprises in the Communication Transmission Equipment industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Yihai Changjin Business Company belongs to the Leasing and Business Services industry.",
+    "Extract from national_industry_status.csv that the number of SSE-listed central state-owned enterprises in the Leasing and Business Services industry is 3.",
+    "Extract from national_industry_status.csv that the total number of enterprises in the Communication Transmission Equipment industry is 120.",
+    "Compare the counts: 120 > 3, so the total number of enterprises in the Communication Transmission Equipment industry is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Yihai Changjin Business Company": "Leasing and Business Services",
+    "Number of SSE-listed central state-owned enterprises in the Leasing and Business Services industry": 3,
+    "Total number of enterprises in the Communication Transmission Equipment industry": 120,
+    "Comparison result (larger one)": "the total number of enterprises in the Communication Transmission Equipment industry"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium085.json b/assets/qa_gold/enterprise_industry_analysis/medium085.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea9fc047accf6b2127edf9c0ab1f4e63761c88f8
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium085.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium085",
+  "question": "Which is larger: the number of HKEX-listed local state-owned enterprises in the industry where Zhongke Zhiyun Data Services Company operates, or the number of BSE-listed enterprises in the Transportation, Storage and Postal Services industry?",
+  "guidelines": "The answer must be either \"Zhongke Zhiyun Data Services Company\" or \"the number of BSE-listed enterprises in the Transportation, Storage and Postal Services industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Zhongke Zhiyun Data Services Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhongke Zhiyun Data Services Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed local state-owned enterprises in Information Transmission, Software and IT Services is 9.",
+    "Extract from national_industry_status.csv that the number of BSE-listed enterprises in the Transportation, Storage and Postal Services industry is 2.",
+    "Compare the counts: 9 > 2, so the number of HKEX-listed local state-owned enterprises in the industry where Zhongke Zhiyun Data Services Company operates is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Zhongke Zhiyun Data Services Company": "Information Transmission, Software and IT Services",
+    "Number of HKEX-listed local state-owned enterprises in Information Transmission, Software and IT Services": 9,
+    "Number of BSE-listed enterprises in the Transportation, Storage and Postal Services industry": 2,
+    "Comparison result (larger one)": "Zhongke Zhiyun Data Services Company"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium086.json b/assets/qa_gold/enterprise_industry_analysis/medium086.json
new file mode 100644
index 0000000000000000000000000000000000000000..732a803c4953ccc64bf04ae7047ba83de1e15cdc
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium086.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium086",
+  "question": "Which is larger: the number of SSE-listed central state-owned enterprises in the industry where Zhongke Zhiyun Data Services Company operates, or the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry?",
+  "guidelines": "The answer must be either \"the number of SSE-listed central state-owned enterprises in the industry where Zhongke Zhiyun Data Services Company operates\" or \"the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhongke Zhiyun Data Services Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the number of SSE-listed central state-owned enterprises in Information Transmission, Software and IT Services is 12.",
+    "Extract from national_industry_status.csv that the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry is 75.",
+    "Compare the counts: 75 > 12, so the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Zhongke Zhiyun Data Services Company": "Information Transmission, Software and IT Services",
+    "Number of SSE-listed central state-owned enterprises in Information Transmission, Software and IT Services": 12,
+    "Number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry": 75,
+    "Comparison result (larger one)": "the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium087.json b/assets/qa_gold/enterprise_industry_analysis/medium087.json
new file mode 100644
index 0000000000000000000000000000000000000000..bc82d8fdebcca095551c252923a42ee344ee4e5a
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium087.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium087",
+  "question": "Which is larger: the number of SZSE-listed central state-owned enterprises in the industry where Wuli Huida Chain Company operates, or the number of HKEX-listed foreign-funded enterprises in the Communication Transmission Equipment industry?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Wuli Huida Chain Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Wuli Huida Chain Company belongs to the Wholesale and Retail industry.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed central state-owned enterprises in the Wholesale and Retail industry is 8.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed foreign-funded enterprises in the Communication Transmission Equipment industry is 1.",
+    "Compare the counts: 8 > 1, so the industry where Wuli Huida Chain Company operates is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Wuli Huida Chain Company": "Wholesale and Retail",
+    "Number of SZSE-listed central state-owned enterprises in the Wholesale and Retail industry": 8,
+    "Number of HKEX-listed foreign-funded enterprises in the Communication Transmission Equipment industry": 1,
+    "Comparison result (larger one)": "Wuli Huida Chain Company"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium088.json b/assets/qa_gold/enterprise_industry_analysis/medium088.json
new file mode 100644
index 0000000000000000000000000000000000000000..157ebb866f1792a6d4e22b45d257d3a94c51498c
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium088.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium088",
+  "question": "Which is higher: the total capitalized R&D expenditure of the industry where Wuli Huida Chain Company operates, or that of the Communication Transmission Equipment industry?",
+  "guidelines": "The answer must be either \"the total capitalized R&D expenditure of the industry where Wuli Huida Chain Company operates\" or \"the Communication Transmission Equipment industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "the Communication Transmission Equipment industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Wuli Huida Chain Company belongs to the Wholesale and Retail industry.",
+    "Extract from national_industry_status.csv that the total capitalized R&D expenditure of the Wholesale and Retail industry is 2537411708.13 Yuan.",
+    "Extract from national_industry_status.csv that the total capitalized R&D expenditure of the Communication Transmission Equipment industry is 5279289114.7 Yuan.",
+    "Compare the totals: 5279289114.7 Yuan > 2537411708.13 Yuan, so the Communication Transmission Equipment industry is higher."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Wuli Huida Chain Company": "Wholesale and Retail",
+    "Total capitalized R&D expenditure of the Wholesale and Retail industry (Yuan)": 2537411708.13,
+    "Total capitalized R&D expenditure of the Communication Transmission Equipment industry (Yuan)": 5279289114.7,
+    "Comparison result (higher one)": "the Communication Transmission Equipment industry"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium089.json b/assets/qa_gold/enterprise_industry_analysis/medium089.json
new file mode 100644
index 0000000000000000000000000000000000000000..440c600305b101c2879506d6e2d4e180a95299c9
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium089.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium089",
+  "question": "What is the difference between the minimum cumulative citation count of core patents in the industry where Huaxin Yuanshi New Materials Company operates and the same metric in the Scientific Research and Technical Services industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Huaxin Yuanshi New Materials Company belongs to the Non-metallic Mineral Products industry.",
+    "Extract from national_industry_status.csv that the minimum cumulative citation count of core patents in the Non-metallic Mineral Products industry is 0.",
+    "Extract from national_industry_status.csv that the minimum cumulative citation count of core patents in the Scientific Research and Technical Services industry is 0.",
+    "Calculate the difference: 0 - 0 = 0.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Huaxin Yuanshi New Materials Company": "Non-metallic Mineral Products",
+    "Minimum cumulative citation count of core patents in the Non-metallic Mineral Products industry": 0,
+    "Minimum cumulative citation count of core patents in the Scientific Research and Technical Services industry": 0,
+    "Difference": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium090.json b/assets/qa_gold/enterprise_industry_analysis/medium090.json
new file mode 100644
index 0000000000000000000000000000000000000000..c515ec6fbc4477edd27435537e2693d9e64351c9
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium090.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium090",
+  "question": "Compared with the same indicator in the Scientific Research and Technical Services industry, what is the difference in the mean cumulative number of PCT invention patent applications for the industry where Huaxin Yuanshi New Materials Company operates?",
+  "guidelines": "The answer must be an exact number, preserving all meaningful decimal places. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": -21.9787234042554,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Huaxin Yuanshi New Materials Company belongs to the Non-metallic Mineral Products industry.",
+    "Extract from national_industry_status.csv that the mean cumulative number of PCT invention patent applications in the Non-metallic Mineral Products industry is 12.063829787234.",
+    "Extract from national_industry_status.csv that the mean cumulative number of PCT invention patent applications in the Scientific Research and Technical Services industry is 34.0425531914894.",
+    "Calculate the difference: 12.063829787234 - 34.0425531914894 = -21.9787234042554."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Huaxin Yuanshi New Materials Company": "Non-metallic Mineral Products",
+    "Mean cumulative number of PCT invention patent applications in the Non-metallic Mineral Products industry": 12.063829787234,
+    "Mean cumulative number of PCT invention patent applications in the Scientific Research and Technical Services industry": 34.0425531914894,
+    "Difference": -21.9787234042554
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium091.json b/assets/qa_gold/enterprise_industry_analysis/medium091.json
new file mode 100644
index 0000000000000000000000000000000000000000..a467cfcb7f29c7fa51bfbf1ac651326eb1ed1c19
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium091.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium091",
+  "question": "What is the difference between the maximum number of international industry awards in the industry where Aijian Yikang Fuzhongxin Company operates and the maximum number of international industry awards in the Chemical Fiber Manufacturing industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Aijian Yikang Fuzhongxin Company belongs to the Health and Social Work industry.",
+    "Extract from national_industry_status.csv that the maximum number of international industry awards in the Health and Social Work industry is 0.",
+    "Extract from national_industry_status.csv that the maximum number of international industry awards in the Chemical Fiber Manufacturing industry is 0.",
+    "Calculate the difference: 0 - 0 = 0.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Aijian Yikang Fuzhongxin Company": "Health and Social Work",
+    "Maximum number of international industry awards in the Health and Social Work industry": 0,
+    "Maximum number of international industry awards in the Chemical Fiber Manufacturing industry": 0,
+    "Difference": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium092.json b/assets/qa_gold/enterprise_industry_analysis/medium092.json
new file mode 100644
index 0000000000000000000000000000000000000000..df832eb80b298dab3afad6364b7a1a818dc46874
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium092.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium092",
+  "question": "What is the difference between the median year-on-year change in operating profit for the industry where Aijian Yikang Fuzhongxin Company operates and that of the Chemical Fiber Manufacturing industry?",
+  "guidelines": "The answer must be a single number with three decimal places. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": 16.675,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Aijian Yikang Fuzhongxin Company belongs to the Health and Social Work industry.",
+    "Extract from national_industry_status.csv that the median year-on-year change in operating profit for the Health and Social Work industry is -15.11 %.",
+    "Extract from national_industry_status.csv that the median year-on-year change in operating profit for the Chemical Fiber Manufacturing industry is -31.785 %.",
+    "Calculate the difference: -15.11 - (-31.785) = 16.675."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Aijian Yikang Fuzhongxin Company": "Health and Social Work",
+    "Median year-on-year change in operating profit for the Health and Social Work industry": -15.11,
+    "Median year-on-year change in operating profit for the Chemical Fiber Manufacturing industry": -31.785,
+    "Difference": 16.675
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium093.json b/assets/qa_gold/enterprise_industry_analysis/medium093.json
new file mode 100644
index 0000000000000000000000000000000000000000..e8a1552e4100697334226dbaf0e2b04710f41293
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium093.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium093",
+  "question": "What is the difference between the number of SSE-listed private enterprises in the industry where Zhongke Keshu Software Company operates and the number of SZSE-listed enterprises in Other Manufacturing in China?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": 71.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhongke Keshu Software Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the number of SSE-listed private enterprises in Information Transmission, Software and IT Services is 96.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed enterprises in Other Manufacturing is 25.",
+    "Calculate the difference: 96 - 25 = 71.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Zhongke Keshu Software Company": "Information Transmission, Software and IT Services",
+    "Number of SSE-listed private enterprises in Information Transmission, Software and IT Services": 96,
+    "Number of SZSE-listed enterprises in Other Manufacturing": 25,
+    "Difference": 71.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium094.json b/assets/qa_gold/enterprise_industry_analysis/medium094.json
new file mode 100644
index 0000000000000000000000000000000000000000..d11b4e0332d9b1310da64f4f11235a1d241c12a3
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium094.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium094",
+  "question": "What is the difference between the number of SZSE-listed private enterprises in the industry where Yihai Changjin Business Company operates and that in China's Transportation, Storage and Postal Services industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": 14.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Yihai Changjin Business Company belongs to the Leasing and Business Services industry.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed private enterprises in the Leasing and Business Services industry is 29.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed private enterprises in the Transportation, Storage and Postal Services industry is 15.",
+    "Calculate the difference: 29 - 15 = 14.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Yihai Changjin Business Company": "Leasing and Business Services",
+    "Number of SZSE-listed private enterprises in the Leasing and Business Services industry": 29,
+    "Number of SZSE-listed private enterprises in the Transportation, Storage and Postal Services industry": 15,
+    "Difference": 14.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium095.json b/assets/qa_gold/enterprise_industry_analysis/medium095.json
new file mode 100644
index 0000000000000000000000000000000000000000..838e399045ca9fdde89301799b72284d89148ace
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium095.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium095",
+  "question": "Which is larger: the number of SSE-listed private enterprises in the industry where Yihai Changjin Business Company operates, or the number of BSE-listed private enterprises in China's Transportation, Storage and Postal Services industry?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Yihai Changjin Business Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Yihai Changjin Business Company belongs to the Leasing and Business Services industry.",
+    "Extract from national_industry_status.csv that the number of SSE-listed private enterprises in the Leasing and Business Services industry is 11.",
+    "Extract from national_industry_status.csv that the number of BSE-listed private enterprises in the Transportation, Storage and Postal Services industry is 2.",
+    "Compare the counts: 11 > 2, so Yihai Changjin Business Company is higher."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Yihai Changjin Business Company": "Leasing and Business Services",
+    "Number of SSE-listed private enterprises in the Leasing and Business Services industry": 11,
+    "Number of BSE-listed private enterprises in the Transportation, Storage and Postal Services industry": 2,
+    "Comparison result (larger one)": "Yihai Changjin Business Company"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium096.json b/assets/qa_gold/enterprise_industry_analysis/medium096.json
new file mode 100644
index 0000000000000000000000000000000000000000..3b8b0b6791c3c7a8756b61291f2e43eeab5b8838
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium096.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium096",
+  "question": "Which is larger: the number of HKEX-listed central state-owned enterprises in the industry where Zhongche Yuanze Shipbuilding Company operates, or the number of HKEX-listed private enterprises in China's Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing industry?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhongche Yuanze Shipbuilding Company belongs to the Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing industry.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed central state-owned enterprises in this industry is 7.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed private enterprises in China's Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing industry is 13.",
+    "Compare the counts: 13 > 7, so the industry side is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Zhongche Yuanze Shipbuilding Company": "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Number of HKEX-listed central state-owned enterprises in Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing": 7,
+    "Number of HKEX-listed private enterprises in Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing": 13,
+    "Comparison result (larger one)": "industry"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium097.json b/assets/qa_gold/enterprise_industry_analysis/medium097.json
new file mode 100644
index 0000000000000000000000000000000000000000..47ec1e526b1b3aeb8c6b6039e31f4b1d5010bd71
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium097.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium097",
+  "question": "Which is larger: the number of HKEX-listed private enterprises in the industry where Zhongche Yuanze Shipbuilding Company operates, or the number of SSE-listed private enterprises in China's Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing industry?",
+  "guidelines": "The answer must be either \"the number of HKEX-listed private enterprises in the industry where Zhongche Yuanze Shipbuilding Company operates\", \"China\", or \"Equal\". Output only the answer itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhongche Yuanze Shipbuilding Company belongs to the Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing industry.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed private enterprises in this industry is 5.",
+    "Extract from national_industry_status.csv that the number of SSE-listed private enterprises in China's Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing industry is 5.",
+    "Compare the counts: 5 = 5, so the result is Equal."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Zhongche Yuanze Shipbuilding Company": "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Number of HKEX-listed private enterprises in Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing": 5,
+    "Number of SSE-listed private enterprises in Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing": 5,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium098.json b/assets/qa_gold/enterprise_industry_analysis/medium098.json
new file mode 100644
index 0000000000000000000000000000000000000000..c2dbb2664cf6f3b07a72943024f70c096c2bfc28
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium098.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium098",
+  "question": "Which is larger: the number of SZSE-listed enterprises in the industry where Yihai Changjin Business Company operates, or the number of SZSE-listed private enterprises in China's Conglomerates industry?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Yihai Changjin Business Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Yihai Changjin Business Company belongs to the Leasing and Business Services industry.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed enterprises in the Leasing and Business Services industry is 44.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed private enterprises in the Conglomerates industry is 4.",
+    "Compare the counts: 44 > 4, so Yihai Changjin Business Company is higher."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Yihai Changjin Business Company": "Leasing and Business Services",
+    "Number of SZSE-listed enterprises in the Leasing and Business Services industry": 44,
+    "Number of SZSE-listed private enterprises in the Conglomerates industry": 4,
+    "Comparison result (larger one)": "Yihai Changjin Business Company"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium099.json b/assets/qa_gold/enterprise_industry_analysis/medium099.json
new file mode 100644
index 0000000000000000000000000000000000000000..71d28f35433d5217ebf30aef46a865e2652da07c
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium099.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium099",
+  "question": "Yi Hai Chang Jin Shang Wu Co., Ltd.industry's Number of SZSE-listed enterprises and China ConglomeratesindustryPrivate enterpriseShenzhen Stock Exchange countcompared with which unitsgreater？",
+  "guidelines": "The answer must a company name or \"industry\", Output onlyname, without any explanation or description.Wu, answer\"No relevant data found\".",
+  "answer": "Yi Hai Chang Jin Shang Wu Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Yi Hai Chang Jin Shang Wu Co., Ltd.industry = Business Services",
+    "national_industry_status.csv in extractBusiness Services Number of SZSE-listed enterprises= 44",
+    "national_industry_status.csv in extractConglomeratesindustry Private enterprise_Number of SZSE-listed enterprises= 4",
+    "Comparecount：44 > 4, Yi Hai Chang Jin Shang Wu Co., Ltd.higher"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Yi Hai Chang Jin Shang Wu Co., Ltd.industry": "Business Services",
+    "Business Services Number of SZSE-listed enterprises": 44,
+    "Conglomerates Private enterprise_Number of SZSE-listed enterprises": 4,
+    "（greater）": "Yi Hai Chang Jin Shang Wu Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium100.json b/assets/qa_gold/enterprise_industry_analysis/medium100.json
new file mode 100644
index 0000000000000000000000000000000000000000..450fc8612733d30270b2739b11d33f6b3ca66ee8
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium100.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium100",
+  "question": "What is the difference between the median year-on-year change in R&D expenditure for the industry where Biyuan Zhize Urban Development Company operates and that of the Pharmaceutical Manufacturing industry in China?",
+  "guidelines": "The answer must be a single number with two decimal places. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": -9.87,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Biyuan Zhize Urban Development Company belongs to the Real Estate industry.",
+    "Extract from national_industry_status.csv that the median year-on-year change in R&D expenditure for the Real Estate industry is 1.68 %.",
+    "Extract from national_industry_status.csv that the median year-on-year change in R&D expenditure for the Pharmaceutical Manufacturing industry is 11.55 %.",
+    "Calculate the gap: 1.68 - 11.55 = -9.87."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Biyuan Zhize Urban Development Company": "Real Estate",
+    "Median year-on-year change in R&D expenditure for the Real Estate industry": 1.68,
+    "Median year-on-year change in R&D expenditure for the Pharmaceutical Manufacturing industry": 11.55,
+    "Gap": -9.87
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium101.json b/assets/qa_gold/enterprise_industry_analysis/medium101.json
new file mode 100644
index 0000000000000000000000000000000000000000..b6230a880c08e6b505cc70e8ad585d5799065895
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium101.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium101",
+  "question": "What is the difference between the number of SSE-listed foreign-funded enterprises in the industry corresponding to Biyuan Zhize Urban Development Company and the number of HKEX-listed state-owned research institute enterprises in China's Pharmaceutical Manufacturing industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": 5.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that the corresponding industry of Biyuan Zhize Urban Development Company is Real Estate.",
+    "Extract from national_industry_status.csv that the number of SSE-listed foreign-funded enterprises in the Real Estate industry is 6.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed state-owned research institute enterprises in the Pharmaceutical Manufacturing industry is 1.",
+    "Calculate the difference: 6 - 1 = 5.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry corresponding to Biyuan Zhize Urban Development Company": "Real Estate",
+    "Number of SSE-listed foreign-funded enterprises in the Real Estate industry": 6,
+    "Number of HKEX-listed state-owned research institute enterprises in the Pharmaceutical Manufacturing industry": 1,
+    "Difference": 5.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium102.json b/assets/qa_gold/enterprise_industry_analysis/medium102.json
new file mode 100644
index 0000000000000000000000000000000000000000..510b3809203e9a11bce7b7ec447108d2376ee195
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium102.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium102",
+  "question": "Which is larger: the total number of Consumer Electronics and Electrical industry enterprises in the province where Changqiao Jinchuang Technology Company is located, or the number of HKEX-listed foreign-funded enterprises in China's Conglomerates industry?",
+  "guidelines": "The answer must be either \"the total number of Consumer Electronics and Electrical industry enterprises in the province where Changqiao Jinchuang Technology Company is located\" or \"the number of HKEX-listed foreign-funded enterprises in China's Conglomerates industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "the number of HKEX-listed foreign-funded enterprises in China's Conglomerates industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Changqiao Jinchuang Technology Company is located in Gansu Province and belongs to the Consumer Electronics and Electrical industry.",
+    "Extract from regional_industry_status.csv that the total number of enterprises in Gansu Province - Consumer Electronics and Electrical industry is 0.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed foreign-funded enterprises in the Conglomerates industry is 5.",
+    "Compare the counts: 5 > 0, so the number of HKEX-listed foreign-funded enterprises in China's Conglomerates industry is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Changqiao Jinchuang Technology Company": "Gansu Province",
+    "Total number of enterprises in Gansu Province - Consumer Electronics and Electrical industry": 0,
+    "Number of HKEX-listed foreign-funded enterprises in the Conglomerates industry": 5,
+    "Comparison result (larger one)": "the number of HKEX-listed foreign-funded enterprises in China's Conglomerates industry"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium103.json b/assets/qa_gold/enterprise_industry_analysis/medium103.json
new file mode 100644
index 0000000000000000000000000000000000000000..e014a6436bf33b64102fc9d889a9af676dc63ce9
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium103.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium103",
+  "question": "Which is larger: the total number of Consumer Electronics and Electrical industry enterprises in the province where Changqiao Jinchuang Technology Company is located, or the number of HKEX-listed private enterprises in China's Conglomerates industry?",
+  "guidelines": "The answer must be either \"the total number of Consumer Electronics and Electrical industry enterprises in the province where Changqiao Jinchuang Technology Company is located\" or \"the number of HKEX-listed private enterprises in China's Conglomerates industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "the number of HKEX-listed private enterprises in China's Conglomerates industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Changqiao Jinchuang Technology Company is located in Gansu Province and belongs to the Consumer Electronics and Electrical industry.",
+    "Extract from regional_industry_status.csv that the total number of enterprises in Gansu Province - Consumer Electronics and Electrical industry is 0.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed private enterprises in the Conglomerates industry is 12.",
+    "Compare the counts: 12 > 0, so the number of HKEX-listed private enterprises in China's Conglomerates industry is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Changqiao Jinchuang Technology Company": "Gansu Province",
+    "Total number of enterprises in Gansu Province - Consumer Electronics and Electrical industry": 0,
+    "Number of HKEX-listed private enterprises in the Conglomerates industry": 12,
+    "Comparison result (larger one)": "the number of HKEX-listed private enterprises in China's Conglomerates industry"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium104.json b/assets/qa_gold/enterprise_industry_analysis/medium104.json
new file mode 100644
index 0000000000000000000000000000000000000000..a6d239c7193649af89dbae183dc51ac904716424
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium104.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium104",
+  "question": "Which value is larger: the number of HKEX-listed foreign-funded enterprises in the province where Jinzhi Hongsheng Asset Management Company is located, or the number of SSE-listed central state-owned enterprises in China's Culture, Sports and Entertainment industry?",
+  "guidelines": "The answer must be either \"the number of HKEX-listed foreign-funded enterprises in the province where Jinzhi Hongsheng Asset Management Company is located\", \"the number of SSE-listed central state-owned enterprises in China's Culture, Sports and Entertainment industry\", or \"Equal\". Output only the answer itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Jinzhi Hongsheng Asset Management Company is located in Shanghai Municipality and belongs to the Financial industry.",
+    "Extract from regional_industry_status.csv that the number of HKEX-listed foreign-funded enterprises in Shanghai Municipality - Financial industry is 2.",
+    "Extract from national_industry_status.csv that the number of SSE-listed central state-owned enterprises in China's Culture, Sports and Entertainment industry is 2.",
+    "Compare the values: 2 = 2, so the result is Equal."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Jinzhi Hongsheng Asset Management Company": "Shanghai Municipality",
+    "Number of HKEX-listed foreign-funded enterprises in Shanghai Municipality - Financial industry": 2,
+    "Number of SSE-listed central state-owned enterprises in Culture, Sports and Entertainment": 2,
+    "Comparison result (larger one)": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium105.json b/assets/qa_gold/enterprise_industry_analysis/medium105.json
new file mode 100644
index 0000000000000000000000000000000000000000..57fbfc58d1e1b6d2eff3f4c34a5c8caa8c4cb8b4
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium105.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium105",
+  "question": "What is the difference between the mean number of provincial or ministerial Science and Technology Progress Awards in the province where Jinzhi Hongsheng Asset Management Company is located and the mean number of the same awards in China's Culture, Sports and Entertainment industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": 3.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Jinzhi Hongsheng Asset Management Company is located in Shanghai Municipality and belongs to the Financial industry.",
+    "Extract from regional_industry_status.csv that the mean number of provincial or ministerial Science and Technology Progress Awards in Shanghai Municipality - Financial industry is 3.",
+    "Extract from national_industry_status.csv that the mean number of provincial or ministerial Science and Technology Progress Awards in the Culture, Sports and Entertainment industry is 0.",
+    "Calculate the difference: 3 - 0 = 3.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Jinzhi Hongsheng Asset Management Company": "Shanghai Municipality",
+    "Mean number of provincial or ministerial Science and Technology Progress Awards in Shanghai Municipality - Financial industry": 3,
+    "Mean number of provincial or ministerial Science and Technology Progress Awards in Culture, Sports and Entertainment": 0,
+    "Difference": 3.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium106.json b/assets/qa_gold/enterprise_industry_analysis/medium106.json
new file mode 100644
index 0000000000000000000000000000000000000000..c05f2d22b0ad8ec82f81198239886606b701e2b0
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium106.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium106",
+  "question": "What is the difference between the total number of Consumer Electronics and Electrical industry enterprises in the province where Shiyang Jinjin Electrical Appliances Company is located and the number of SZSE-listed enterprises in China's Rubber and Plastic Products industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": -68.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Shiyang Jinjin Electrical Appliances Company is located in the Inner Mongolia Autonomous Region and belongs to the Consumer Electronics and Electrical industry.",
+    "Extract from regional_industry_status.csv that the total number of enterprises in Inner Mongolia Autonomous Region - Consumer Electronics and Electrical industry is 0.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed enterprises in the Rubber and Plastic Products industry is 68.",
+    "Calculate the difference: 0 - 68 = -68.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Shiyang Jinjin Electrical Appliances Company": "Inner Mongolia Autonomous Region",
+    "Total number of enterprises in Inner Mongolia Autonomous Region - Consumer Electronics and Electrical industry": 0,
+    "Number of SZSE-listed enterprises in the Rubber and Plastic Products industry": 68,
+    "Difference": -68.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium107.json b/assets/qa_gold/enterprise_industry_analysis/medium107.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d12a5d2cbc272fb9eb3828ae840bda7d8217cd3
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium107.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium107",
+  "question": "What is the difference between the total number of Consumer Electronics and Electrical industry enterprises in the province where Shiyang Jinjin Electrical Appliances Company is located and the number of enterprises in China's Rubber and Plastic Products industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": -107.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Shiyang Jinjin Electrical Appliances Company is located in the Inner Mongolia Autonomous Region and belongs to the Consumer Electronics and Electrical industry.",
+    "Extract from regional_industry_status.csv that the total number of enterprises in Inner Mongolia Autonomous Region - Consumer Electronics and Electrical industry is 0.",
+    "Extract from national_industry_status.csv that the number of enterprises in the Rubber and Plastic Products industry is 107.",
+    "Calculate the difference: 0 - 107 = -107.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Shiyang Jinjin Electrical Appliances Company": "Inner Mongolia Autonomous Region",
+    "Total number of enterprises in Inner Mongolia Autonomous Region - Consumer Electronics and Electrical industry": 0,
+    "Number of enterprises in the Rubber and Plastic Products industry": 107,
+    "Difference": -107.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium108.json b/assets/qa_gold/enterprise_industry_analysis/medium108.json
new file mode 100644
index 0000000000000000000000000000000000000000..a2464618012809012ab8637e0232a99e5fc72bf9
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium108.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium108",
+  "question": "Which value is larger: the total number of industry enterprises in the province where Zhongke Keshu Software Company is located, or the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry?",
+  "guidelines": "The answer must be either \"the total number of industry enterprises in the province where Zhongke Keshu Software Company is located\" or \"the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhongke Keshu Software Company is located in Guangxi Zhuang Autonomous Region and belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from regional_industry_status.csv that the total number of enterprises in Guangxi Zhuang Autonomous Region - Information Transmission, Software and IT Services is 4.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed private enterprises in the Communication Transmission Equipment industry is 50.",
+    "Compare the values: 50 > 4, so the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Zhongke Keshu Software Company": "Guangxi Zhuang Autonomous Region",
+    "Total number of enterprises in Guangxi Zhuang Autonomous Region - Information Transmission, Software and IT Services": 4,
+    "Number of SZSE-listed private enterprises in the Communication Transmission Equipment industry": 50,
+    "Comparison result (larger one)": "the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium109.json b/assets/qa_gold/enterprise_industry_analysis/medium109.json
new file mode 100644
index 0000000000000000000000000000000000000000..37090b982230ff0ccf342a09735dae20dbea99a5
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium109.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium109",
+  "question": "Which is larger: the number of SSE-listed private enterprises in the Information Transmission, Software and IT Services industry in the province where Zhongke Keshu Software Company is located, or the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry?",
+  "guidelines": "The answer must be either \"the number of SSE-listed private enterprises in the Information Transmission, Software and IT Services industry in the province where Zhongke Keshu Software Company is located\" or \"the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhongke Keshu Software Company is located in Guangxi Zhuang Autonomous Region and belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from regional_industry_status.csv that the number of SSE-listed private enterprises in Guangxi Zhuang Autonomous Region - Information Transmission, Software and IT Services is 2.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed private enterprises in the Communication Transmission Equipment industry is 50.",
+    "Compare the counts: 50 > 2, so the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry is larger."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Zhongke Keshu Software Company": "Guangxi Zhuang Autonomous Region",
+    "Number of SSE-listed private enterprises in Guangxi Zhuang Autonomous Region - Information Transmission, Software and IT Services": 2,
+    "Number of SZSE-listed private enterprises in the Communication Transmission Equipment industry": 50,
+    "Comparison result (larger one)": "the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium110.json b/assets/qa_gold/enterprise_industry_analysis/medium110.json
new file mode 100644
index 0000000000000000000000000000000000000000..84c1e78ea57aea0db4ec40f2fd0a3f139ed22fe5
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium110.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium110",
+  "question": "What is the difference between the total annual number of China patent grants in the province where Xingkuwen Arts and Crafts Company is located and that in China's Chemical Fiber Manufacturing industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": -1314.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Xingkuwen Arts and Crafts Company is located in Zhejiang Province and belongs to the Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing industry.",
+    "Extract from regional_industry_status.csv that the total annual number of China patent grants in Zhejiang Province - Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing is 193.",
+    "Extract from national_industry_status.csv that the total annual number of China patent grants in the Chemical Fiber Manufacturing industry is 1507.",
+    "Calculate the gap: 193 - 1507 = -1314.0."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Xingkuwen Arts and Crafts Company": "Zhejiang Province",
+    "Total annual number of China patent grants in Zhejiang Province - Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing": 193,
+    "Total annual number of China patent grants in the Chemical Fiber Manufacturing industry": 1507,
+    "Gap": -1314.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_analysis/medium111.json b/assets/qa_gold/enterprise_industry_analysis/medium111.json
new file mode 100644
index 0000000000000000000000000000000000000000..cafefb76586934aa34ea4ae2a25c0df592663e07
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_analysis/medium111.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium111",
+  "question": "Which is larger: the number of SZSE-listed enterprises in the province where Xingkuwen Arts and Crafts Company is located, or the number of SSE-listed private enterprises in China's Chemical Fiber Manufacturing industry?",
+  "guidelines": "The answer must be either \"the number of SZSE-listed enterprises in the province where Xingkuwen Arts and Crafts Company is located\" or \"the number of SSE-listed private enterprises in China's Chemical Fiber Manufacturing industry\". Output only one of these, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "answer": "the number of SSE-listed private enterprises in China's Chemical Fiber Manufacturing industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Determine from company_profile.csv that Xingkuwen Arts and Crafts Company is located in Zhejiang Province.",
+    "Extract from regional_industry_status.csv that the number of SZSE-listed enterprises in Zhejiang Province, Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing is 3.",
+    "Extract from national_industry_status.csv that the number of SSE-listed private enterprises in the Chemical Fiber Manufacturing industry is 7.",
+    "Compare 3 and 7; 7 is larger, so output \"the number of SSE-listed private enterprises in China's Chemical Fiber Manufacturing industry\"."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Xingkuwen Arts and Crafts Company": "Zhejiang Province",
+    "Number of SZSE-listed enterprises in Zhejiang Province, Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing": 3,
+    "Number of SSE-listed private enterprises in Chemical Fiber Manufacturing": 7,
+    "Comparison result (larger one)": "the number of SSE-listed private enterprises in China's Chemical Fiber Manufacturing industry"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/easy067.json b/assets/qa_gold/enterprise_industry_policy_analysis/easy067.json
new file mode 100644
index 0000000000000000000000000000000000000000..72775197e7002ebfbd1baddb512ec6ef466271c0
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/easy067.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy067",
+  "question": "“江西省人民政府印发关于做优做强我省锂电新能源产业若干政策措施的通知”对众白昌锦商贸公司的行业是否适用？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "answer": "否",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "从company_profile.csv中抽取众白昌锦商贸公司所在省份为上海市",
+    "从policy_resource.csv中抽取政策《江西省人民政府印发关于做优做强我省锂电新能源产业若干政策措施的通知》的适用区域为江西省",
+    "判断企业所在省份是否在政策适用区域内，不在则输出「否」"
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "众白昌锦商贸公司所在省份": "上海市",
+    "政策适用区域": "江西省",
+    "是否适用": "否"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/easy068.json b/assets/qa_gold/enterprise_industry_policy_analysis/easy068.json
new file mode 100644
index 0000000000000000000000000000000000000000..bdc293f1c81dd37453e04df0418efeccf5728a43
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/easy068.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy068",
+  "question": "大花表仪医疗科技公司所在行业是否会受到“合肥市促进“两强一增”行动若干政策”的影响？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "answer": "否",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "从company_profile.csv中抽取大花表仪医疗科技公司所在省份为四川省",
+    "从policy_resource.csv中抽取政策《合肥市促进“两强一增”行动若干政策》的适用区域为合肥市",
+    "判断企业所在省份是否在政策适用区域内，不在则输出「否」"
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "大花表仪医疗科技公司所在省份": "四川省",
+    "政策适用区域": "合肥市",
+    "是否适用": "否"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/easy069.json b/assets/qa_gold/enterprise_industry_policy_analysis/easy069.json
new file mode 100644
index 0000000000000000000000000000000000000000..df07ad9d2a5087808f0da73c3f3683de13af118f
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/easy069.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy069",
+  "question": "丽群汇通零售公司所在行业是否会受到“教育部办公厅　工业和信息化部办公厅　国家知识产权局办公室关于组织开展“千校万企”协同创新伙伴行动的通知”的影响？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "answer": "否",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "从company_profile.csv中抽取丽群汇通零售公司所在行业为批发和零售业",
+    "从policy_resource.csv中抽取政策《教育部办公厅 工业和信息化部办公厅 国家知识产权局办公室关于组织开展“千校万企”协同创新伙伴行动的通知》的适用行业为教育、科学研究和技术服务业",
+    "判断企业所在行业是否命中政策适用行业，不命中则输出「否」"
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "丽群汇通零售公司所在行业": "批发和零售业",
+    "政策适用行业": "教育、科学研究和技术服务业",
+    "是否适用": "否"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/easy070.json b/assets/qa_gold/enterprise_industry_policy_analysis/easy070.json
new file mode 100644
index 0000000000000000000000000000000000000000..e7fd456619a5013142ae9ad3d5a927adf21de1df
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/easy070.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy070",
+  "question": "“关于印发重庆市促进大中小企业融通发展工作方案（2022—2025年）的通知”对乐动乐博娱乐用品公司的行业是否适用？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "answer": "否",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "从company_profile.csv中抽取乐动乐博娱乐用品公司所在省份为广东省",
+    "从policy_resource.csv中抽取政策《关于印发重庆市促进大中小企业融通发展工作方案（2022—2025年）的通知》的适用区域为重庆市",
+    "判断企业所在省份是否在政策适用区域内，不在则输出「否」"
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "乐动乐博娱乐用品公司所在省份": "广东省",
+    "政策适用区域": "重庆市",
+    "是否适用": "否"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/easy071.json b/assets/qa_gold/enterprise_industry_policy_analysis/easy071.json
new file mode 100644
index 0000000000000000000000000000000000000000..6814fac73fbdca0b7fd542da5ac4f89859156f05
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/easy071.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy071",
+  "question": "“科技部等九部门关于印发《“十四五” 东西部科技合作实施方案》的通知”对亚玮工泽机床公司的行业是否适用？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "answer": "否",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "从company_profile.csv中抽取亚玮工泽机床公司所在行业为通用设备制造业",
+    "从policy_resource.csv中抽取政策《科技部等九部门关于印发《“十四五” 东西部科技合作实施方案》的通知》适用行业为科学研究和技术服务业",
+    "判断企业所在行业是否命中政策适用行业，不命中则输出「否」"
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "亚玮工泽机床公司所在行业": "通用设备制造业",
+    "政策适用行业": "科学研究和技术服务业",
+    "是否适用": "否"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/easy072.json b/assets/qa_gold/enterprise_industry_policy_analysis/easy072.json
new file mode 100644
index 0000000000000000000000000000000000000000..e5cbcdf44bf7627d7c7eda180e853df121f419ce
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/easy072.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy072",
+  "question": "果融泽鸿资产管理公司所在行业是否受益于“自治区发展改革委关于印发《宁夏回族自治区氢能产业发展规划》的通知”？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "answer": "否",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "从company_profile.csv中抽取果融泽鸿资产管理公司所在省份为山东省",
+    "从policy_resource.csv中抽取政策《自治区发展改革委关于印发《宁夏回族自治区氢能产业发展规划》的通知》适用区域为宁夏回族自治区",
+    "判断企业所在省份是否在政策适用区域内，不在则输出「否」"
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "果融泽鸿资产管理公司所在省份": "山东省",
+    "政策适用区域": "宁夏回族自治区",
+    "是否适用": "否"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/easy073.json b/assets/qa_gold/enterprise_industry_policy_analysis/easy073.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8ce28ba4a935783a274294060cb0dea2b4737ea
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/easy073.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy073",
+  "question": "“商务部等14部门关于开展内外贸一体化试点的通知”是否可以推进物丽昌源批发公司所属行业发展？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "answer": "是",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "从company_profile.csv中抽取物丽昌源批发公司所在行业为批发和零售业",
+    "从policy_resource.csv中抽取政策《商务部等14部门关于开展内外贸一体化试点的通知》适用行业列表",
+    "判断企业所在行业是否命中政策适用行业，命中则输出「是」"
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "物丽昌源批发公司所在行业": "批发和零售业",
+    "政策适用行业": "批发和零售业；互联网和相关服务；交通运输、仓储和邮政业；租赁和商务服务业；电信、广播电视和卫星传输服务；居民服务、修理和其他服务业",
+    "是否适用": "是"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/easy074.json b/assets/qa_gold/enterprise_industry_policy_analysis/easy074.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d4fdcf984bc5b0c563bd2ef1100b858029084d5
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/easy074.json
@@ -0,0 +1,24 @@
+{
+  "id": "easy074",
+  "question": "“广东省人民政府关于印发中国（韶关）等8个  跨境电子商务综合试验区实施方案的通知”对恒通达达信息技术公司的行业是否适用？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "answer": "是",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "从company_profile.csv中抽取恒通达达信息技术公司所在行业为批发和零售业且所在省份为广东省",
+    "从policy_resource.csv中抽取政策《广东省人民政府关于印发中国（韶关）等8个 跨境电子商务综合试验区实施方案的通知》的适用行业和适用省份",
+    "判断企业行业与省份是否同时命中政策适用范围，命中则输出「是」"
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "恒通达达信息技术公司所在行业": "批发和零售业",
+    "恒通达达信息技术公司所在省份": "广东省",
+    "政策适用行业": "批发和零售业",
+    "政策适用省份": "广东省",
+    "是否适用": "是"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/easy075.json b/assets/qa_gold/enterprise_industry_policy_analysis/easy075.json
new file mode 100644
index 0000000000000000000000000000000000000000..70909a4c934754b3116abee8f00ecd55790944de
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/easy075.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy075",
+  "question": "众课科数软件公司所在行业是否受益于“民政部、中央政法委、中央网信办、发展改革委、工业和信息化部、公安部、财政部、住房城乡建设部、农业农村部印发《关于深入推进智慧社区建设的意见》的通知”？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "answer": "是",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "从company_profile.csv中抽取众课科数软件公司所在行业为信息传输、软件和信息技术服务业",
+    "从policy_resource.csv中抽取政策《民政部、中央政法委、中央网信办、发展改革委、工业和信息化部、公安部、财政部、住房城乡建设部、农业农村部印发〈关于深入推进智慧社区建设的意见〉的通知》的适用行业列表",
+    "判断企业所在行业是否命中政策适用行业，命中则输出「是」"
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "众课科数软件公司所在行业": "信息传输、软件和信息技术服务业",
+    "政策适用行业": "信息传输、软件和信息技术服务业；互联网和相关服务；电信、广播电视和卫星传输服务",
+    "是否适用": "是"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/easy076.json b/assets/qa_gold/enterprise_industry_policy_analysis/easy076.json
new file mode 100644
index 0000000000000000000000000000000000000000..6900cd410dc9b6c3f49e24ed68a1ab0e1095cedb
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/easy076.json
@@ -0,0 +1,22 @@
+{
+  "id": "easy076",
+  "question": "“四部门关于公布农业、建筑、医疗、矿山领域机器人典型应用场景名单的通知”对以山泽辰医疗器械公司的行业是否适用？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "answer": "否",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "从company_profile.csv中抽取以山泽辰医疗器械公司所在行业为医药制造业",
+    "从policy_resource.csv中抽取政策《四部门关于公布农业、建筑、医疗、矿山领域机器人典型应用场景名单的通知》的适用行业列表",
+    "判断企业所在行业是否命中政策适用行业，不命中则输出「否」"
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "以山泽辰医疗器械公司所在行业": "医药制造业",
+    "政策适用行业": "农、林、牧、渔业；建筑业；卫生和社会工作；采矿业",
+    "是否适用": "否"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium001.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium001.json
new file mode 100644
index 0000000000000000000000000000000000000000..75b1d8a953f4e478cbafd815f60095d1d536b2dc
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium001.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium001",
+  "question": "What is the gap between the number of central ministry/agency policies issued by the Ministry of Commerce in the ministerial policies for the industry of Wu Li Hui Da Lian Suo Co., Ltd. and the number of local policies issued by the Hainan Province Department of Industry and Information Technology for the industry of Xing Ku Wen Gong Yi Mei Shu Pin Co., Ltd.?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that the industry of Wu Li Hui Da Lian Suo Co., Ltd. is Wholesale and Retail",
+    "Extract from policy_release_status.csv that in ministerial policies for Wholesale and Retail, the number of policies issued by the Ministry of Commerce is 3",
+    "Extract from company_profile.csv that the industry of Xing Ku Wen Gong Yi Mei Shu Pin Co., Ltd. is Cultural, Arts, Sports and Entertainment Goods Manufacturing",
+    "Extract from policy_release_status.csv that in local policies for Cultural, Arts, Sports and Entertainment Goods Manufacturing, the number of policies issued by the Hainan Province Department of Industry and Information Technology is 1",
+    "Compute the gap: 3 - 1 = 2.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Wu Li Hui Da Lian Suo Co., Ltd.": "Wholesale and Retail",
+    "Wholesale and Retail ministerial policies_Ministry of Commerce number of policies": 3,
+    "Industry of Xing Ku Wen Gong Yi Mei Shu Pin Co., Ltd.": "Cultural, Arts, Sports and Entertainment Goods Manufacturing",
+    "Cultural, Arts, Sports and Entertainment Goods Manufacturing local policies_Hainan Province Department of Industry and Information Technology number of policies": 1,
+    "Gap (Ministry of Commerce number of policies - Hainan Province Department of Industry and Information Technology number of policies)": 2.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium002.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium002.json
new file mode 100644
index 0000000000000000000000000000000000000000..4fbeae5a2145303a311521bbd228b5f8fdceb65b
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium002.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium002",
+  "question": "What is the difference between the number of policies issued by the Gansu Province General Office of the People's Government in the local policies for the Wholesale and Retail industry and the number of policies issued by the Ministry of Transport in the ministerial policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without any unit, comma, or explanatory text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Wu Li Hui Da Lian Suo Co., Ltd.'s industry is Wholesale and Retail",
+    "From policy_release_status.csv, extract that in the local policies for Wholesale and Retail, the number of policies issued by the Gansu Province General Office of the People's Government is 1",
+    "From company_profile.csv, extract that Xing Ku Wen Gong Yi Mei Shu Pin Co., Ltd.'s industry is Cultural, Arts, Sports and Entertainment Goods Manufacturing",
+    "From policy_release_status.csv, extract that in the ministerial policies for Cultural, Arts, Sports and Entertainment Goods Manufacturing, the number of policies issued by the Ministry of Transport is 1",
+    "Compute the difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Wu Li Hui Da Lian Suo Co., Ltd.": "Wholesale and Retail",
+    "Wholesale and Retail local policies_Gansu Province General Office of the People's Government number of policies": 1,
+    "Industry of Xing Ku Wen Gong Yi Mei Shu Pin Co., Ltd.": "Cultural, Arts, Sports and Entertainment Goods Manufacturing",
+    "Cultural, Arts, Sports and Entertainment Goods Manufacturing ministerial policies_Ministry of Transport number of policies": 1,
+    "Difference (Gansu Province General Office of the People's Government number of policies - Ministry of Transport number of policies)": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium003.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium003.json
new file mode 100644
index 0000000000000000000000000000000000000000..cbc9a4eb0df060eb61a91e34dd05fdcdb3ab3d89
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium003.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium003",
+  "question": "Which is greater: the number of policies issued by the General Administration of Customs in the ministerial policies for the industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd., or the number of policies issued by the General Office of the State Council in the State Council policies for the industry of Bei Kong Ze Jing Water Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Bei Kong Ze Jing Water Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Ying Tai Sheng Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in Financial Services ministerial policies, the General Administration of Customs policy count is 1",
+    "Extract from company_profile.csv that Bei Kong Ze Jing Water Co., Ltd.'s industry is Water Conservancy, Environment and Public Facilities Management",
+    "Extract from policy_release_status.csv that in Water Conservancy, Environment and Public Facilities Management State Council policies, the General Office of the State Council policy count is 1",
+    "Compare 1 and 1; when equal, output \"Bei Kong Ze Jing Water Co., Ltd.\" as required by the question"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_General Administration of Customs policy count": 1,
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.": "Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management State Council policies_General Office of the State Council policy count": 1,
+    "Comparison result": "Both are equal; per question requirement, output Bei Kong Ze Jing Water Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium004.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium004.json
new file mode 100644
index 0000000000000000000000000000000000000000..61e338d919cbce9442af24d3d73c662380960d46
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium004.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium004",
+  "question": "Which is larger: the number of central ministry/agency policies issued by the Ministry of Agriculture and Rural Affairs in the ministerial policies for the Financial Services industry, or the number of central ministry/agency policies issued by the National Health Commission in the ministerial policies for the Water Conservancy, Environment and Public Facilities Management industry?",
+  "guidelines": "The answer must be a company name; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Bei Kong Ze Jing Water Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Hua Ying Tai Sheng Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "From policy_release_status.csv, extract that in the ministerial policies for Financial Services, the number of policies issued by the Ministry of Agriculture and Rural Affairs is 1",
+    "From company_profile.csv, extract that Bei Kong Ze Jing Water Co., Ltd.'s industry is Water Conservancy, Environment and Public Facilities Management",
+    "From policy_release_status.csv, extract that in the ministerial policies for Water Conservancy, Environment and Public Facilities Management, the number of policies issued by the National Health Commission is 2",
+    "Compare 1 and 2; 2 is larger; output \"Bei Kong Ze Jing Water Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_Ministry of Agriculture and Rural Affairs number of policies": 1,
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.": "Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management ministerial policies_National Health Commission number of policies": 2,
+    "Comparison result": "2 is larger; output Bei Kong Ze Jing Water Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium005.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium005.json
new file mode 100644
index 0000000000000000000000000000000000000000..a963e94654ab100531c85e0a1c587c8f07e1d553
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium005.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium005",
+  "question": "Is the number of central ministry/agency policies issued by the Development and Reform Commission in the ministerial policies for the industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. higher than the number of local policies issued by the Shenzhen Municipality Bureau of Commerce in the local policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "From policy_release_status.csv, extract that in the ministerial policies for Information Transmission, Software and IT Services, the number of policies issued by the Development and Reform Commission is 2",
+    "From company_profile.csv, extract that Yi Hai Chang Jin Shang Wu Co., Ltd.'s industry is Leasing and Business Services",
+    "From policy_release_status.csv, extract that in the local policies for Leasing and Business Services, the number of policies issued by the Shenzhen Municipality Bureau of Commerce is 1",
+    "Determine whether 2 is greater than 1; if so, output \"Yes\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies_Development and Reform Commission number of policies": 2,
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Leasing and Business Services local policies_Shenzhen Municipality Bureau of Commerce number of policies": 1,
+    "Whether higher (Development and Reform Commission number of policies > Shenzhen Municipality Bureau of Commerce number of policies)": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium006.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium006.json
new file mode 100644
index 0000000000000000000000000000000000000000..6df7e5679081574a29e9b2cb5304a282242816e6
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium006.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium006",
+  "question": "Which is greater: the number of local policies issued by the Yunnan Province Communications Administration in the local policies for the industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd., or the number of central ministry/agency policies issued by the Development and Reform Commission in the ministerial policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd.?",
+  "guidelines": "The answer must be either a company name or \"Equal\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "From policy_release_status.csv, extract that in the local policies for Information Transmission, Software and IT Services, the number of policies issued by the Yunnan Province Communications Administration is 1",
+    "From company_profile.csv, extract that Yi Hai Chang Jin Shang Wu Co., Ltd.'s industry is Leasing and Business Services",
+    "From policy_release_status.csv, extract that in the ministerial policies for Leasing and Business Services, the number of policies issued by the Development and Reform Commission is 1",
+    "Compare 1 and 1; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies_Yunnan Province Communications Administration number of policies": 1,
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Leasing and Business Services ministerial policies_Development and Reform Commission number of policies": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium007.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium007.json
new file mode 100644
index 0000000000000000000000000000000000000000..1acaf4a9333966b466ceaee890813630ebcf8128
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium007.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium007",
+  "question": "What is the difference between the number of policies issued by the Ministry of Industry and Information Technology in the ministerial policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd. and the number of local policies issued by the Guangdong Provincial People's Government for the industry of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 1.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Yi Hai Chang Jin Shang Wu Co., Ltd.'s industry is Leasing and Business Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Leasing and Business Services, the Ministry of Industry and Information Technology policy count is 2",
+    "Extract from company_profile.csv that Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.'s industry is Health and Social Work",
+    "Extract from policy_release_status.csv that in the local policies for Health and Social Work, the Guangdong Provincial People's Government policy count is 1",
+    "Calculate the difference: 2 - 1 = 1.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Leasing and Business Services ministerial policies_Ministry of Industry and Information Technology policy count": 2,
+    "Industry of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.": "Health and Social Work",
+    "Health and Social Work local policies_Guangdong Provincial People's Government policy count": 1,
+    "Difference (Ministry of Industry and Information Technology policy count - Guangdong Provincial People's Government policy count)": 1.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium008.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium008.json
new file mode 100644
index 0000000000000000000000000000000000000000..061b409dfdc85080868f702de55e322cfeff63c1
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium008.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium008",
+  "question": "Which is greater: the number of local policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd., or the number of policies issued by the General Office of the State Council in the State Council policies for the industry of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yi Hai Chang Jin Shang Wu Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Yi Hai Chang Jin Shang Wu Co., Ltd.'s industry is Leasing and Business Services",
+    "Extract from policy_release_status.csv that the local policies policy count for Leasing and Business Services is 12",
+    "Extract from company_profile.csv that Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.'s industry is Health and Social Work",
+    "Extract from policy_release_status.csv that in Health and Social Work State Council policies, the policy count issued by the General Office of the State Council is 1",
+    "Compare 12 and 1; 12 is larger; output \"Yi Hai Chang Jin Shang Wu Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Leasing and Business Services local policies policy count": 12,
+    "Industry of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.": "Health and Social Work",
+    "Health and Social Work State Council policies_General Office of the State Council policy count": 1,
+    "Comparison result": "12 is larger; output Yi Hai Chang Jin Shang Wu Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium009.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium009.json
new file mode 100644
index 0000000000000000000000000000000000000000..b1a74fb089a2aee8af693e38f85474f8040d30b9
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium009.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium009",
+  "question": "Which is greater: the number of local policies issued by the Guangdong Province General Office of the People's Government in the local policies for the industry of Zhang Qiao Jin Chuang Technology Co., Ltd., or the number of central ministry/agency policies issued by the National Cryptography Administration in the ministerial policies for the industry of Heng Li Ke Zhi Ruan Jian Co., Ltd.?",
+  "guidelines": "The answer must be either \"Number of local policies issued by the Guangdong Province General Office of the People's Government for the industry of Zhang Qiao Jin Chuang Technology Co., Ltd.\" or \"Number of central ministry/agency policies issued by the National Cryptography Administration for the industry of Heng Li Ke Zhi Ruan Jian Co., Ltd.\" or \"Equal\". Output only the selected answer text, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhang Qiao Jin Chuang Technology Co., Ltd.'s industry is Consumer Electronics and Electrical Equipment",
+    "From policy_release_status.csv, extract that in the local policies for Consumer Electronics and Electrical Equipment, the number of policies issued by the Guangdong Province General Office of the People's Government is 1",
+    "From company_profile.csv, extract that Heng Li Ke Zhi Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "From policy_release_status.csv, extract that in the ministerial policies for Information Transmission, Software and IT Services, the number of policies issued by the National Cryptography Administration is 1",
+    "Compare 1 and 1; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Consumer Electronics and Electrical Equipment",
+    "Consumer Electronics and Electrical Equipment local policies_Guangdong Province General Office of the People's Government number of policies": 1,
+    "Industry of Heng Li Ke Zhi Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies_National Cryptography Administration number of policies": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium010.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium010.json
new file mode 100644
index 0000000000000000000000000000000000000000..edec9515164f5663bb45851508a8d93467e16aa0
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium010.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium010",
+  "question": "Which is greater: the number of local policies issued by the Hainan Province Department of Finance for the industry of Zhang Qiao Jin Chuang Technology Co., Ltd., or the number of central ministry/agency policies issued by the National Health Commission for the industry of Heng Li Ke Zhi Ruan Jian Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"Equal\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhang Qiao Jin Chuang Technology Co., Ltd.'s industry is Consumer Electronics and Electrical Equipment",
+    "Extract from policy_release_status.csv that in the local policies for Consumer Electronics and Electrical Equipment, the policy count issued by the Hainan Province Department of Finance is 1",
+    "Extract from company_profile.csv that Heng Li Ke Zhi Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Information Transmission, Software and IT Services, the policy count issued by the National Health Commission is 1",
+    "Compare 1 and 1; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Consumer Electronics and Electrical Equipment",
+    "Consumer Electronics and Electrical Equipment local policies_Hainan Province Department of Finance policy count": 1,
+    "Industry of Heng Li Ke Zhi Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies_National Health Commission policy count": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium011.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium011.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd943013daad30c6309c9e56cc9bf67b3d6236bc
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium011.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium011",
+  "question": "What is the difference between the number of central ministry/agency policies issued by the State Administration of Foreign Exchange in the ministerial policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. and the number of local policies issued by the Guangdong Provincial Committee of the CPC for the Financial Services industry in Guangdong Province where Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "From policy_release_status.csv, extract that in the ministerial policies for Information Transmission, Software and IT Services, the policy count issued by the State Administration of Foreign Exchange is 1",
+    "From company_profile.csv, extract that Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located in Guangdong Province and its industry is Financial Services",
+    "From policy_release_status.csv, filter by province = Guangdong Province and industry = Financial Services, and extract that in local policies, the policy count issued by the Guangdong Provincial Committee of the CPC is 1",
+    "Compute the difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies_State Administration of Foreign Exchange policy count": 1,
+    "Province of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Financial Services local policies_Guangdong Provincial Committee of the CPC policy count": 1,
+    "Difference (State Administration of Foreign Exchange policy count - Guangdong Provincial Committee of the CPC policy count)": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium012.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium012.json
new file mode 100644
index 0000000000000000000000000000000000000000..2b404953e91ecd3ed27e4003dd2af7ae87c3ea44
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium012.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium012",
+  "question": "Which is greater: the number of central ministry/agency policies issued by the People's Bank of China in the ministerial policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd., or the number of local policies issued by the Guangdong Provincial Committee of the CPC for the Financial Services industry in Guangdong Province where Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"Equal\"; output only the selected answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "From policy_release_status.csv, extract that in the ministerial policies for Information Transmission, Software and IT Services, the policy count issued by the People's Bank of China is 1",
+    "From company_profile.csv, extract that Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located in Guangdong Province and its industry is Financial Services",
+    "From policy_release_status.csv, filter by province = Guangdong Province and industry = Financial Services, and extract that in local policies, the policy count issued by the Guangdong Provincial Committee of the CPC is 1",
+    "Compare 1 and 1; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies_People's Bank of China policy count": 1,
+    "Province of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Financial Services local policies_Guangdong Provincial Committee of the CPC policy count": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium013.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium013.json
new file mode 100644
index 0000000000000000000000000000000000000000..2091a0fd87c6ec4f1b25a2e7bea92e0017a4119b
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium013.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium013",
+  "question": "Which is greater: the number of local policies issued by the Yunnan Province Development and Reform Commission for the industry of Bei Kong Ze Jing Water Co., Ltd., or the number of local policies issued by the Shanghai Municipality Science and Technology Commission for the Health and Social Work industry in Shanghai Municipality where Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"Equal\"; output only the selected answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Bei Kong Ze Jing Water Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Bei Kong Ze Jing Water Co., Ltd.'s industry is Water Conservancy, Environment and Public Facilities Management",
+    "Extract from policy_release_status.csv that in the local policies for Water Conservancy, Environment and Public Facilities Management, the policy count issued by the Yunnan Province Development and Reform Commission is 2",
+    "Extract from company_profile.csv that Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd. is located in Shanghai Municipality and its industry is Health and Social Work",
+    "Extract from policy_release_status.csv that in the local policies for Health and Social Work in Shanghai Municipality, the policy count issued by the Shanghai Municipality Science and Technology Commission is 1",
+    "Compare 2 and 1; 2 is greater; output \"Bei Kong Ze Jing Water Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.": "Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management local policies_Yunnan Province Development and Reform Commission policy count": 2,
+    "Province of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Health and Social Work local policies_Shanghai Municipality Science and Technology Commission policy count": 1,
+    "Comparison result": "2 is greater; output Bei Kong Ze Jing Water Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium014.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium014.json
new file mode 100644
index 0000000000000000000000000000000000000000..82d14590f51de69c7f6014ab02ee121fb055d1dd
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium014.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium014",
+  "question": "Which is higher: the number of central ministry/agency policies issued by the Ministry of Agriculture and Rural Affairs in the ministerial policies for the industry of Bei Kong Ze Jing Water Co., Ltd., or the number of local policies issued by the Shanghai Municipality Health Commission for the Health and Social Work industry in Shanghai Municipality where Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Bei Kong Ze Jing Water Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Bei Kong Ze Jing Water Co., Ltd.'s industry is Water Conservancy, Environment and Public Facilities Management",
+    "Extract from policy_release_status.csv that in the ministerial policies for this industry, the policy count issued by the Ministry of Agriculture and Rural Affairs is 2",
+    "Extract from company_profile.csv that Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd. is located in Shanghai Municipality and its industry is Health and Social Work",
+    "Extract from policy_release_status.csv that in the local policies for Health and Social Work in Shanghai Municipality, the policy count issued by the Shanghai Municipality Health Commission is 1",
+    "Compare 2 and 1; 2 is higher; output \"Bei Kong Ze Jing Water Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.": "Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management ministerial policies_Ministry of Agriculture and Rural Affairs policy count": 2,
+    "Province of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Health and Social Work local policies_Shanghai Municipality Health Commission policy count": 1,
+    "Comparison result": "2 is higher; output Bei Kong Ze Jing Water Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium015.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium015.json
new file mode 100644
index 0000000000000000000000000000000000000000..f36363e2ee98efb1098583ed60c308abe89deec6
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium015.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium015",
+  "question": "Which is greater: the number of central ministry/agency policies issued by the State Administration of Foreign Exchange in the ministerial policies for the industry of Wu Li Hui Da Lian Suo Co., Ltd., or the number of local policies for the Water Conservancy, Environment and Public Facilities Management industry in Guangdong Province where Bei Kong Ze Jing Water Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"Equal\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Wu Li Hui Da Lian Suo Co., Ltd.'s industry is Wholesale and Retail",
+    "From policy_release_status.csv, extract that in the ministerial policies for Wholesale and Retail, the policy count issued by the State Administration of Foreign Exchange is 1",
+    "From company_profile.csv, extract that Bei Kong Ze Jing Water Co., Ltd. is located in Guangdong Province and its industry is Water Conservancy, Environment and Public Facilities Management",
+    "From policy_release_status.csv, filter by province = Guangdong Province and industry = Water Conservancy, Environment and Public Facilities Management, and extract that the local policies policy count is 1",
+    "Compare 1 and 1; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Wu Li Hui Da Lian Suo Co., Ltd.": "Wholesale and Retail",
+    "Wholesale and Retail ministerial policies_State Administration of Foreign Exchange policy count": 1,
+    "Province of Bei Kong Ze Jing Water Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Water Conservancy, Environment and Public Facilities Management local policies policy count": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium016.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium016.json
new file mode 100644
index 0000000000000000000000000000000000000000..87c50b94c2376b4a79b296de8a0befdd742bbd91
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium016.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium016",
+  "question": "Which is greater: the number of local policies issued by the Shenzhen Municipal People's Government for the industry of Wu Li Hui Da Lian Suo Co., Ltd., or the total number of policies for the Water Conservancy, Environment and Public Facilities Management industry in Guangdong Province where Bei Kong Ze Jing Water Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"Equal\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Wu Li Hui Da Lian Suo Co., Ltd.'s industry is Wholesale and Retail",
+    "From policy_release_status.csv, extract that in the local policies for Wholesale and Retail, the policy count issued by the Shenzhen Municipal People's Government is 1",
+    "From company_profile.csv, extract that Bei Kong Ze Jing Water Co., Ltd. is located in Guangdong Province and its industry is Water Conservancy, Environment and Public Facilities Management",
+    "From policy_release_status.csv, extract that the total policy count for Water Conservancy, Environment and Public Facilities Management in Guangdong Province is 1",
+    "Compare 1 and 1; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Wu Li Hui Da Lian Suo Co., Ltd.": "Wholesale and Retail",
+    "Wholesale and Retail local policies_Shenzhen Municipal People's Government policy count": 1,
+    "Province of Bei Kong Ze Jing Water Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Water Conservancy, Environment and Public Facilities Management total policy count": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium017.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium017.json
new file mode 100644
index 0000000000000000000000000000000000000000..a4b1aae7e107240dd94f0348d539c34abe4891a2
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium017.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium017",
+  "question": "Which is greater: the number of central ministry/agency policies issued by the National Health Commission in the ministerial policies for the industry of Bei Kong Ze Jing Water Co., Ltd., or the number of local policies issued by the Guangdong Provincial Committee of the CPC for the Financial Services industry in Guangdong Province where Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Bei Kong Ze Jing Water Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Bei Kong Ze Jing Water Co., Ltd.'s industry is Water Conservancy, Environment and Public Facilities Management",
+    "Extract from policy_release_status.csv that in the ministerial policies for this industry, the policy count issued by the National Health Commission is 2",
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located in Guangdong Province and its industry is Financial Services",
+    "Extract from policy_release_status.csv that in Guangdong Province Financial Services local policies, the policy count issued by the Guangdong Provincial Committee of the CPC is 1",
+    "Compare 2 and 1; 2 is greater; output \"Bei Kong Ze Jing Water Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.": "Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management ministerial policies_National Health Commission policy count": 2,
+    "Province of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Financial Services local policies_Guangdong Provincial Committee of the CPC policy count": 1,
+    "Comparison result": "2 is greater; output Bei Kong Ze Jing Water Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium018.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium018.json
new file mode 100644
index 0000000000000000000000000000000000000000..eebb9435d246a13059d666b4aba4bf70028cf5a4
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium018.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium018",
+  "question": "Which is greater: the number of local policies issued by the Liaoning Province People's Government for the industry of Bei Kong Ze Jing Water Co., Ltd., or the number of local policies issued by the Guangdong Province General Office of the People's Government for the Financial Services industry in Guangdong Province where Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Hui Jin Jin Rui Cai Fu Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Bei Kong Ze Jing Water Co., Ltd.'s industry is Water Conservancy, Environment and Public Facilities Management",
+    "Extract from policy_release_status.csv that in the local policies for this industry, the policy count issued by the Liaoning Province People's Government is 1",
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located in Guangdong Province and its industry is Financial Services",
+    "Extract from policy_release_status.csv that in the local policies for Financial Services in Guangdong Province, the policy count issued by the Guangdong Province General Office of the People's Government is 2",
+    "Compare 1 and 2; 2 is greater; output \"Hui Jin Jin Rui Cai Fu Management Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.": "Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management local policies_Liaoning Province People's Government policy count": 1,
+    "Province of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Financial Services local policies_Guangdong Province General Office of the People's Government policy count": 2,
+    "Comparison result": "2 is greater; output Hui Jin Jin Rui Cai Fu Management Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium019.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium019.json
new file mode 100644
index 0000000000000000000000000000000000000000..966de25acae7e3814a34e91fdebbc366f9296c35
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium019.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium019",
+  "question": "What is the difference between the number of local policies issued by the Shaanxi Province Development and Reform Commission for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. and the total number of policies for the Conglomerates industry in Guangdong Province where Hua Cheng Sheng Yuan Integrated Development Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -1.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for this industry in Shaanxi Province, the policy count issued by the Shaanxi Province Development and Reform Commission is 1",
+    "Extract from company_profile.csv that Hua Cheng Sheng Yuan Integrated Development Co., Ltd. is located in Guangdong Province and its industry is Conglomerates",
+    "Extract from policy_release_status.csv that the total policy count for Conglomerates in Guangdong Province is 2",
+    "Calculate the difference: 1 - 2 = -1.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies_Shaanxi Province Development and Reform Commission policy count": 1,
+    "Province of Hua Cheng Sheng Yuan Integrated Development Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Conglomerates total policy count": 2,
+    "Difference (Shaanxi Province Development and Reform Commission policy count - total policy count)": -1.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium020.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium020.json
new file mode 100644
index 0000000000000000000000000000000000000000..93c16ba98b21790e2547969b4afc99079db86159
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium020.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium020",
+  "question": "Which is greater: the number of central ministry/agency policies issued by the Ministry of Housing and Urban-Rural Development in the ministerial policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd., or the number of local policies for the Conglomerates industry in Guangdong Province where Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"Equal\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for this industry, the policy count issued by the Ministry of Housing and Urban-Rural Development is 2",
+    "Extract from company_profile.csv that Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. is located in Guangdong Province and its industry is Conglomerates",
+    "Extract from policy_release_status.csv that in Guangdong Province, the Conglomerates local policies policy count is 2",
+    "Compare 2 and 2; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies_Ministry of Housing and Urban-Rural Development policy count": 2,
+    "Province of Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Conglomerates local policies policy count": 2,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium021.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium021.json
new file mode 100644
index 0000000000000000000000000000000000000000..1ef5b34393ef96a966cff7351afa89f9aab910d3
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium021.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium021",
+  "question": "Which is greater: the number of specific local policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd., or the number of similar local policies for the industry of Bao Xin Hui Hui Wang Luo Co., Ltd. in its province?",
+  "guidelines": "The answer must be a company name or \"industry\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Bao Xin Hui Hui Wang Luo Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for this industry, the policy count issued by the Sichuan Province People's Government is 2",
+    "Extract from company_profile.csv that Bao Xin Hui Hui Wang Luo Co., Ltd. is located in Beijing Municipality and its industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for this industry in Beijing Municipality, the policy count issued by the Beijing Municipality Bureau of Economy and Information Technology is 7",
+    "Compare 2 and 7; 7 is greater; output \"Bao Xin Hui Hui Wang Luo Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies_Sichuan Province People's Government policy count": 2,
+    "Province of Bao Xin Hui Hui Wang Luo Co., Ltd.": "Beijing Municipality",
+    "Beijing Municipality Information Transmission, Software and IT Services local policies_Beijing Municipality Bureau of Economy and Information Technology policy count": 7,
+    "Comparison result": "7 is greater; output Bao Xin Hui Hui Wang Luo Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium022.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium022.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d6c9289d49f71f769f175e6759ab12a20193e4c
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium022.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium022",
+  "question": "Between Zhong Ke Ke Shu Ruan Jian Co., Ltd. and Bao Xin Hui Hui Wang Luo Co., Ltd., which obtains a greater number of local policy supports?",
+  "guidelines": "The answer must be a company name or \"Equal\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for this industry, the policy count issued by the Shenzhen Municipality Science and Technology Innovation Commission is 3",
+    "Extract from company_profile.csv that Bao Xin Hui Hui Wang Luo Co., Ltd. is located in Beijing Municipality and its industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for this industry in Beijing Municipality, the policy count issued by the Beijing Municipality Science and Technology Commission is 3",
+    "Compare 3 and 3; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies_Shenzhen Municipality Science and Technology Innovation Commission policy count": 3,
+    "Province of Bao Xin Hui Hui Wang Luo Co., Ltd.": "Beijing Municipality",
+    "Beijing Municipality Information Transmission, Software and IT Services local policies_Beijing Municipality Science and Technology Commission policy count": 3,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium023.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium023.json
new file mode 100644
index 0000000000000000000000000000000000000000..86f1326a79de1cddcab2fc9e5ee40d7f1a0bbaf9
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium023.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium023",
+  "question": "What is the difference between the number of local policies issued by the Hunan Province General Office of the People's Government for the industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. and the total number of policies for the Consumer Electronics and Electrical Equipment industry in Guangdong Province where Zhang Qiao Jin Chuang Technology Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.'s industry is Real Estate",
+    "Extract from policy_release_status.csv that in the local policies for Real Estate, the policy count issued by the Hunan Province General Office of the People's Government is 1",
+    "Extract from company_profile.csv that Zhang Qiao Jin Chuang Technology Co., Ltd. is located in Guangdong Province and its industry is Consumer Electronics and Electrical Equipment",
+    "Extract from policy_release_status.csv that the total policy count for Consumer Electronics and Electrical Equipment in Guangdong Province is 1",
+    "Calculate the difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.": "Real Estate",
+    "Real Estate local policies_Hunan Province General Office of the People's Government policy count": 1,
+    "Province of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Consumer Electronics and Electrical Equipment total policy count": 1,
+    "Difference (Hunan Province General Office of the People's Government policy count - total policy count)": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium024.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium024.json
new file mode 100644
index 0000000000000000000000000000000000000000..855246f5afa9f74b0672901a09698f8e186d07e3
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium024.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium024",
+  "question": "Which is greater: the number of central ministry/agency policies issued by the Ministry of Science and Technology in the ministerial policies for the industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd., or the number of local policies for the Consumer Electronics and Electrical Equipment industry in Guangdong Province where Zhang Qiao Jin Chuang Technology Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.'s industry is Real Estate",
+    "Extract from policy_release_status.csv that in the ministerial policies for Real Estate, the policy count issued by the Ministry of Science and Technology is 1",
+    "Extract from company_profile.csv that Zhang Qiao Jin Chuang Technology Co., Ltd. is located in Guangdong Province and its industry is Consumer Electronics and Electrical Equipment",
+    "Extract from policy_release_status.csv that the local policies policy count for Consumer Electronics and Electrical Equipment in Guangdong Province is 1",
+    "Compare 1 and 1; if equal, output \"Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.\" as required by the question"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.": "Real Estate",
+    "Real Estate ministerial policies_Ministry of Science and Technology policy count": 1,
+    "Province of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Consumer Electronics and Electrical Equipment local policies policy count": 1,
+    "Comparison result": "Equal; per question requirement, output Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium025.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium025.json
new file mode 100644
index 0000000000000000000000000000000000000000..874a4145e4f7e1d13bcaddbaae61eeb770a8d765
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium025.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium025",
+  "question": "What is the difference between the number of local policies issued by the Anhui Province People's Government for the industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd. and the total number of local policies for the Consumer Electronics and Electrical Equipment industry in Guangdong Province where Zhang Qiao Jin Chuang Technology Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Ying Tai Sheng Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the local policies for Financial Services, the policy count issued by the Anhui Province People's Government is 1",
+    "Extract from company_profile.csv that Zhang Qiao Jin Chuang Technology Co., Ltd. is located in Guangdong Province and its industry is Consumer Electronics and Electrical Equipment",
+    "Extract from policy_release_status.csv that the local policies policy count for Consumer Electronics and Electrical Equipment in Guangdong Province is 1",
+    "Calculate the difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services local policies_Anhui Province People's Government policy count": 1,
+    "Province of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Consumer Electronics and Electrical Equipment local policies policy count": 1,
+    "Difference (Anhui Province People's Government policy count - local policies policy count)": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium026.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium026.json
new file mode 100644
index 0000000000000000000000000000000000000000..d7cfee70bed8d81d86ae0f4ca1802ff8efd11494
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium026.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium026",
+  "question": "What is the difference between the number of central ministry/agency policies issued by the General Office of the China Banking and Insurance Regulatory Commission in the ministerial policies for the industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd. and the number of local policies for the Consumer Electronics and Electrical Equipment industry in Guangdong Province where Zhang Qiao Jin Chuang Technology Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Ying Tai Sheng Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Financial Services, the policy count issued by the General Office of the China Banking and Insurance Regulatory Commission is 1",
+    "Extract from company_profile.csv that Zhang Qiao Jin Chuang Technology Co., Ltd. is located in Guangdong Province and its industry is Consumer Electronics and Electrical Equipment",
+    "Extract from policy_release_status.csv that the local policies policy count for Consumer Electronics and Electrical Equipment in Guangdong Province is 1",
+    "Calculate the difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_General Office of the China Banking and Insurance Regulatory Commission policy count": 1,
+    "Province of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Consumer Electronics and Electrical Equipment local policies policy count": 1,
+    "Difference (General Office of the China Banking and Insurance Regulatory Commission policy count - local policies policy count)": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium027.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium027.json
new file mode 100644
index 0000000000000000000000000000000000000000..83c7522c2bb9703daa5e47ede1b9e1781e1fb708
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium027.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium027",
+  "question": "Between the number of local policies issued by the General Office of the People's Government of Henan Province in the local policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd. and the number of local policies issued by the Shanghai Municipality Finance Bureau in the local policies for the industry of Lang Ji Hui Ruan Technology Co., Ltd. in Shanghai, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Lang Ji Hui Ruan Technology Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Yi Hai Chang Jin Shang Wu Co., Ltd.'s industry is Leasing and Business Services",
+    "Extract from policy_release_status.csv that in the local policies for Leasing and Business Services, the policy count issued by Henan Province General Office of the People's Government is 1",
+    "Extract from company_profile.csv that Lang Ji Hui Ruan Technology Co., Ltd. is located in Shanghai Municipality and its industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for Information Transmission, Software and IT Services in Shanghai Municipality, the policy count issued by the Shanghai Municipality Finance Bureau is 2",
+    "Compare 1 and 2; since 2 is greater, output \"Lang Ji Hui Ruan Technology Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Leasing and Business Services local policies_Henan Province General Office of the People's Government policy count": 1,
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.": "Shanghai Municipality",
+    "Information Transmission, Software and IT Services local policies in Shanghai Municipality_Shanghai Municipality Finance Bureau policy count": 2,
+    "Comparison result": "2 is greater; output Lang Ji Hui Ruan Technology Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium028.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium028.json
new file mode 100644
index 0000000000000000000000000000000000000000..c16b324438c27e677d012586008148c08a5ee06c
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium028.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium028",
+  "question": "What is the difference between the number of central ministry/agency policies issued by the China Federation of Logistics & Purchasing in the ministerial policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd. and the number of local policies for the Information Transmission, Software and IT Services industry in Shanghai Municipality where Lang Ji Hui Ruan Technology Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -14.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Yi Hai Chang Jin Shang Wu Co., Ltd.'s industry is Leasing and Business Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Leasing and Business Services, the policy count issued by the China Federation of Logistics & Purchasing is 1",
+    "Extract from company_profile.csv that Lang Ji Hui Ruan Technology Co., Ltd. is located in Shanghai Municipality and its industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that the local policies policy count for Information Transmission, Software and IT Services in Shanghai Municipality is 15",
+    "Calculate the difference: 1 - 15 = -14.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Leasing and Business Services ministerial policies_China Federation of Logistics & Purchasing policy count": 1,
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Information Transmission, Software and IT Services local policies policy count": 15,
+    "Difference (China Federation of Logistics & Purchasing policy count - local policies policy count)": -14.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium029.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium029.json
new file mode 100644
index 0000000000000000000000000000000000000000..1b6861cfd71c1bfabedfc863b759ba863a3186ac
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium029.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium029",
+  "question": "Between the number of central ministry/agency policies in the ministerial policies for the industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd. and the number of local policies in the local policies for the industry of Hua Cheng Sheng Yuan Integrated Development Co., Ltd. in Guangdong Province, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Zhao Ye Hua Chang Real Estate Development Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhao Ye Hua Chang Real Estate Development Co., Ltd.'s industry is Real Estate",
+    "Extract from policy_release_status.csv that the ministerial policies policy count for Real Estate is 5",
+    "Extract from company_profile.csv that Hua Cheng Sheng Yuan Integrated Development Co., Ltd. is located in Guangdong Province and its industry is Integrated",
+    "Extract from policy_release_status.csv that the local policies policy count for Integrated in Guangdong Province is 2",
+    "Compare 5 and 2; since 5 is greater, output \"Zhao Ye Hua Chang Real Estate Development Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd.": "Real Estate",
+    "Real Estate ministerial policies policy count": 5,
+    "Province of Hua Cheng Sheng Yuan Integrated Development Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Integrated local policies policy count": 2,
+    "Comparison result": "5 is greater; output Zhao Ye Hua Chang Real Estate Development Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium030.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium030.json
new file mode 100644
index 0000000000000000000000000000000000000000..a4f3e6c20ce286f081be9087b51deaf594844c18
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium030.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium030",
+  "question": "What is the difference between the number of central ministry/agency policies issued by the Ministry of Housing and Urban-Rural Development in the ministerial policies for the industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd. and the number of policies for the Integrated industry in Guangdong Province where Hua Cheng Sheng Yuan Integrated Development Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -1.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhao Ye Hua Chang Real Estate Development Co., Ltd.'s industry is Real Estate",
+    "Extract from policy_release_status.csv that in the ministerial policies for Real Estate, the policy count issued by the Ministry of Housing and Urban-Rural Development is 1",
+    "Extract from company_profile.csv that Hua Cheng Sheng Yuan Integrated Development Co., Ltd. is located in Guangdong Province and its industry is Integrated",
+    "Extract from policy_release_status.csv that the policy count for Integrated in Guangdong Province is 2",
+    "Calculate the difference: 1 - 2 = -1.0"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd.": "Real Estate",
+    "Real Estate ministerial policies_Ministry of Housing and Urban-Rural Development policy count": 1,
+    "Province of Hua Cheng Sheng Yuan Integrated Development Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Integrated policy count": 2,
+    "Difference (Ministry of Housing and Urban-Rural Development policy count - policy count)": -1.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium031.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium031.json
new file mode 100644
index 0000000000000000000000000000000000000000..26226e5275c5828cfc4c5c8178f026ba30915caa
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium031.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium031",
+  "question": "What is the difference between the number of central ministry/agency policies issued by the Ministry of Culture and Tourism in the ministerial policies for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the number of local policies issued by the Chengdu Municipality Bureau of Economy and Information Technology in the local policies for the Commercial Electrical Machinery and Equipment Manufacturing industry?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Financial Services, the policy count issued by the Ministry of Culture and Tourism is 1",
+    "Extract from policy_release_status.csv that in the local policies for Commercial Electrical Machinery and Equipment Manufacturing, the policy count issued by the Chengdu Municipality Bureau of Economy and Information Technology is 1",
+    "Calculate the difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_Ministry of Culture and Tourism policy count": 1,
+    "Commercial Electrical Machinery and Equipment Manufacturing local policies_Chengdu Municipality Bureau of Economy and Information Technology policy count": 1,
+    "Difference (Ministry of Culture and Tourism policy count - Chengdu Municipality Bureau of Economy and Information Technology policy count)": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium032.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium032.json
new file mode 100644
index 0000000000000000000000000000000000000000..796459f53f132022134368e405da02885ae1c71b
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium032.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium032",
+  "question": "Between the number of local policies related to Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the corresponding number of local policies for the Commercial Electrical Machinery and Equipment Manufacturing industry, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Hui Jin Jin Rui Cai Fu Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the local policies for Financial Services, the policy count issued by Henan Province General Office of the People's Government is 2",
+    "Extract from policy_release_status.csv that in the local policies for Commercial Electrical Machinery and Equipment Manufacturing, the policy count issued by the Hainan Province Department of Industry and Information Technology is 1",
+    "Compare 2 and 1; since 2 is greater, output \"Hui Jin Jin Rui Cai Fu Management Co., Ltd.\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services local policies_Henan Province General Office of the People's Government policy count": 2,
+    "Commercial Electrical Machinery and Equipment Manufacturing local policies_Hainan Province Department of Industry and Information Technology policy count": 1,
+    "Comparison result": "2 is greater; output Hui Jin Jin Rui Cai Fu Management Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium033.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium033.json
new file mode 100644
index 0000000000000000000000000000000000000000..e18895c69b00d965315d78ffadc99e15df669fed
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium033.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium033",
+  "question": "Between the number of local policies issued by the General Office of the Guangxi Zhuang Autonomous Regional People's Government in the local policies for the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. and the number of local policies issued by the Guizhou Province Department of Housing and Urban-Rural Development in the local policies for the Information Transmission, Software and IT Services industry, which is greater?",
+  "guidelines": "The answer must be \"Equal\", a company name, or the word \"industry\"; output only one word or company name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.'s industry is Education",
+    "Extract from policy_release_status.csv that in the local policies for Education, the policy count issued by the General Office of the Guangxi Zhuang Autonomous Regional People's Government is 1",
+    "Extract from policy_release_status.csv that in the local policies for Information Transmission, Software and IT Services, the policy count issued by the Guizhou Province Department of Housing and Urban-Rural Development is 1",
+    "Compare 1 and 1; if they are equal, output \"Equal\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.": "Education",
+    "Education local policies_General Office of the Guangxi Zhuang Autonomous Regional People's Government policy count": 1,
+    "Information Transmission, Software and IT Services local policies_Guizhou Province Department of Housing and Urban-Rural Development policy count": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium034.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium034.json
new file mode 100644
index 0000000000000000000000000000000000000000..5c6a732eb53bf8a9e6929627c0095686e936e981
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium034.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium034",
+  "question": "Between the number of policies issued by the Yunnan Province Department of Industry and Information Technology in the local policies for the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. and the number of policies issued by the Provincial General Office of the People's Government in the local policies for the Information Transmission, Software and IT Services industry, which is greater?",
+  "guidelines": "The answer must be \"Number of policies for the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. issued by the Yunnan Province Department of Industry and Information Technology\", \"Number of policies for the Information Transmission, Software and IT Services industry issued by the Provincial General Office of the People's Government\", or \"Equal\". Output only the specified answer text, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.'s industry is Education",
+    "Extract from policy_release_status.csv that in the local policies for Education, the policy count issued by the Yunnan Province Department of Industry and Information Technology is 1",
+    "Extract from policy_release_status.csv that in the local policies for Information Transmission, Software and IT Services, the policy count issued by the Provincial General Office of the People's Government is 1",
+    "Compare 1 and 1; if they are equal, output \"Equal\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.": "Education",
+    "Education local policies_Yunnan Province Department of Industry and Information Technology policy count": 1,
+    "Information Transmission, Software and IT Services local policies_Provincial General Office of the People's Government policy count": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium035.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium035.json
new file mode 100644
index 0000000000000000000000000000000000000000..32f27b7e8da154c10e142b4ef1b5a0c619893ae4
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium035.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium035",
+  "question": "What is the difference between the number of local policies issued by the Shanghai Municipality Commission of Economy and Information Technology in the local policies for the industry of Rui Xing Jian Kang Zhi Yao Co., Ltd. and the number of local policies issued by the Liaoning Province People's Government in the local policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Rui Xing Jian Kang Zhi Yao Co., Ltd.'s industry is Pharmaceutical Manufacturing",
+    "Extract from policy_release_status.csv that in the local policies for Pharmaceutical Manufacturing, the policy count issued by the Shanghai Municipality Commission of Economy and Information Technology is 3",
+    "Extract from policy_release_status.csv that in the local policies for Cultural, Arts, Sports and Entertainment Goods Manufacturing, the policy count issued by the Liaoning Province People's Government is 1",
+    "Calculate the difference: 3 - 1 = 2.0"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Rui Xing Jian Kang Zhi Yao Co., Ltd.": "Pharmaceutical Manufacturing",
+    "Pharmaceutical Manufacturing local policies_Shanghai Municipality Commission of Economy and Information Technology policy count": 3,
+    "Cultural, Arts, Sports and Entertainment Goods Manufacturing local policies_Liaoning Province People's Government policy count": 1,
+    "Difference (Shanghai Municipality Commission of Economy and Information Technology policy count - Liaoning Province People's Government policy count)": 2.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium036.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium036.json
new file mode 100644
index 0000000000000000000000000000000000000000..a585df0c455fd55f014f5f412c2fb7a5969249c2
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium036.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium036",
+  "question": "Between the number of local policies issued by the Shanghai Municipality Commission of Economy and Information Technology in the local policies for the industry of Rui Xing Jian Kang Zhi Yao Co., Ltd. and the number of central ministry/agency policies issued by the Ministry of Agriculture and Rural Affairs in the ministerial policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Rui Xing Jian Kang Zhi Yao Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Rui Xing Jian Kang Zhi Yao Co., Ltd.'s industry is Pharmaceutical Manufacturing",
+    "Extract from policy_release_status.csv that in the local policies for Pharmaceutical Manufacturing, the policy count issued by the Shanghai Municipality Commission of Economy and Information Technology is 3",
+    "Extract from policy_release_status.csv that in the ministerial policies for Cultural, Arts, Sports and Entertainment Goods Manufacturing, the policy count issued by the Ministry of Agriculture and Rural Affairs is 1",
+    "Compare 3 and 1; since 3 is greater, output \"Rui Xing Jian Kang Zhi Yao Co., Ltd.\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Rui Xing Jian Kang Zhi Yao Co., Ltd.": "Pharmaceutical Manufacturing",
+    "Pharmaceutical Manufacturing local policies_Shanghai Municipality Commission of Economy and Information Technology policy count": 3,
+    "Cultural, Arts, Sports and Entertainment Goods Manufacturing ministerial policies_Ministry of Agriculture and Rural Affairs policy count": 1,
+    "Comparison result": "3 is greater; output Rui Xing Jian Kang Zhi Yao Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium037.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium037.json
new file mode 100644
index 0000000000000000000000000000000000000000..7637c188c3ee19e672b3b337c567f2649dd3340f
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium037.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium037",
+  "question": "Between the number of local policies issued by the Secretariat of the Office of Nansha Development Zone Administrative Committee, Guangzhou in the local policies for the industry of Bao Xin Hui Hui Wang Luo Co., Ltd. and the number of local policies issued by the Shanghai Municipality Finance Bureau in the local policies for the General Equipment Manufacturing industry, which is greater?",
+  "guidelines": "The answer must be \"Number of local policies for the industry of Bao Xin Hui Hui Wang Luo Co., Ltd. issued by the Secretariat of the Office of Nansha Development Zone Administrative Committee, Guangzhou\" or \"Number of local policies for the General Equipment Manufacturing industry issued by the Shanghai Municipality Finance Bureau\". Output only the policy title, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "General Equipment Manufacturing",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Bao Xin Hui Hui Wang Luo Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for Information Transmission, Software and IT Services, the policy count issued by the Secretariat of the Office of Nansha Development Zone Administrative Committee, Guangzhou is 1",
+    "Extract from policy_release_status.csv that in the local policies for General Equipment Manufacturing, the policy count issued by the Shanghai Municipality Finance Bureau is 1",
+    "Compare 1 and 1; since they are equal, output \"General Equipment Manufacturing\" according to the question requirement"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Bao Xin Hui Hui Wang Luo Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies_Secretariat of the Office of Nansha Development Zone Administrative Committee, Guangzhou policy count": 1,
+    "General Equipment Manufacturing local policies_Shanghai Municipality Finance Bureau policy count": 1,
+    "Comparison result": "Equal; according to the question requirement, output General Equipment Manufacturing"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium038.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium038.json
new file mode 100644
index 0000000000000000000000000000000000000000..15d0e9e1f55a766bb37540605bd9f0f6a271b41b
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium038.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium038",
+  "question": "What is the difference between the number of central ministry/agency policies issued by the General Office of the China National Intellectual Property Administration in the ministerial policies for the industry of Bao Xin Hui Hui Wang Luo Co., Ltd. and the number of local policies issued by the Sichuan Province People's Government in the local policies for the General Equipment Manufacturing industry?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 1.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Bao Xin Hui Hui Wang Luo Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Information Transmission, Software and IT Services, the policy count issued by the General Office of the China National Intellectual Property Administration is 2",
+    "Extract from policy_release_status.csv that in the local policies for General Equipment Manufacturing, the policy count issued by the Sichuan Province People's Government is 1",
+    "Calculate the difference: 2 - 1 = 1.0"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Bao Xin Hui Hui Wang Luo Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies_General Office of the China National Intellectual Property Administration policy count": 2,
+    "General Equipment Manufacturing local policies_Sichuan Province People's Government policy count": 1,
+    "Difference (General Office of the China National Intellectual Property Administration policy count - Sichuan Province People's Government policy count)": 1.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium039.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium039.json
new file mode 100644
index 0000000000000000000000000000000000000000..295ae150a1449d2cb382f8ff1b1cba72c700d78e
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium039.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium039",
+  "question": "Between the number of local policies issued by the Hefei Municipality Office of the People's Government in the local policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. and the number of local policies issued by the Shandong Province People's Government in the local policies for the Scientific Research and Technical Services industry, which is greater?",
+  "guidelines": "The answer must be \"Number of local policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. issued by the Hefei Municipality Office of the People's Government\" or \"Number of local policies for the Scientific Research and Technical Services industry issued by the Shandong Province People's Government\". Output only the policy title, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Scientific Research and Technical Services",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for Information Transmission, Software and IT Services, the policy count issued by the Hefei Municipality Office of the People's Government is 2",
+    "Extract from policy_release_status.csv that in the local policies for Scientific Research and Technical Services, the policy count issued by the Shandong Province People's Government is 1",
+    "Compare 2 and 1; although 2 is greater, follow the question's requirement and output \"Scientific Research and Technical Services\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies_Hefei Municipality Office of the People's Government policy count": 2,
+    "Scientific Research and Technical Services local policies_Shandong Province People's Government policy count": 1,
+    "Comparison result": "2 is greater; according to the question statement, output Scientific Research and Technical Services"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium040.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium040.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba5deedb719154379ed5c8dcfa2bbe942932da14
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium040.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium040",
+  "question": "What is the difference between the number of local policies issued by the Guangzhou Development District Bureau of Economy and Information Technology in the local policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. and the number of local policies issued by the Yunnan Province General Office of the People's Government in the local policies for the Scientific Research and Technical Services industry?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for Information Transmission, Software and IT Services, the policy count issued by the Guangzhou Development District Bureau of Economy and Information Technology is 2",
+    "Extract from policy_release_status.csv that in the local policies for Scientific Research and Technical Services, the policy count issued by the Yunnan Province General Office of the People's Government is 2",
+    "Calculate the difference: 2 - 2 = 0.0"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies_Guangzhou Development District Bureau of Economy and Information Technology policy count": 2,
+    "Scientific Research and Technical Services local policies_Yunnan Province General Office of the People's Government policy count": 2,
+    "Difference (Guangzhou Development District Bureau of Economy and Information Technology policy count - Yunnan Province General Office of the People's Government policy count)": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium041.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium041.json
new file mode 100644
index 0000000000000000000000000000000000000000..105bfaa30776d897338af64c4fe10054d450937b
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium041.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium041",
+  "question": "Between the number of central ministry/agency policies issued by the Ministry of Public Security for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the number of local policies for the Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing industry in Anhui Province, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Hui Jin Jin Rui Cai Fu Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Financial Services, the policy count issued by the Ministry of Public Security is 1",
+    "Extract from policy_release_status.csv that in the local policies for Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing in Anhui Province, the policy count is 1",
+    "Compare 1 and 1; since they are equal, output \"Hui Jin Jin Rui Cai Fu Management Co., Ltd.\" according to the question requirement"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_Ministry of Public Security policy count": 1,
+    "Anhui Province Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing local policies policy count": 1,
+    "Comparison result": "Equal; according to the question requirement, output Hui Jin Jin Rui Cai Fu Management Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium042.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium042.json
new file mode 100644
index 0000000000000000000000000000000000000000..28226b33495bc212cb3569bf1613ed080b548409
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium042.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium042",
+  "question": "Between the number of central ministry/agency policies issued by the General Administration of Sport of China in the ministerial policies for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the total number of policies for the Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing industry in Anhui Province, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Hui Jin Jin Rui Cai Fu Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Financial Services, the policy count issued by the General Administration of Sport of China is 1",
+    "Extract from policy_release_status.csv that in Anhui Province, the policy count for the Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing industry is 1",
+    "Compare 1 and 1; since they are equal, output \"Hui Jin Jin Rui Cai Fu Management Co., Ltd.\" according to the question requirement"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_General Administration of Sport of China policy count": 1,
+    "Anhui Province Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing policy count": 1,
+    "Comparison result": "Equal; according to the question requirement, output Hui Jin Jin Rui Cai Fu Management Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium043.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium043.json
new file mode 100644
index 0000000000000000000000000000000000000000..66178644e8407a663073def90dbe43f70122bbd5
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium043.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium043",
+  "question": "Between the number of local policies issued by the Hunan Province People's Government in the local policies for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the number of policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry in Guangdong Province, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Hui Jin Jin Rui Cai Fu Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the local policies for Financial Services, the policy count issued by the Hunan Province People's Government is 1",
+    "Extract from policy_release_status.csv that in Guangdong Province, the policy count for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry is 1",
+    "Compare 1 and 1; since they are equal, output \"Hui Jin Jin Rui Cai Fu Management Co., Ltd.\" according to the question requirement"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services local policies_Hunan Province People's Government policy count": 1,
+    "Guangdong Province Cultural, Arts, Sports and Entertainment Goods Manufacturing policy count": 1,
+    "Comparison result": "Equal; according to the question requirement, output Hui Jin Jin Rui Cai Fu Management Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium044.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium044.json
new file mode 100644
index 0000000000000000000000000000000000000000..3792364ffe0f73e404c9b8629a5b9784f940d6c5
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium044.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium044",
+  "question": "Between the number of central ministry/agency policies issued by the State Administration of Foreign Exchange in the ministerial policies for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the number of local policies issued by the Shenzhen Municipality People's Government in the local policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry in Guangdong Province, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Hui Jin Jin Rui Cai Fu Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Financial Services, the policy count issued by the State Administration of Foreign Exchange is 1",
+    "Extract from policy_release_status.csv that in the local policies for Cultural, Arts, Sports and Entertainment Goods Manufacturing in Guangdong Province, the policy count issued by the Shenzhen Municipality People's Government is 1",
+    "Compare 1 and 1; since they are equal, output \"Hui Jin Jin Rui Cai Fu Management Co., Ltd.\" according to the question requirement"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_State Administration of Foreign Exchange policy count": 1,
+    "Guangdong Province Cultural, Arts, Sports and Entertainment Goods Manufacturing local policies_Shenzhen Municipality People's Government policy count": 1,
+    "Comparison result": "Equal; according to the question requirement, output Hui Jin Jin Rui Cai Fu Management Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium045.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium045.json
new file mode 100644
index 0000000000000000000000000000000000000000..bad51d5d5c979e532b074a39e2a49479f38468d8
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium045.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium045",
+  "question": "Is the number of central ministry/agency policies issued by the Ministry of Housing and Urban-Rural Development in the ministerial policies for the industry of Zhao Ye Ze Jin Real Estate Holdings Co., Ltd. greater than the total number of policies for the Metal Smelting and Rolling Processing industry in Gansu Province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhao Ye Ze Jin Real Estate Holdings Co., Ltd.'s industry is Real Estate",
+    "Extract from policy_release_status.csv that in the ministerial policies for Real Estate, the policy count issued by the Ministry of Housing and Urban-Rural Development is 1",
+    "Extract from policy_release_status.csv that in Gansu Province, the policy count for the Metal Smelting and Rolling Processing industry is 1",
+    "Determine whether 1 is greater than 1; since it is not, output \"No\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Zhao Ye Ze Jin Real Estate Holdings Co., Ltd.": "Real Estate",
+    "Real Estate ministerial policies_Ministry of Housing and Urban-Rural Development policy count": 1,
+    "Gansu Province Metal Smelting and Rolling Processing industry policy count": 1,
+    "Whether greater (Ministry of Housing and Urban-Rural Development policy count > industry policy count)": "No"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium046.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium046.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e2d8e978f6fef54d977e9253dcc33aa4900a18d
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium046.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium046",
+  "question": "Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.industry's Ministry of Housing and Urban-Rural DevelopmentNumber of policies (indicator) and Gansu Province Development and Reform CommissionNumber of policies (indicator)what is the gap?",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.industry = Real Estate",
+    "Extracted from policy_release_status.csv: Real Estateministerial policies in Ministry of Housing and Urban-Rural DevelopmentNumber of policies (indicator)=1",
+    "policy_release_status.csv in by province=Gansu Province、industry=extractlocal policies in Gansu ProvinceDevelopment and Reform CommissionNumber of policies (indicator)=1",
+    "Calculate difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.industry": "Real Estate",
+    "Real Estateministerial policies_Ministry of Housing and Urban-Rural DevelopmentNumber of policies (indicator)": 1,
+    "Gansu Provincelocal policies_Gansu ProvinceDevelopment and Reform CommissionNumber of policies (indicator)": 1,
+    "Difference(Ministry of Housing and Urban-Rural DevelopmentNumber of policies (indicator)-Gansu ProvinceDevelopment and Reform CommissionNumber of policies (indicator))": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium047.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium047.json
new file mode 100644
index 0000000000000000000000000000000000000000..8762e648db29f5b533d78d1633d69cd8251870a9
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium047.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium047",
+  "question": "Between the number of local policies issued by the Chengdu Municipality Bureau of Economy and Information Technology in the local policies for the industry of Hua Xin Yuan Shi New Materials Co., Ltd. and the number of local policies issued by the Gansu Province General Office of the People's Government in the local policies for the Pharmaceutical Manufacturing industry in Gansu Province, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"Equal\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Xin Yuan Shi New Materials Co., Ltd.'s industry is Non-metallic Mineral Products",
+    "Extract from policy_release_status.csv that in the local policies for Non-metallic Mineral Products, the policy count issued by the Chengdu Municipality Bureau of Economy and Information Technology is 1",
+    "Extract from policy_release_status.csv that in the local policies for Pharmaceutical Manufacturing in Gansu Province, the policy count issued by the Gansu Province General Office of the People's Government is 1",
+    "Compare 1 and 1; if they are equal, output \"Equal\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Hua Xin Yuan Shi New Materials Co., Ltd.": "Non-metallic Mineral Products",
+    "Non-metallic Mineral Products local policies_Chengdu Municipality Bureau of Economy and Information Technology policy count": 1,
+    "Pharmaceutical Manufacturing local policies_Gansu Province General Office of the People's Government policy count": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium048.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium048.json
new file mode 100644
index 0000000000000000000000000000000000000000..0940abd19a6090ba2f88b86e28f74529eee57832
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium048.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium048",
+  "question": "Between the number of policies issued by the Shandong Province Department of Industry and Information Technology in the local policies for the industry of Hua Xin Yuan Shi Xin Cai Liao Co., Ltd. and the number of policies for the Pharmaceutical Manufacturing industry in Gansu Province, which is higher?",
+  "guidelines": "The answer must be a company name or the word \"Equal\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Xin Yuan Shi Xin Cai Liao Co., Ltd.'s industry is Non-metallic Mineral Products",
+    "Extract from policy_release_status.csv that in the local policies for Non-metallic Mineral Products, the policy count issued by the Shandong Province Department of Industry and Information Technology is 1",
+    "Extract from policy_release_status.csv that in Gansu Province, the policy count for the Pharmaceutical Manufacturing industry is 1",
+    "Compare 1 and 1; if they are equal, output \"Equal\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Hua Xin Yuan Shi Xin Cai Liao Co., Ltd.": "Non-metallic Mineral Products",
+    "Non-metallic Mineral Products local policies_Shandong Province Department of Industry and Information Technology policy count": 1,
+    "Gansu Province Pharmaceutical Manufacturing policy count": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium049.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium049.json
new file mode 100644
index 0000000000000000000000000000000000000000..000ba6aa19f7e5e20ddab833f069cd9fe34aa68d
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium049.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium049",
+  "question": "Between the number of local policies for the industry of Wan Hui Jin Sheng Real Estate Development Co., Ltd. and the number of local policies issued by the Jiangxi Province People's Government for the Commercial Electrical Machinery and Equipment Manufacturing industry in Jiangxi Province, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Wan Hui Jin Sheng Real Estate Development Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Wan Hui Jin Sheng Real Estate Development Co., Ltd.'s industry is Real Estate",
+    "Extract from policy_release_status.csv that the local policies policy count for Real Estate is 2",
+    "Extract from policy_release_status.csv that in the local policies for Commercial Electrical Machinery and Equipment Manufacturing in Jiangxi Province, the policy count issued by the Jiangxi Province People's Government is 1",
+    "Compare 2 and 1; since 2 is greater, output \"Wan Hui Jin Sheng Real Estate Development Co., Ltd.\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Wan Hui Jin Sheng Real Estate Development Co., Ltd.": "Real Estate",
+    "Real Estate local policies policy count": 2,
+    "Jiangxi Province Commercial Electrical Machinery and Equipment Manufacturing local policies_Jiangxi Province People's Government policy count": 1,
+    "Comparison result": "2 is greater; output Wan Hui Jin Sheng Real Estate Development Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium050.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium050.json
new file mode 100644
index 0000000000000000000000000000000000000000..a7317f413efe75aba0c547e6f544da74610fe8c9
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium050.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium050",
+  "question": "Wan Hui Jin Sheng Fang Di Chan Kai Fa Co., Ltd.industry's Shandong ProvinceNumber of policies (indicator) and Jiangxi Province Number of policies (indicator)how much？",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Wan Hui Jin Sheng Fang Di Chan Kai Fa Co., Ltd.industry = Real Estate",
+    "Extracted from policy_release_status.csv: Real Estatelocal policies in Shandong ProvincepersonsNumber of policies (indicator)=1",
+    "policy_release_status.csv in by province=Jiangxi Province、industry=extractlocal policies in Jiangxi ProvincepersonsNumber of policies (indicator)=1",
+    "Calculate difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Wan Hui Jin Sheng Fang Di Chan Kai Fa Co., Ltd.industry": "Real Estate",
+    "Real Estatelocal policies_Shandong ProvincepersonsNumber of policies (indicator)": 1,
+    "Jiangxi Provincelocal policies_Jiangxi ProvincepersonsNumber of policies (indicator)": 1,
+    "Difference(Shandong ProvincepersonsNumber of policies (indicator)-Jiangxi ProvincepersonsNumber of policies (indicator))": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium051.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium051.json
new file mode 100644
index 0000000000000000000000000000000000000000..8bbdbcf9de053ab20ca6cac4e8d83fc2dbc246a2
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium051.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium051",
+  "question": "Wu Li Hui Da Chain Co., Ltd.industry's ministerial policies_Number of policies (indicator) and China Shenzhen CitypersonsNumber of policies (indicator)compared with difference how much？",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Wu Li Hui Da Chain Co., Ltd.industry = Retail",
+    "Extracted from policy_release_status.csv: Retailministerial policies in Number of policies (indicator)=1",
+    "Extracted from policy_release_status.csv: ChinaConglomerateslocal policies in Shenzhen CitypersonsNumber of policies (indicator)=1",
+    "Calculate difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Wu Li Hui Da Chain Co., Ltd.industry": "Retail",
+    "Retailministerial policies_Number of policies (indicator)": 1,
+    "ChinaConglomerateslocal policies_Shenzhen CitypersonsNumber of policies (indicator)": 1,
+    "Difference(Number of policies (indicator)-Shenzhen CitypersonsNumber of policies (indicator))": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium052.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium052.json
new file mode 100644
index 0000000000000000000000000000000000000000..3c0c69295236702fefb901f5aaff8afba0f6b78d
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium052.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium052",
+  "question": "Between the number of local policies issued by the Sichuan Province People's Government in the local policies for the industry of Wu Li Hui Da Chain Co., Ltd. and the number of central ministry/agency policies in the China conglomerates category, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Wu Li Hui Da Chain Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Wu Li Hui Da Chain Co., Ltd.'s industry is Wholesale and Retail",
+    "Extract from policy_release_status.csv that in the local policies for Wholesale and Retail, the policy count issued by the Sichuan Province People's Government is 1",
+    "Extract from policy_release_status.csv that in the China conglomerates central ministry/agency policies, the policy count is 1",
+    "Compare 1 and 1; since they are equal, output \"Wu Li Hui Da Chain Co., Ltd.\" according to the question requirement"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Wu Li Hui Da Chain Co., Ltd.": "Wholesale and Retail",
+    "Wholesale and Retail local policies_Sichuan Province People's Government policy count": 1,
+    "China conglomerates central ministry/agency policies policy count": 1,
+    "Comparison result": "Equal; according to the question requirement, output Wu Li Hui Da Chain Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium053.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium053.json
new file mode 100644
index 0000000000000000000000000000000000000000..79e18c0aeaed7034f1effd0123619e842cb56899
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium053.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium053",
+  "question": "Between the number of policies issued by the State Administration of Foreign Exchange in the ministerial policies for the industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd. and the number of local policies issued by the Chongqing Municipality General Office of the People's Government in the local policies for the Financial Services industry in China, which is greater?",
+  "guidelines": "The answer must be \"Equal\", a company name, or the word \"industry\"; output only one word or company name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Chongqing Municipality General Office of the People's Government",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Ying Tai Sheng Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Financial Services, the policy count issued by the State Administration of Foreign Exchange is 1",
+    "Extract from policy_release_status.csv that in the China Financial Services local policies, the policy count issued by the Chongqing Municipality General Office of the People's Government is 2",
+    "Compare 1 and 2; since 2 is greater, output \"Chongqing Municipality General Office of the People's Government\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_State Administration of Foreign Exchange policy count": 1,
+    "China Financial Services local policies_Chongqing Municipality General Office of the People's Government policy count": 2,
+    "Comparison result": "2 is greater; according to the question requirement, output Chongqing Municipality General Office of the People's Government"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium054.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium054.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4d8e9ebd87cde61b357823b57284013d8537e3c
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium054.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium054",
+  "question": "Hua Ying Tai Sheng Wealth Management Co., Ltd.industry's Shanghai Municipalitypersons Number of policies (indicator) and China Number of policies issued by Ministry of Educationcompared with which unitsgreater？",
+  "guidelines": "The answer must a company name or \"industry\"\"Equal\", Output onlyname, without any explanation or description.Wu, answer\"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Hua Ying Tai Sheng Wealth Management Co., Ltd.industry = Financial Industry",
+    "Extracted from policy_release_status.csv: Financial Industrylocal policies in Shanghai MunicipalitypersonsNumber of policies (indicator)=1",
+    "Extracted from policy_release_status.csv: ChinaFinancial Industryministerial policies in EducationNumber of policies (indicator)=1",
+    "Compare1 and 1, Equalthen output\"Equal\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Hua Ying Tai Sheng Wealth Management Co., Ltd.industry": "Financial Industry",
+    "Financial Industrylocal policies_Shanghai MunicipalitypersonsNumber of policies (indicator)": 1,
+    "ChinaFinancial Industryministerial policies_EducationNumber of policies (indicator)": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium055.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium055.json
new file mode 100644
index 0000000000000000000000000000000000000000..35a0cbe247bea04b95b1efe11d4cc5e182258910
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium055.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium055",
+  "question": "Between the number of policies issued by the Ministry of Public Security in the ministerial policies for the industry of Tong Tong Ze Hong Securities Co., Ltd. and the number of policies issued by the State Administration of Foreign Exchange in the ministerial policies for the Leasing and Business Services industry in China, which is greater?",
+  "guidelines": "The answer must be \"Equal\", a company name, or the word \"industry\"; output only one word or company name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Tong Tong Ze Hong Securities Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Financial Services, the policy count issued by the Ministry of Public Security is 1",
+    "Extract from policy_release_status.csv that in the ministerial policies for Leasing and Business Services in China, the policy count issued by the State Administration of Foreign Exchange is 1",
+    "Compare 1 and 1; if they are equal, output \"Equal\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Industry of Tong Tong Ze Hong Securities Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_Ministry of Public Security policy count": 1,
+    "China Leasing and Business Services ministerial policies_State Administration of Foreign Exchange policy count": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium056.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium056.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eec231eaa15d87dd13a8089b9447760b8f02aef
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium056.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium056",
+  "question": "Tong Tong Ze Hong Zheng Quan Co., Ltd.industry's local policiesGuangzhou CitypersonsNumber of policies (indicator) and China ministerial policiesNational Development and Reform CommissionNumber of policies (indicator)compared with difference how much？",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Tong Tong Ze Hong Zheng Quan Co., Ltd.industry = Financial Industry",
+    "Extracted from policy_release_status.csv: Financial Industrylocal policies in Guangzhou CitypersonsNumber of policies (indicator)=1",
+    "Extracted from policy_release_status.csv: ChinaBusiness Servicesministerial policies in National Development and Reform CommissionNumber of policies (indicator)=1",
+    "Calculate difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Tong Tong Ze Hong Zheng Quan Co., Ltd.industry": "Financial Industry",
+    "Financial Industrylocal policies_Guangzhou CitypersonsNumber of policies (indicator)": 1,
+    "ChinaBusiness Servicesministerial policies_National Development and Reform CommissionNumber of policies (indicator)": 1,
+    "Difference(Guangzhou CitypersonsNumber of policies (indicator)-National Development and Reform CommissionNumber of policies (indicator))": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium057.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium057.json
new file mode 100644
index 0000000000000000000000000000000000000000..fa9ee4914247b3e6bb107877a7167456204d30c9
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium057.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium057",
+  "question": "Hua Ying Tai Sheng Wealth Management Co., Ltd.industry's Number of policies (indicator) and China Sichuan ProvinceNumber of policies (indicator)what is the gap?",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "answer": 2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Hua Ying Tai Sheng Wealth Management Co., Ltd.industry = Financial Industry",
+    "Extracted from policy_release_status.csv: Financial Industryministerial policies in Number of policies (indicator)=3",
+    "Extracted from policy_release_status.csv: ChinaTransportation、Postal Serviceslocal policies in Sichuan ProvinceNumber of policies (indicator)=1",
+    "Calculate difference: 3 - 1 = 2.0"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Hua Ying Tai Sheng Wealth Management Co., Ltd.industry": "Financial Industry",
+    "Financial Industryministerial policies_Number of policies (indicator)": 3,
+    "ChinaTransportation、Postal Serviceslocal policies_Sichuan ProvinceNumber of policies (indicator)": 1,
+    "Difference(Number of policies (indicator)-Sichuan ProvinceNumber of policies (indicator))": 2.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium058.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium058.json
new file mode 100644
index 0000000000000000000000000000000000000000..0994cf41da100f2554343eadae7f6ccf152b78a3
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium058.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium058",
+  "question": "Shi Yang Jin Jin Electrical Appliances Co., Ltd.province Guangdong ProvinceNumber of policies (indicator) and China Hainan ProvinceNumber of policies (indicator)compared with difference how much？",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Shi Yang Jin Jin Electrical Appliances Co., Ltd.province = Guangdong Province and industry = Transportation、Postal Services",
+    "policy_release_status.csv in by province=Guangdong Province、industry=Transportation、Postal Servicesextractlocal policies in Guangdong ProvinceNumber of policies (indicator)=1",
+    "Extracted from policy_release_status.csv: ChinaFood and Beveragelocal policies in Hainan ProvinceNumber of policies (indicator)=1",
+    "Calculate difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Shi Yang Jin Jin Electrical Appliances Co., Ltd.province": "Guangdong Province",
+    "Guangdong ProvinceTransportation、Postal Serviceslocal policies_Guangdong ProvinceNumber of policies (indicator)": 1,
+    "ChinaFood and Beveragelocal policies_Hainan ProvinceNumber of policies (indicator)": 1,
+    "Difference(Guangdong ProvinceNumber of policies (indicator)-Hainan ProvinceNumber of policies (indicator))": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium059.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium059.json
new file mode 100644
index 0000000000000000000000000000000000000000..c37979960c36b434b0374f4757c44d8c75ed5833
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium059.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium059",
+  "question": "Between the number of policies issued by the Guangdong Province Development and Reform Commission in the local policies for the province where Shi Yang Jin Jin Electrical Appliances Co., Ltd. is located and the number of local policies issued by the Department of Digitalization and Future Industries in the Food and Beverage industry in China, which is greater?",
+  "guidelines": "The answer must be a company name, the word \"industry\", or \"Equal\"; output only one word or one company name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Shi Yang Jin Jin Electrical Appliances Co., Ltd. is located in Guangdong Province and its industry is Transportation, Warehousing and Postal Services",
+    "Extract from policy_release_status.csv that in the local policies for Transportation, Warehousing and Postal Services in Guangdong Province, the policy count issued by the Guangdong Province Development and Reform Commission is 1",
+    "Extract from policy_release_status.csv that in the China Food and Beverage local policies, the policy count issued by the Department of Digitalization and Future Industries is 1",
+    "Compare 1 and 1; if they are equal, output \"Equal\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Shi Yang Jin Jin Electrical Appliances Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Transportation, Warehousing and Postal Services local policies_Guangdong Province Development and Reform Commission policy count": 1,
+    "China Food and Beverage local policies_Department of Digitalization and Future Industries policy count": 1,
+    "Comparison result": "Equal"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium060.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium060.json
new file mode 100644
index 0000000000000000000000000000000000000000..2871427f37dd28a810ec48efa14eabe487cb54a4
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium060.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium060",
+  "question": "Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.province Number of policies (indicator) and China  in Guangdong ProvinceNumber of policies (indicator)compared with which unitsgreater？",
+  "guidelines": "The answer must a company name or \"industry\", Output onlyname, without any explanation or description.Wu, answer\"No relevant data found\".",
+  "answer": "Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.province = Shanghai Municipality and industry = ",
+    "policy_release_status.csv in by province=Shanghai Municipality、industry=extractNumber of policies (indicator)=11",
+    "Extracted from policy_release_status.csv: ChinaSemiconductor Industrylocal policies in  in Guangdong ProvinceNumber of policies (indicator)=1",
+    "Compare11 and 1, 11greater, output\"Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.province": "Shanghai Municipality",
+    "Shanghai MunicipalityNumber of policies (indicator)": 11,
+    "ChinaSemiconductor Industrylocal policies_ in Guangdong ProvinceNumber of policies (indicator)": 1,
+    "Comparison result": "11greater，outputJian Fan Ning Ze Yang Lao Fu Wu Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium061.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium061.json
new file mode 100644
index 0000000000000000000000000000000000000000..4eab7cb313c42b59b51b1643f0913298ae8f02a5
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium061.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium061",
+  "question": "Hua Cheng Sheng Yuan Integrated Development Co., Ltd.province Guangzhou CityHuang Pu DistrictNumber of policies (indicator) and China Number of policies (indicator)compared with which unitsgreater？",
+  "guidelines": "The answer must a company name or \"industry\", Output onlyname, without any explanation or description.Wu, answer\"No relevant data found\".",
+  "answer": "Hua Cheng Sheng Yuan Integrated Development Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Hua Cheng Sheng Yuan Integrated Development Co., Ltd.province = Guangdong Province and industry = Information Transmission, Software and IT Services",
+    "policy_release_status.csv in by province=Guangdong Province、industry=Information Transmission, Software and IT Servicesextractlocal policies in Guangzhou CityHuang Pu DistrictNumber of policies (indicator)=2",
+    "Extracted from policy_release_status.csv: ChinaSemiconductor Industrylocal policies in Number of policies (indicator)=1",
+    "Compare2 and 1, 2greater, output\"Hua Cheng Sheng Yuan Integrated Development Co., Ltd.\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Hua Cheng Sheng Yuan Integrated Development Co., Ltd.province": "Guangdong Province",
+    "Guangdong ProvinceInformation Transmission, Software and IT Serviceslocal policies_Guangzhou CityHuang Pu DistrictNumber of policies (indicator)": 2,
+    "ChinaSemiconductor Industrylocal policies_Number of policies (indicator)": 1,
+    "Comparison result": "2greater，outputHua Cheng Sheng Yuan Integrated Development Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium062.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium062.json
new file mode 100644
index 0000000000000000000000000000000000000000..eaf8574d954c3a068a6360e4801a5deac74fb7e3
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium062.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium062",
+  "question": "What is the difference between the number of local policies issued by the Guangdong Provincial Committee of the CPC in the local policies for the industry of Hua Cheng Sheng Yuan Integrated Development Co., Ltd. and the number of local policies issued by the Fujian Province Department of Industry and Information Technology in the local policies for the Semiconductor Industry in China?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Cheng Sheng Yuan Integrated Development Co., Ltd. is located in Guangdong Province and its industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for Information Transmission, Software and IT Services in Guangdong Province, the policy count issued by the Guangdong Provincial Committee of the CPC is 1",
+    "Extract from policy_release_status.csv that in the China Semiconductor Industry local policies, the policy count issued by the Fujian Province Department of Industry and Information Technology is 1",
+    "Calculate the difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Hua Cheng Sheng Yuan Integrated Development Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Information Transmission, Software and IT Services local policies_Guangdong Provincial Committee of the CPC policy count": 1,
+    "China Semiconductor Industry local policies_Fujian Province Department of Industry and Information Technology policy count": 1,
+    "Difference (Guangdong Provincial Committee of the CPC policy count - Fujian Province Department of Industry and Information Technology policy count)": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium063.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium063.json
new file mode 100644
index 0000000000000000000000000000000000000000..ebdcdcf38c60de2edd92331947cb61f5ceb79346
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium063.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium063",
+  "question": "Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.province Number of policies (indicator) and China TransportationNumber of policies (indicator)compared with  how much？",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.province = Guangdong Province and industry = Electronics",
+    "policy_release_status.csv in by province=Guangdong Province、industry=Electronicsextractlocal policies in Guangdong ProvincepersonsNumber of policies (indicator)=1",
+    "Extracted from policy_release_status.csv: ChinaPlastic Productsministerial policies in TransportationNumber of policies (indicator)=1",
+    "Calculate difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.province": "Guangdong Province",
+    "Guangdong ProvinceElectronicslocal policies_Guangdong ProvincepersonsNumber of policies (indicator)": 1,
+    "ChinaPlastic Productsministerial policies_TransportationNumber of policies (indicator)": 1,
+    "Difference(Guangdong ProvincepersonsNumber of policies (indicator)-TransportationNumber of policies (indicator))": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium064.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium064.json
new file mode 100644
index 0000000000000000000000000000000000000000..7cb3a8adec112f325ed3d1f4df356dab83993308
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium064.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium064",
+  "question": "Is the number of specific local policies in the province where Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd. is located the same as the number of policies issued by the Hainan Province Department of Industry and Information Technology for China's rubber and plastic products industry local policies?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd. is in Guangdong Province and its industry is Consumer Electronics and Electrical Equipment",
+    "From policy_release_status.csv, filter by province = Guangdong Province and industry = Consumer Electronics and Electrical Equipment, and extract the number of local policies issued by the Guangdong Province General Office of the People's Government: 1",
+    "From policy_release_status.csv, extract the number of China rubber and plastic products industry local policies issued by the Hainan Province Department of Industry and Information Technology: 1",
+    "Compare whether 1 equals 1; if so, output \"Yes\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Consumer Electronics and Electrical Equipment local policies_Guangdong Province General Office of the People's Government number of policies": 1,
+    "China rubber and plastic products industry local policies_Hainan Province Department of Industry and Information Technology number of policies": 1,
+    "Whether the same": "Yes"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium065.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium065.json
new file mode 100644
index 0000000000000000000000000000000000000000..54921265dd672423881f149f5defd6d5fc0ea3f4
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium065.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium065",
+  "question": "Zhong Ke Ke Shu Software Co., Ltd.province  Number of policies (indicator) and China Sichuan Provincepersons Number of policies (indicator)what is the gap?",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Zhong Ke Ke Shu Software Co., Ltd.province = Guangdong Province and industry = Communication Transmission Equipment",
+    "policy_release_status.csv in by province=Guangdong Province、industry=Communication Transmission Equipmentextractlocal policies in Guangdong ProvinceNumber of policies (indicator)=1",
+    "Extracted from policy_release_status.csv: ChinaMetal Productslocal policies in Sichuan ProvincepersonsNumber of policies (indicator)=1",
+    "Calculate difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Zhong Ke Ke Shu Software Co., Ltd.province": "Guangdong Province",
+    "Guangdong ProvinceCommunication Transmission Equipmentlocal policies_Guangdong ProvinceNumber of policies (indicator)": 1,
+    "ChinaMetal Productslocal policies_Sichuan ProvincepersonsNumber of policies (indicator)": 1,
+    "Difference(Guangdong ProvinceNumber of policies (indicator)-Sichuan ProvincepersonsNumber of policies (indicator))": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/enterprise_industry_policy_analysis/medium066.json b/assets/qa_gold/enterprise_industry_policy_analysis/medium066.json
new file mode 100644
index 0000000000000000000000000000000000000000..62264611b789cb44787445200644d4067d7e828f
--- /dev/null
+++ b/assets/qa_gold/enterprise_industry_policy_analysis/medium066.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium066",
+  "question": "Zhong Ke Ke Shu Software Co., Ltd.province Guangzhou CitypersonsNumber of policies (indicator) and China Hunan ProvincepersonsNumber of policies (indicator)compared with which unitsgreater？",
+  "guidelines": "The answer must a company name or \"industry\", Output onlyname, without any explanation or description.Wu, answer\"No relevant data found\".",
+  "answer": "Zhong Ke Ke Shu Software Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Zhong Ke Ke Shu Software Co., Ltd.province = Guangdong Province and industry = Communication Transmission Equipment",
+    "policy_release_status.csv in by province=Guangdong Province、industry=Communication Transmission Equipmentextractlocal policies in Guangzhou CitypersonsNumber of policies (indicator)=2",
+    "Extracted from policy_release_status.csv: ChinaMetal Productslocal policies in Hunan ProvincepersonsNumber of policies (indicator)=1",
+    "Compare2 and 1, 2greater, output\"Zhong Ke Ke Shu Software Co., Ltd.\""
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Zhong Ke Ke Shu Software Co., Ltd.province": "Guangdong Province",
+    "Guangdong ProvinceCommunication Transmission Equipmentlocal policies_Guangzhou CitypersonsNumber of policies (indicator)": 2,
+    "ChinaMetal Productslocal policies_Hunan ProvincepersonsNumber of policies (indicator)": 1,
+    "Comparison result": "2greater，outputZhong Ke Ke Shu Software Co., Ltd."
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard001.json b/assets/qa_gold/hypothesis_verification/hard001.json
new file mode 100644
index 0000000000000000000000000000000000000000..99b978054a8010218c0560d6055acb20025e57d1
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard001.json
@@ -0,0 +1,36 @@
+{
+  "id": "hard001",
+  "question": "作为研发驱动型产业，医药制造业的企业创新投入与收入表现之间的关联一直备受研究者关注，而地方政策环境可能影响这种关联的强弱。请以2022年数据为基础，将医药制造业上市企业按所在省份是否出台了地方生物医药产业发展促进政策(政策名称含生物医药，且含发展或促进)分为两组，分别计算两组企业的研发投入占比与营业收入同比增减幅之间的斯皮尔曼等级相关系数，并报告两个相关系数的差值（有政策省份系数减去无政策省份系数）。",
+  "guidelines": "依次回答出台政策省份的相关系数和未出台政策省份的相关系数及两者差值。相关系数保留4位小数，差值保留2位小数。如[\"0.2356\", \"-0.1048\", \"0.34\"]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    0.1142,
+    -0.2132,
+    0.33
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"医药\"的政策记录，找到80条医药相关政策，其中地方政策55条、部委/国务院政策25条。",
+    "从policy_resource.csv中读取并匹配上述地方政策，按“政策名称含生物医药，且含发展或促进”识别“地方生物医药产业发展促进政策”，共得到8条，政策id为：92、141、381、430、431、432、433、436。",
+    "根据这8条政策的省份归属，提取出台了生物医药产业促进政策的省份共7个：上海市（id:92）、江苏省（id:430）、云南省（id:141）、广东省（id:431）、安徽省（id:432）、浙江省（id:381、433）、天津市（id:436）。",
+    "从company_profile.csv筛选行业=\"医药制造业\"的企业，共449家。与company_operation_status.csv关联后，筛选研发投入占比和营业收入同比增减幅均非空的有效企业，共417家。",
+    "将417家有效企业按省份分为两组：有生物医药政策省份200家（上海市45家、广东省46家、浙江省45家、江苏省46家、天津市8家、云南省6家、安徽省4家），无生物医药政策省份217家。",
+    "分别计算两组企业研发投入占比与营业收入同比增减幅的斯皮尔曼等级相关系数：有政策省份r=0.1142（p=0.1075，不显著），无政策省份r=-0.2132（p=0.0016，显著）。",
+    "计算差值 = 0.1142 - (-0.2132) = 0.33。有政策省份研发投入与营收增长呈弱正相关（不显著），无政策省份呈显著负相关，差值为0.33。"
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "医药相关政策总数（条）": 80,
+    "标题含生物医药的地方政策数（条）": 8,
+    "涉及省份数（个）": 7,
+    "医药制造业有效企业数（家）": 417,
+    "有政策省份企业数（家）": 200,
+    "无政策省份企业数（家）": 217,
+    "有政策省份斯皮尔曼系数": 0.1142,
+    "无政策省份斯皮尔曼系数": -0.2132,
+    "相关系数差值": 0.33
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard002.json b/assets/qa_gold/hypothesis_verification/hard002.json
new file mode 100644
index 0000000000000000000000000000000000000000..dff5ec7e1ae316d702123bd8f662275cab5c6370
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard002.json
@@ -0,0 +1,33 @@
+{
+  "id": "hard002",
+  "question": "有观点认为，在推动消费电子及电气业数字化转型与智能制造升级方面，国家层面的产业政策与地方政府政策在表述深度和覆盖方向上存在系统性差异。请检验这一判断：针对所有与消费电子及电气业相关的政策文件，分别统计国家级政策（含国务院及各部委发布的政策）与地方级政策中，明确提出数字化转型或智能制造相关目标或措施的政策数量及其占各自总数的比例，并给出国家级覆盖率减去地方级覆盖率的差值（以百分点计）。",
+  "guidelines": "依次回答国家级政策覆盖率、地方级政策覆盖率、差值（国家级占比减去地方级占比）。覆盖率和差值均以百分点表示，保留2位小数。如[80.00, 71.43, 8.57]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    75.0,
+    66.67,
+    8.33
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv中筛选industry字段包含\"消费电子\"的政策记录，共找到10条消费电子及电气业相关政策。按policyClassification字段将其划分为国家级（国务院政策+部委政策）和地方级（地方政策）两类：国家级政策4条（id: 19、150、156、167），地方级政策6条（id: 15、89、96、253、303、375）。",
+    "从policy_resource.csv中读取上述10条政策的全文内容，逐条分析是否明确提出数字化转型或智能制造相关目标或措施（关键词包括：数字化转型、数字化、数智化、智能制造、智能化、工业互联网、两化融合等）。",
+    "对4条国家级政策全文进行内容分析：id=19（加快电力装备绿色低碳创新发展行动计划）正文极短，仅为转发通知，无数字化转型或智能制造相关目标措施，判定为不包含；id=150（工业和信息化部关于开展2022\"三品\"全国行活动的通知）明确提出\"加快推进数字化助力消费品工业\"三品\"战略实施\"，包含；id=156（五部门数字化助力消费品工业\"三品\"行动方案）全文以数字化为核心主题，包含；id=167（推进国家级质量标准实验室建设的指导意见）明确提出\"围绕质量管理数字化\"等数字化目标，包含。国家级政策中含数字化转型/智能制造目标的有3条。",
+    "对6条地方政策全文进行内容分析：id=15（广东省进一步促进工业经济平稳增长措施）明确提出推动企业开展\"高端化、智能化、绿色化技术改造\"，包含；id=89（江西省打造全国新兴产业培育发展高地实施方案）明确提出\"产业链核心环节数字化转型\"目标，包含；id=96（重庆市促进大中小企业融通发展工作方案）明确提出\"加快全产业链数字化、网络化转型\"，包含；id=253（成都市\"十四五\"制造业高质量发展规划）正文为简短转发通知，无数字化/智能制造相关目标措施，不包含；id=303（海南省激励企业上规模奖励资金管理实施细则）内容为企业产值达标奖励规则，无数字化转型内容，不包含；id=375（四川省承接制造业有序转移实施意见）明确提出\"推动传统劳动密集型产业向数字化、智能化、高端化转型升级\"，包含。地方政策中含数字化转型/智能制造目标的有4条。",
+    "计算国家级政策覆盖率：3 ÷ 4 × 100% = 75.00%；计算地方级政策覆盖率：4 ÷ 6 × 100% ≈ 66.67%；差值 = 75.00% - 66.67% = 8.33个百分点，国家级政策覆盖率高于地方级。"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "消费电子及电气业相关政策总数（条）": 10,
+    "国家级政策总数（条）": 4,
+    "国家级含数字化转型/智能制造政策数（条）": 3,
+    "地方级政策总数（条）": 6,
+    "地方级含数字化转型/智能制造政策数（条）": 4,
+    "国家级政策覆盖率（%）": 75.0,
+    "地方级政策覆盖率（%）": 66.67,
+    "覆盖率差值（百分点）": 8.33
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard003.json b/assets/qa_gold/hypothesis_verification/hard003.json
new file mode 100644
index 0000000000000000000000000000000000000000..9adf19c91648ae475dcac52fc646bf509ced9682
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard003.json
@@ -0,0 +1,37 @@
+{
+  "id": "hard003",
+  "question": "在2022年通用设备制造业上市企业中，计算有地方制造业创新与科技发展促进政策支撑的省份（有政策省份）与无政策省份的企业政府补贴金额与年度中国发明专利申请数之间的斯皮尔曼等级相关系数，并给出两组系数之差（有政策省份系数减无政策省份系数）。要求依次回答：有政策省份的相关系数、无政策省份的相关系数、两组系数之差（均保留4位小数）。",
+  "guidelines": "依次回答有政策省份的相关系数、无政策省份的相关系数和两者差值（有政策-无政策）。相关系数和差值均保留4位小数。如[\"0.6812\", \"0.3590\", \"0.3222\"]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    0.7531,
+    0.4245,
+    0.3285
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选policyClassification为\"地方政策\"且industry包含\"通用设备\"的政策记录，找到36条涉及通用设备制造业的地方政策。",
+    "从policy_resource.csv筛选政策正文涉及制造业创新与科技发展促进的政策，得到8条地方制造业创新与科技发展促进政策（id:75，78，175，238，385，541，562，590），涉及7个省份：上海市、宁夏回族自治区、安徽省、广东省、新疆维吾尔自治区、福建省、陕西省。",
+    "从policy_resource.csv中读取这8条政策的全文内容，分析发现这些政策主要涉及制造业创新中心建设（上海市、福建省）、专精特新中小企业倍增培育（安徽省）、科技型企业创新发展倍增（陕西省）、创新产品研制引导（宁夏回族自治区）、创新链产业链融合发展（广东省广州市）、未来产业创新高地建设（上海市）、技术创新中心建设（新疆维吾尔自治区）和长三角科技创新共同体建设（上海市）等方向，政策内容均强调对通用设备等先进制造业企业的创新支持和补贴引导。",
+    "从company_profile.csv筛选行业为\"通用设备制造业\"的企业，共213家。",
+    "从company_operation_status.csv提取这213家企业的政府奖励资金、补贴和年度中国发明专利申请数，筛选两项指标均非空的有效企业，得到189家。其中有政策省份44家，无政策省份145家。",
+    "计算有政策省份组（44家企业）政府补贴与年度中国发明专利申请数的斯皮尔曼等级相关系数 = 0.7531。",
+    "计算无政策省份组（145家企业）政府补贴与年度中国发明专利申请数的斯皮尔曼等级相关系数 = 0.4245。",
+    "两者差值 = 0.7531 - 0.4245 = 0.3285。有政策省份的政府补贴与专利产出之间呈现更强的正相关关系，表明在出台了创新科技促进政策的省份中，政府补贴对企业专利产出的激励效应显著更强。"
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "涉及通用设备的地方政策总数（条）": 36,
+    "创新科技主题地方政策数（条）": 8,
+    "涉及省份数（个）": 7,
+    "通用设备制造业企业总数（家）": 213,
+    "有效企业数（家）": 189,
+    "有政策省份有效企业数（家）": 44,
+    "无政策省份有效企业数（家）": 145,
+    "有政策省份斯皮尔曼相关系数": 0.7531,
+    "无政策省份斯皮尔曼相关系数": 0.4245
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard004.json b/assets/qa_gold/hypothesis_verification/hard004.json
new file mode 100644
index 0000000000000000000000000000000000000000..45aea544efd279435e5dffb1b3b2b312d4ade9bb
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard004.json
@@ -0,0 +1,37 @@
+{
+  "id": "hard004",
+  "question": "在2022年纺织鞋服业上市企业中对比政策支持对企业盈亏的影响，计算有地方制造业转型升级政策支撑的省份（有政策省份）与无政策省份的企业营业利润亏损占比，并给出两组占比之差（有政策省份占比减无政策省份占比，以百分点计）。要求依次回答：有政策省份亏损占比、无政策省份亏损占比、两组占比的差值（均保留2位小数）。",
+  "guidelines": "依次回答有政策省份亏损占比、无政策省份亏损占比和差值（有政策组占比减去无政策组占比）。占比以百分数表示，保留2位小数。如[38.46, 27.50, 10.96]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    43.75,
+    31.96,
+    11.79
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选涉及行业包含\"纺织鞋服业\"的地方政策，共找到14条。",
+    "从policy_resource.csv中读取这14条政策全文，对每条政策进行深度分析，判断其是否属于制造业转型升级类政策：含有\"转型升级\"、\"智能制造\"、\"绿色制造\"、\"制造业创新\"、\"数字化转型\"、\"技术改造\"、\"升级改造\"、\"高端化\"等核心政策目标词汇，且非以节能减排约束或落后产能退出为主要定位的政策。",
+    "经过政策内容分析，认定以下10条为制造业转型升级政策，并匹配policy_resource.csv中的id：广东省工业经济平稳增长措施（id=15）、福建省制造业创新中心名单（id=22）、上海市制造业创新中心建设工程实施方案（id=75）、成都市“十四五”制造业高质量发展规划（id=253）、四川省承接制造业有序转移实施意见（id=375）、广西壮族自治区强龙头壮产业行动（id=263）、山东省新旧动能转换重大产业攻关项目管理实施细则（id=274）、湖南省智能制造标杆示范行动实施方案（id=276）、河北省推进规模以上工业企业培育工作若干措施（id=409）、新疆维吾尔自治区技术创新中心建设工作指引（id=541）。涉及省份9个：广东省、福建省、上海市、四川省、广西壮族自治区、山东省、湖南省、河北省、新疆维吾尔自治区。",
+    "从company_profile.csv筛选industry='纺织鞋服业'的企业，共177家，分布于19个省份。",
+    "从company_operation_status.csv获取这177家企业的营业利润金额数据，所有企业营业利润均非空，有效企业共177家。",
+    "按是否在有政策省份（上海市、四川省、山东省、广东省、广西壮族自治区、新疆维吾尔自治区、河北省、湖南省、福建省）将177家有效企业分为两组：有政策组80家，无政策组97家。",
+    "计算各组亏损（营业利润金额<0）企业占比：有政策组亏损企业35家，占比=35/80×100%=43.75%；无政策组亏损企业31家，占比=31/97×100%=31.96%；差值=43.75%-31.96%=11.79个百分点。"
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "纺织鞋服业相关地方政策数（条）": 14,
+    "制造业转型升级政策数（条）": 10,
+    "有政策省份数（个）": 9,
+    "纺织鞋服业有效企业总数（家）": 177,
+    "有政策组有效企业数（家）": 80,
+    "无政策组有效企业数（家）": 97,
+    "有政策组亏损企业数（家）": 35,
+    "无政策组亏损企业数（家）": 31,
+    "有政策组亏损占比（%）": 43.75,
+    "无政策组亏损占比（%）": 31.96
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard005.json b/assets/qa_gold/hypothesis_verification/hard005.json
new file mode 100644
index 0000000000000000000000000000000000000000..a8cf8d46ed044670e6e8cb3594c4ebf1b4b7ec2c
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard005.json
@@ -0,0 +1,42 @@
+{
+  "id": "hard005",
+  "question": "在化学原料和化学制品制造业中，碳达峰与节能减排政策的推进可能给部分企业带来额外合规成本，由此引发一种反常现象：企业获得的政府补贴较高（以全体有效企业政府补贴金额的中位数作为划定高补贴的分界点），但利润同比却在下滑。本题以2022年度该行业的上市企业为分析对象。请分别计算出台了地方碳达峰或节能减排促进政策的省份、以及未出台此类政策的省份中，反常企业占各组有效企业总数的比例，以及两组比例之差（有政策省份减无政策省份，以百分点计）。",
+  "guidelines": "依次回答有政策省份的反常企业占比、无政策省份的反常企业占比和两组占比的差值。占比和差值均以百分数表示，保留2位小数。如[32.14, 21.05, 11.09]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "29.29",
+    "24.43",
+    "4.87"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"化学原料\"且policyClassification为\"地方政策\"的记录，找到48条化学原料相关地方政策。",
+    "从policy_resource.csv中读取这48条政策全文，按“碳达峰/节能减排/绿色低碳转型”口径筛得12条相关政策，policy id为：575、116、163、489、106、36、7、477、268、376、512、161。",
+    "上述12条政策主要包括碳达峰实施方案、节能减排综合工作方案、清洁生产推行方案和绿色低碳转型实施意见等；对应省份共10个：上海市、四川省、宁夏回族自治区、安徽省、江西省、河南省、湖南省、甘肃省、贵州省、辽宁省。",
+    "从company_profile.csv筛选行业=\"化学原料和化学制品制造业\"的企业共364家，与company_operation_status.csv关联获取运营数据。",
+    "筛选政府奖励资金、补贴和营业利润同比增减幅均非空的有效企业，共361家。其中有政策省份99家，无政策省份262家。",
+    "计算全行业361家有效企业的政府奖励资金、补贴中位数为10050282.75元。",
+    "在有政策省份99家企业中，政府补贴高于中位数的有56家，其中营业利润同比下滑的有29家，反常企业占比=29/99×100%=29.29%。",
+    "在无政策省份262家企业中，政府补贴高于中位数的有124家，其中营业利润同比下滑的有64家，反常企业占比=64/262×100%=24.43%。",
+    "两组占比的差值=29.29%-24.43%=4.87个百分点，有政策省份的反常占比反而更高。"
+  ],
+  "steps_num": 9,
+  "milestone": {
+    "化学原料地方政策总数（条）": 48,
+    "碳达峰/节能减排相关政策数（条）": 12,
+    "涉及省份数（个）": 10,
+    "全行业有效企业数（家）": 361,
+    "有政策省份有效企业数（家）": 99,
+    "无政策省份有效企业数（家）": 262,
+    "全行业政府补贴中位数（元）": 10050282.75,
+    "有政策省份高补贴企业数（家）": 56,
+    "有政策省份反常企业数（家）": 29,
+    "有政策省份反常占比（%）": 29.29,
+    "无政策省份高补贴企业数（家）": 124,
+    "无政策省份反常企业数（家）": 64,
+    "无政策省份反常占比（%）": 24.43
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard006.json b/assets/qa_gold/hypothesis_verification/hard006.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e97716f601ba697b3a0453f5a47c01735f32a41
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard006.json
@@ -0,0 +1,38 @@
+{
+  "id": "hard006",
+  "question": "在2022年铁路、船舶、航空航天和其他运输设备制造业上市企业中，计算有地方先进制造与装备产业促进政策支撑的省份（有政策省份）与无政策省份的企业总资产与累计中国发明专利授权数之间的斯皮尔曼等级相关系数，并给出两组系数之差（有政策省份系数减无政策省份系数，保留2位小数）。要求依次回答：有政策省份的相关系数、无政策省份的相关系数、两组系数的差值。",
+  "guidelines": "依次回答有政策省份的相关系数、无政策省份的相关系数和两者差值（有政策-无政策）。相关系数保留2位小数。如[\"0.63\", \"0.48\", \"0.15\"]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "0.56",
+    "0.72",
+    "-0.16"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"铁路\"或\"船舶\"或\"航空\"的政策记录，找到46条铁路、船舶、航空航天和其他运输设备制造业相关政策。",
+    "筛选其中policyClassification为\"地方政策\"的记录，得到36条地方政策。",
+    "从policy_resource.csv中读取这36条地方政策的全文内容，分析哪些政策涉及先进制造、装备制造、智能制造、首台套或重大技术装备等内容，筛得21条含相关内容的地方政策，policy id为：42、75、78、87、89、139、153、154、175、176、189、238、274、276、303、375、385、448、476、541、590；涉及13个省份：上海市、四川省、天津市、安徽省、山东省、广东省、新疆维吾尔自治区、江西省、河南省、海南省、湖南省、福建省、陕西省。",
+    "从company_profile.csv筛选industry=\"铁路、船舶、航空航天和其他运输设备制造业\"的企业，共99家。",
+    "关联company_operation_status.csv获取总资产和累计中国发明专利授权数，排除任一指标为空的企业后，得到96家有效企业。其中有政策省份36家，无政策省份60家。",
+    "分别计算两组企业总资产与累计中国发明专利授权数的斯皮尔曼等级相关系数：有政策省份ρ=0.5589≈0.56，无政策省份ρ=0.7216≈0.72。",
+    "计算差值：0.56-0.72=-0.16。无政策省份的规模-创新相关性反而更强。"
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "铁路船舶航空相关政策总数（条）": 46,
+    "地方政策数（条）": 36,
+    "含先进制造/装备相关内容的地方政策数（条）": 21,
+    "涉及省份数（个）": 13,
+    "行业企业总数（家）": 99,
+    "有效企业数（家）": 96,
+    "有政策省份有效企业数（家）": 36,
+    "无政策省份有效企业数（家）": 60,
+    "有政策省份斯皮尔曼相关系数": 0.56,
+    "无政策省份斯皮尔曼相关系数": 0.72,
+    "差值": -0.16
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard007.json b/assets/qa_gold/hypothesis_verification/hard007.json
new file mode 100644
index 0000000000000000000000000000000000000000..a051fe70ea93d86844c27eaaf684294691ec39b1
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard007.json
@@ -0,0 +1,45 @@
+{
+  "id": "hard007",
+  "question": "在2022年非金属矿物制品业上市企业中研究政策对不同规模企业营业利润率的影响，计算已出台专项推动建材行业碳达峰或节能减排政策的省份（有政策省份）中大型企业与小型企业的平均营业利润率差距（有政策省份规模差距），以及未出台此类政策的省份（无政策省份）中同一差距（无政策省份规模差距），并计算两者的差值（有政策省份规模差距减无政策省份规模差距）。要求依次回答：有政策省份规模差距、无政策省份规模差距、两类省份差距之差（均以百分点表示，保留2位小数）。",
+  "guidelines": "依次回答有政策省份规模差距、无政策省份规模差距、两类省份差距之差。均以百分点表示，保留2位小数。如[5.46, 1.23, 4.23]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    7.72,
+    0.69,
+    7.03
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"非金属矿物制品业\"的地方政策记录，共找到12条。",
+    "从policy_resource.csv中读取这12条政策的全文内容，分析政策是否涉及建材行业（水泥、玻璃、陶瓷等非金属矿物制品）的碳达峰、节能减排或绿色低碳要求；匹配到的政策id为：42、116、163、489、106、36、7、477、268、376、512、161。经全文分析，确认这些政策对应省份共9个：湖南省、河南省、四川省、江西省、辽宁省、宁夏回族自治区、甘肃省、贵州省、安徽省。",
+    "从company_profile.csv筛选industry=\"非金属矿物制品业\"的企业，共找到125家有效企业（总资产、营业利润金额、营业收入金额均非空且营业收入非零）。",
+    "按总资产全行业三分位分层：Q33=2,805,969,330元，Q67=10,145,495,897元。大型企业（总资产>Q67）42家，小型企业（总资产<=Q33）42家。",
+    "按省份分组：有政策省份（9个）有效企业34家（大型13家、小型12家），无政策省份有效企业91家（大型29家、小型30家）。",
+    "从company_operation_status.csv提取营业利润金额和营业收入金额，计算各企业营业利润率=营业利润金额/营业收入金额×100%。",
+    "计算有政策省份规模差距=大型企业平均利润率-小型企业平均利润率=11.02%-3.30%=7.72个百分点。",
+    "计算无政策省份规模差距=7.37%-6.68%=0.69个百分点。",
+    "计算差距之差=7.72%-0.69%=7.03个百分点。"
+  ],
+  "steps_num": 9,
+  "milestone": {
+    "碳达峰/节能减排政策条数（条）": 12,
+    "有政策省份数（个）": 9,
+    "非金属矿物制品业有效企业总数（家）": 125,
+    "全行业总资产Q33（元）": 2805969329.91,
+    "全行业总资产Q67（元）": 10145495897.25,
+    "有政策省份大型企业数（家）": 13,
+    "有政策省份小型企业数（家）": 12,
+    "有政策省份大型企业平均营业利润率（%）": 11.02,
+    "有政策省份小型企业平均营业利润率（%）": 3.3,
+    "有政策省份规模差距（百分点）": 7.72,
+    "无政策省份大型企业数（家）": 29,
+    "无政策省份小型企业数（家）": 30,
+    "无政策省份大型企业平均营业利润率（%）": 7.37,
+    "无政策省份小型企业平均营业利润率（%）": 6.68,
+    "无政策省份规模差距（百分点）": 0.69,
+    "两类省份差距之差（百分点）": 7.03
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard008.json b/assets/qa_gold/hypothesis_verification/hard008.json
new file mode 100644
index 0000000000000000000000000000000000000000..23fa193a0f82941c8ffda258193e6ad3bb3daf57
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard008.json
@@ -0,0 +1,36 @@
+{
+  "id": "hard008",
+  "question": "2022年，集成电路产业的地方政策竞争进入白热化阶段，各省在专项激励力度上差异显著。本题统计口径说明如下：①统计对象为营业利润与营业收入数据均完整且营业收入非零的内地企业，港澳台地区企业不纳入；②营业利润率 = 营业利润 ÷ 营业收入 × 100%；③认定为专项集成电路产业促进政策，须是专门针对集成电路或半导体产业的地方政策，且明确包含流片补贴、企业落户奖励、研发设计人才支持、产业规模发展目标等专项措施中的至少一项，仅泛提数字经济或科技创新的通用政策不符合要求。在此基础上，请计算2022年半导体业中出台了上述专项政策的省份与未出台省份的企业平均营业利润率，并给出差值（有政策省份均值减去无政策省份均值，以百分点计）。",
+  "guidelines": "依次回答有政策省份平均营业利润率、无政策省份平均营业利润率、两者差值（有政策-无政策）。数值均保留2位小数，以百分点表示。如[10.55, 13.87, -3.32]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    7.02,
+    16.33,
+    -9.31
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"半导体\"的政策记录，找到44条半导体相关政策；进一步筛选policyClassification为\"地方政策\"的记录，共35条。",
+    "从policy_resource.csv中读取这35条地方政策的全文内容，分析哪些政策是专门针对集成电路或半导体产业的专项促进政策（而非泛制造业政策中附带覆盖半导体）。经深度阅读政策正文，筛选出4条包含明确IC专项扶持措施的政策，policy id为：80、125、196、290。对应政策分别为：广东省横琴粤澳深度合作区促进集成电路产业发展若干措施（id=80，含实缴资本奖励最高500万元、总部项目奖励最高2000万元等落户奖励）、浙江省杭州市促进集成电路产业高质量发展实施意见（id=125，含到2025年产业规模实现800亿元发展目标及集成电路企业研发费用>5%要求）、上海市新时期促进集成电路产业和软件产业高质量发展若干政策（id=196，含研发设计人员奖励最高50万元及企业核心团队分级奖励）、安徽省合肥市加快推进集成电路产业发展若干政策（id=290，含流片补贴最高1000万元、EDA工具补贴最高200万元、IP研发补贴）。",
+    "经政策内容分析确认，出台专项集成电路产业促进政策的省份为4个：广东省（id=80）、浙江省（id=125）、上海市（id=196）、安徽省（id=290）。",
+    "从company_profile.csv筛选industry=\"半导体业\"且province不属于港澳台地区的内地企业，共160家，其中有政策省份（广东省、浙江省、上海市、安徽省）97家，无政策省份63家。",
+    "从company_operation_status.csv获取这160家企业的营业利润金额和营业收入金额，全部数据完整且营业收入非零，保留全部160家为有效企业。计算每家企业营业利润率=营业利润金额/营业收入金额×100%。",
+    "计算有政策省份97家企业的平均营业利润率：（各企业营业利润率之和）÷97=7.02%。其中广东省54家均值3.72%、上海市27家均值18.85%、浙江省13家均值-4.42%、安徽省3家均值9.57%。",
+    "计算无政策省份63家企业的平均营业利润率：（各企业营业利润率之和）÷63=16.33%。两组差值=7.02-16.33=-9.31个百分点，有政策省份企业平均营业利润率低于无政策省份，假说未被支持。"
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "半导体相关政策总数（条）": 44,
+    "地方半导体政策数（条）": 35,
+    "专项IC促进政策数（条）": 4,
+    "有政策省份数（个）": 4,
+    "内地半导体有效企业总数（家）": 160,
+    "有政策省份企业数（家）": 97,
+    "无政策省份企业数（家）": 63,
+    "有政策省份平均营业利润率（%）": 7.02,
+    "无政策省份平均营业利润率（%）": 16.33
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard009.json b/assets/qa_gold/hypothesis_verification/hard009.json
new file mode 100644
index 0000000000000000000000000000000000000000..eafd3b292355e5ecacf06fe2725e2e2c3140eb04
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard009.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard009",
+  "question": "在2022年汽车制造业上市企业中验证政策对企业营业利润有更好的促进作用的假设，计算出台了新能源汽车产业专项促进政策省份（有政策省份）的企业营业利润同比增减幅中位数，未出台此类政策省份（无政策省份）的企业营业利润同比增减幅中位数，以及两者的差值（有政策省份中位数减无政策省份中位数，以百分点计）。要求依次回答：有政策省份企业中位数、无政策省份企业中位数、两组中位数之差（均保留2位小数，单位为百分点）。",
+  "guidelines": "依次回答有政策省份企业中位数、无政策省份企业中位数和差值。数值均保留2位小数，单位为百分点（%）。差值=有政策省份中位数-无政策省份中位数。如[4.16, -3.80, 7.96]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    2.73,
+    -5.52,
+    8.25
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选policyClassification为地方政策且industry字段包含汽车的政策，共找到53条汽车制造业相关地方政策。",
+    "从policy_resource.csv读取这53条政策的全文内容，筛选专门指向新能源汽车推广、换电模式应用、燃料电池汽车示范或智能网联汽车管理的政策（即题干定义的新能源汽车产业专项促进政策），共找到11条，政策id为：58、157、165、206、251、322、484、503、580、584、586。经内容分析涉及7个省份：广东省、上海市、海南省、重庆市、四川省、山东省、江苏省。",
+    "从company_profile.csv筛选industry为汽车制造业的企业，排除港澳台地区企业，得到226家内地汽车制造业有效企业。",
+    "从company_operation_status.csv获取226家企业的营业利润同比增减幅数据，所有企业数据均完整。",
+    "按省份分组：有新能源汽车产业专项促进政策省份（7个）共118家企业；无新能源汽车产业专项促进政策省份共108家企业。",
+    "计算两组企业营业利润同比增减幅的中位数：有政策省份中位数=2.73%，无政策省份中位数=-5.52%。差值=2.73% - (-5.52%) = 8.25个百分点。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "汽车相关地方政策总数（条）": 53,
+    "新能源汽车产业专项促进政策数（条）": 11,
+    "有政策省份数（个）": 7,
+    "内地汽车制造业有效企业总数（家）": 226,
+    "有政策省份企业数（家）": 118,
+    "无政策省份企业数（家）": 108,
+    "有政策省份营业利润同比增减幅中位数（%）": 2.73,
+    "无政策省份营业利润同比增减幅中位数（%）": -5.52,
+    "两组中位数差值（百分点）": 8.25
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard010.json b/assets/qa_gold/hypothesis_verification/hard010.json
new file mode 100644
index 0000000000000000000000000000000000000000..9b2d8bca7b389a8a5b062491976bc3fdf4cc5fcb
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard010.json
@@ -0,0 +1,42 @@
+{
+  "id": "hard010",
+  "question": "在2022年金属冶炼和压延加工业上市企业中验证政策对民因企业有更好的资产收益率假设，计算有专项推动有色金属冶炼产业高质量发展政策省份（有政策省份）的所有制效率差距（国有均值减民营均值）、无政策省份的所有制效率差距，以及两类省份差距之差（有政策省份差距减无政策省份差距，以百分点计）。要求依次回答：有政策省份的所有制效率差距、无政策省份的所有制效率差距、两类省份差距之差（均以百分点表示，保留2位小数）。",
+  "guidelines": "依次回答有政策省份的所有制效率差距、无政策省份的所有制效率差距、两类省份差距之差，均以百分点表示，保留2位小数。如[-7.52, -2.10, -5.42]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    -9.89,
+    -3.28,
+    -6.61
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选涉及行业包含\"金属冶炼和压延加工业\"的政策，共得到41条；其中地方政策32条，涉及省份包括云南省、四川省、河南省、贵州省等17个省份。",
+    "从policy_resource.csv中读取上述32条地方政策全文，逐一分析政策目标和主要措施，区分两类：一是以节能减排、碳达峰为主要目标的通用环保政策；二是专项推动有色金属冶炼（绿色铝、钒钛、功能材料等）产业高质量发展的专项产业政策。",
+    "经过深度分析，识别出4个省份出台专项有色金属冶炼产业高质量发展政策，对应政策id为：云南省（id=281《绿色铝产业发展三年行动》、id=128《新材料产业发展三年行动》）、四川省（id=511《促进钒钛产业高质量发展的实施意见》）、河南省（id=87《加快材料产业优势再造换道领跑行动计划》）、贵州省（id=266《支持铜仁市打造国家级新型功能材料战略性新兴产业集群的若干政策措施》）。",
+    "从company_profile.csv筛选行业为\"金属冶炼和压延加工业\"且省份不属于港澳台地区的企业，共139家；与company_operation_status.csv合并后，筛选营业利润金额和总资产均非空且总资产>0的有效企业，得到139家；剔除外资企业（3家）和集体企业（1家），保留国有企业（61家）和民营企业（75家）共136家。",
+    "按省份分组：有政策省份（云南省、四川省、河南省、贵州省）共18家企业（国有10家、民营8家）；无政策省份共118家（国有51家、民营67家）。",
+    "计算各分组的平均资产收益率（资产收益率=营业利润金额/总资产×100%）：有政策省份国有均值=6.29%，民营均值=16.18%，所有制效率差距=6.29%-16.18%=-9.89%；无政策省份国有均值=2.88%，民营均值=6.17%，所有制效率差距=2.88%-6.17%=-3.28%。",
+    "计算两类省份差距之差=(-9.89%) - (-3.28%) = -6.61个百分点。"
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "金属冶炼地方政策总条数（条）": 32,
+    "专项有色金属产业高质量发展政策数（条）": 6,
+    "涉及有政策省份数（个）": 4,
+    "有政策省份有效企业数（家）": 18,
+    "无政策省份有效企业数（家）": 118,
+    "有政策省份国有企业数（家）": 10,
+    "有政策省份民营企业数（家）": 8,
+    "无政策省份国有企业数（家）": 51,
+    "无政策省份民营企业数（家）": 67,
+    "有政策省份国有企业均值资产收益率（%）": 6.29,
+    "有政策省份民营企业均值资产收益率（%）": 16.18,
+    "无政策省份国有企业均值资产收益率（%）": 2.88,
+    "无政策省份民营企业均值资产收益率（%）": 6.17,
+    "有政策省份所有制效率差距（百分点）": -9.89,
+    "无政策省份所有制效率差距（百分点）": -3.28
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard011.json b/assets/qa_gold/hypothesis_verification/hard011.json
new file mode 100644
index 0000000000000000000000000000000000000000..7dbb05769f413049d73d972bfeae8a76b2832c64
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard011.json
@@ -0,0 +1,40 @@
+{
+  "id": "hard011",
+  "question": "在2022年专用设备制造业上市企业中验证政策对头头部企业影响更大的假设，计算有地方重大技术装备专项促进政策或先进制造业专项法规省份（有政策省份）的CR20%、无政策省份的CR20%，以及两者的差值（有政策省份CR20%减无政策省份CR20%，以百分点计）。要求依次回答：有政策省份CR20%、无政策省份CR20%、差值（均保留2位小数，单位为百分点）。",
+  "guidelines": "依次回答有政策省份CR20%、无政策省份CR20%、差值。数值保留2位小数，以百分点表示。如[82.35, 70.14, 12.21]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    79.88,
+    73.71,
+    6.17
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选涉及专用设备制造业的地方政策，共找到47条。",
+    "从policy_resource.csv中读取这47条政策的全文内容，深入分析政策核心目标与具体措施，筛选以'推动重大技术装备创新示范应用'或'专项促进先进制造业/装备产业高端化（含工程机械等专用设备子行业为核心集群目标）'为主要政策目标的地方专项政策。经逐条分析，共识别出2条符合条件的政策：天津市促进首台（套）重大技术装备示范应用若干措施（id=176）和湖南省先进制造业促进条例（id=189）。两条政策涉及天津市和湖南省共2个省份。",
+    "从company_profile.csv筛选行业='专用设备制造业'的企业，共447家；关联company_operation_status.csv获取营业收入金额，排除港澳台地区企业及营业收入为空或为零的企业，得到有效内地企业440家（天津市10家、湖南省11家、其他内地省份419家）。",
+    "将有效企业分为两组：有政策省份（天津市、湖南省，共21家）和无政策省份（其他22个内地省份，共419家）。",
+    "计算有政策省份CR20%：21家企业，按营业收入降序排列，前20%企业数=ceil(21×0.2)=5家；前5家营业收入合计=1326.87亿元，21家总营业收入=1661.09亿元；CR20%=1326.87/1661.09×100%=79.88%。",
+    "计算无政策省份CR20%：419家企业，按营业收入降序排列，前20%企业数=ceil(419×0.2)=84家；前84家营业收入合计=9836.00亿元，419家总营业收入=13344.68亿元；CR20%=9836.00/13344.68×100%=73.71%。",
+    "差值（有政策-无政策）=79.88%-73.71%=6.17个百分点。"
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "专用设备制造业地方政策总数（条）": 47,
+    "符合条件的专项政策数（条）": 2,
+    "政策涉及省份数（个）": 2,
+    "有政策省份有效企业数（家）": 21,
+    "无政策省份有效企业数（家）": 419,
+    "有政策省份前20%企业数（家）": 5,
+    "无政策省份前20%企业数（家）": 84,
+    "有政策省份前20%营业收入合计（亿元）": 1326.87,
+    "有政策省份营业收入总计（亿元）": 1661.09,
+    "无政策省份前20%营业收入合计（亿元）": 9836.0,
+    "无政策省份营业收入总计（亿元）": 13344.68,
+    "有政策省份CR20%（%）": 79.88,
+    "无政策省份CR20%（%）": 73.71
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard012.json b/assets/qa_gold/hypothesis_verification/hard012.json
new file mode 100644
index 0000000000000000000000000000000000000000..188c5ea0cbcebdeefcb56def43a6eca55427c68c
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard012.json
@@ -0,0 +1,39 @@
+{
+  "id": "hard012",
+  "question": "一个行业内营业利润率的分布离散程度，在一定条件下可反映企业间的竞争分化态势或政策的结构性影响效应。以四分位距（IQR = Q3减去Q1，均基于企业个体营业利润率的分布计算）作为离散程度的测量工具，对2022年橡胶和塑料制品业（排除港澳台）数据进行分析，营业利润率=营业利润金额/营业收入金额×100%。若以出台了面向制造业企业、按发展阶段设定梯度化现金奖励或培育支持机制且将橡胶和塑料制品业列为受益行业的专项产业培育激励政策的省份为一组，其余省份为另一组，两组企业营业利润率的IQR分别是多少个百分点？两组IQR的差值（有政策省份减无政策省份）为多少个百分点？",
+  "guidelines": "依次回答有政策省份的IQR、无政策省份的IQR、两者差值。数值以百分点表示，保留2位小数。如[5.82, 9.45, -3.63]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    4.37,
+    11.59,
+    -7.23
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"橡胶\"的地方政策，共找到17条橡胶和塑料制品业相关地方政策，涉及湖南省、山东省、上海市、四川省、云南省、辽宁省、安徽省、广西壮族自治区、海南省、陕西省、河北省、新疆维吾尔自治区等12个省份。",
+    "从policy_resource.csv中读取上述17条地方政策的全文内容，对每条政策进行深度分析，识别其政策类型：（1）安徽省（id=175，安徽省专精特新中小企业倍增行动方案）：对省专精特新冠军企业给予一次性奖补80万元，对国家级专精特新小巨人和单项冠军企业分别奖补100万元、200万元，并设置从创新型中小企业到单项冠军的五级梯度培育通道，明确将橡胶和塑料制品业列为覆盖行业；（2）河北省（id=409，推进规模以上工业企业培育工作若干措施）：建立企业培育库、梯次升级培育机制，涵盖橡胶和塑料制品业，通过资金倾斜、要素保障推动临规企业升规壮大；（3）海南省（id=303，激励企业上规模奖励资金实施细则）：对年产值首次突破3亿至50亿的橡胶和塑料制品业企业分别给予30万至500万元一次性奖励，但海南省无上市橡胶和塑料制品业企业。其余14条政策均为节能减排综合工作实施方案、先进制造业条例、智能制造标杆示范方案、新材料目录、落后产能退出公告、农业现代化规划等非企业梯度激励类政策，或仅将橡胶和塑料制品业列为附带受益行业而无专项企业培育奖励措施。",
+    "确定政策分组：出台了面向制造业企业梯度化现金奖励或培育激励政策（且覆盖橡胶和塑料制品业）的省份为安徽省和河北省（对应政策id：175、409）；海南省虽有同类型政策（id=303）但无上市橡胶和塑料制品业企业。其余13个有橡胶和塑料制品业企业的内地省份均无此类政策。",
+    "从company_profile.csv筛选industry字段为\"橡胶和塑料制品业\"且province不属于港澳台的内地企业，共得到105家有效企业。其中安徽省8家、河北省3家（有政策组共11家），其余省份合计94家（无政策组）。",
+    "从company_operation_status.csv获取105家企业的营业利润金额和营业收入金额，均为非空值且营业收入均不为零，全部105家企业满足有效条件。计算各企业营业利润率 = 营业利润金额 / 营业收入金额 × 100%。",
+    "计算有政策省份（安徽省+河北省，n=11）的营业利润率IQR：Q1 = 7.9124%，Q3 = 12.2784%，IQR = 12.2784 − 7.9124 = 4.3660，保留2位小数为4.37个百分点。",
+    "计算无政策省份（n=94）的营业利润率IQR：Q1 = 1.6678%，Q3 = 13.2604%，IQR = 13.2604 − 1.6678 = 11.5926，保留2位小数为11.59个百分点。",
+    "计算差值：精确值 = 4.3660 − 11.5926 = −7.2266，保留2位小数为−7.23个百分点。"
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "橡胶和塑料制品业相关地方政策总数（条）": 17,
+    "具有企业梯度激励机制的地方政策数（条）": 2,
+    "有政策省份数（个）": 2,
+    "有政策组企业数（家）": 11,
+    "无政策组企业数（家）": 94,
+    "有政策组Q1（%）": 7.91,
+    "有政策组Q3（%）": 12.28,
+    "无政策组Q1（%）": 1.67,
+    "无政策组Q3（%）": 13.26,
+    "有政策省份IQR（百分点）": 4.37,
+    "无政策省份IQR（百分点）": 11.59
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard013.json b/assets/qa_gold/hypothesis_verification/hard013.json
new file mode 100644
index 0000000000000000000000000000000000000000..d87ff2e15b756f14e3b49637bb194b7ecff054a1
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard013.json
@@ -0,0 +1,43 @@
+{
+  "id": "hard013",
+  "question": "国家与地方政策的协同效应，在产业经济学中通常以政策叠加框架加以讨论——双重政策覆盖的企业是否在劳动效率上具有系统性优势，是检验政策层级互补性的核心命题之一。人均营业收入作为劳动效率的代理变量，以2022年食品饮料业为例，将企业所在省份按政策覆盖状态分为三组：第一组，省份同时被国家消费品工业促进政策明确列为活动实施省份且已出台地方食品相关产业政策；第二组，省份仅有地方食品相关产业政策、未被前述国家政策明确覆盖；第三组，上述两类政策均无。有效企业须营业收入金额非空且雇员总数为正的内地企业（排除港澳台）。三组企业的平均人均营业收入分别为多少万元？第一组与第二组的均值差为多少万元？",
+  "guidelines": "依次回答双重覆盖组、仅地方覆盖组、无政策覆盖组的平均人均营业收入（万元/人），以及双重覆盖组与仅地方覆盖组的差值（万元/人）。所有数值保留2位小数。如[185.30, 152.75, 126.40, 32.55]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    201.94,
+    168.2,
+    140.96,
+    33.74
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"食品\"的政策，共找到16条食品相关政策。其中国家级（部委政策）3条、地方政策13条。",
+    "从policy_resource.csv中读取3条国家级食品相关政策全文，政策id为：140、150、156。其中国家级政策包括：《关于推动轻工业高质量发展的指导意见》（id=140，覆盖食品等8个轻工业子行业）、《关于开展2022“三品”全国行活动的通知》（id=150，明确将上海、江苏、浙江、福建、山东、湖北、广东、重庆、四川等9个省份列为活动实施地区）、《数字化助力消费品工业“三品”行动方案》（id=156，全国性数字化转型指导方案）。其中id=150通过点名9省形成差异化国家级覆盖，其余2条为全国普适性指导文件。",
+    "从policy_resource.csv中读取13条地方食品相关政策全文，地方政策id为：43、111、131、134、253、276、303、317、375、377、399、409、512。经分析，有地方食品相关政策的省份共9个：甘肃省（id=43）、河南省（id=111、317）、云南省（id=131、377）、四川省（id=253、375）、湖南省（id=276）、海南省（id=303）、宁夏回族自治区（id=399）、河北省（id=409）、贵州省（id=512）。另有1条吉林省政策（id=134）在policy_release_status.csv中province字段标记为全国，不纳入地方政策省份统计。",
+    "将省份按政策覆盖层级分为三组：双重覆盖组（省份同时在国家三品全国行9省名单内且有地方食品相关政策）= 四川省；仅地方覆盖组（有地方食品相关政策但不在国家三品全国行9省名单内）= 甘肃省、河南省、云南省、湖南省、海南省、宁夏回族自治区、河北省、贵州省（8个省份）；无政策覆盖组（不在国家三品全国行9省名单内且无地方食品相关政策）= 天津市、安徽省、吉林省等13个省份。",
+    "从company_profile.csv筛选industry为食品饮料业且排除港澳台地区的内地企业，共234家。将各企业按所属省份分配到三组：双重覆盖组9家（四川省）、仅地方覆盖组37家（8省合计）、无政策覆盖组59家（13省合计）。",
+    "从company_operation_status.csv获取234家企业的营业收入金额和雇员总数，筛选有效企业（营业收入非空且雇员总数>0）：双重覆盖组有效9家、仅地方覆盖组有效37家、无政策覆盖组有效58家（1家因雇员总数缺失被排除）。",
+    "按组计算各企业人均营业收入（= 营业收入金额 / 雇员总数），再取组内均值：双重覆盖组均值 = 2019384.10元 / 10000 = 201.94万元/人；仅地方覆盖组均值 = 1681974.75元 / 10000 = 168.20万元/人；无政策覆盖组均值 = 1409640.38元 / 10000 = 140.96万元/人；双重覆盖组与仅地方覆盖组差值 = 201.94 - 168.20 = 33.74万元/人。"
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "食品相关政策总数（条）": 16,
+    "国家级食品政策数（条）": 3,
+    "地方食品政策数（条，省份非全国）": 12,
+    "三品全国行明确列名省份数（个）": 9,
+    "有地方食品政策的省份数（个）": 9,
+    "双重覆盖省份数（个）": 1,
+    "仅地方覆盖省份数（个）": 8,
+    "无政策覆盖省份数（个）": 13,
+    "双重覆盖有效企业数（家）": 9,
+    "仅地方覆盖有效企业数（家）": 37,
+    "无政策覆盖有效企业数（家）": 58,
+    "双重覆盖组平均人均营业收入（万元/人）": 201.94,
+    "仅地方覆盖组平均人均营业收入（万元/人）": 168.2,
+    "无政策覆盖组平均人均营业收入（万元/人）": 140.96,
+    "双重覆盖组与仅地方覆盖组差值（万元/人）": 33.74
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard014.json b/assets/qa_gold/hypothesis_verification/hard014.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf660d745a660df5ed3403675f5f9ea8f1e4e3f8
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard014.json
@@ -0,0 +1,34 @@
+{
+  "id": "hard014",
+  "question": "针对验证政府补贴能高效地转化为了企业的创新产出对于不同企业存在差异的假设。以通信传输设备业为例，这一转化效率在有无专项研发创新激励政策的省份之间是否存在差异。认定的专项扶持企业研发创新的地方促进政策须同时满足：①政策涵盖通信传输设备业；②政策明确包含金额确定的企业研发奖励、科技创新补贴或专项创新资金等直接企业激励措施。有效企业指政府补贴金额大于0且年度中国发明专利申请数有完整记录的内地企业。请以每百万元政府补贴所对应的年度发明专利申请数作为衡量补贴转化效率的指标，分别计算并给出2022年有政策省份与无政策省份中有效企业的该指标均值，并给出差值（有政策省份均值减去无政策省份均值）。",
+  "guidelines": "依次回答有政策省份的补贴转化效率、无政策省份的补贴转化效率、两者差值。均保留2位小数，单位为件/百万元。如[3.05, 2.47, 0.58]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    2.38,
+    3.18,
+    -0.81
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv中筛选涉及行业包含「通信传输设备」且政策分类为「地方政策」的记录，共得到46条地方政策，涉及18个省份（含少数省份归属「全国」的政策）。",
+    "从policy_resource.csv中读取上述46条地方政策全文，逐条分析政策是否满足以下两个条件：（1）政策明确将通信传输设备业列为主要支持行业之一；（2）政策正文中包含直接面向企业的、金额明确的创新研发奖励或补贴措施（如专精特新企业研发奖励、科技型企业研发投入补助、创新平台建设补贴等）。经深度内容分析，筛选出5个省份的政策符合上述标准，对应政策id为：湖北省北斗产业高质量发展政策（id=69，对专精特新企业给予50万-100万元一次性奖励，对研发平台给予最高1000万元补助）、陕西省科技型企业创新发展倍增计划（id=238，对研发成果给予最高40%/200万元奖励，对专精特新企业研发新增投入补贴最高500万元）、广东省促进工业经济平稳增长政策（id=15，对通信传输设备等制造业企业实施研发投入奖补）、安徽省专精特新中小企业倍增行动方案（id=175，对专精特新冠军/小巨人企业给予80万-200万元创新奖补，含5G等新型信息技术应用奖补）、上海市推进高端制造业发展若干措施（id=386，对5G/工业互联网应用场景项目给予最高800万元奖励，对制造业创新平台给予最高2000万元支持）。",
+    "从company_profile.csv中筛选行业为「通信传输设备业」且省份不在港澳台的内陆企业，共得到118家，分属17个省份/直辖市。按政策分组：湖北、陕西、广东、安徽、上海5省共62家（有政策组），其余12省共56家（无政策组）。",
+    "从company_operation_status.csv中关联企业运营数据，筛选有效企业（政府奖励资金、补贴>0且年度中国发明专利申请数非空）：有政策组56家，无政策组43家，合计99家有效企业。",
+    "计算各企业的补贴转化效率 = 年度中国发明专利申请数 / (政府奖励资金、补贴 / 1,000,000)，单位为件/百万元。有政策组：56家企业效率值之和 / 56 = 2.38 件/百万元；无政策组：43家企业效率值之和 / 43 = 3.18 件/百万元。",
+    "计算差值：有政策省份均值 - 无政策省份均值 = 2.38 - 3.18 = -0.81 件/百万元，即有政策省份的补贴转化效率反而低于无政策省份，呈现出反常的「政策补贴挤出创新效率」现象。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "通信传输设备业地方政策总条数（条）": 46,
+    "符合专项研发创新扶持条件的政策省份数（个）": 5,
+    "通信传输设备业内陆企业总数（家）": 118,
+    "有政策省份有效企业数（家）": 56,
+    "无政策省份有效企业数（家）": 43,
+    "有政策省份补贴转化效率均值（件/百万元）": 2.38,
+    "无政策省份补贴转化效率均值（件/百万元）": 3.18,
+    "差值（件/百万元）": -0.81
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/hard015.json b/assets/qa_gold/hypothesis_verification/hard015.json
new file mode 100644
index 0000000000000000000000000000000000000000..7561b07d6077826539bce14cab6b0102c257e09c
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/hard015.json
@@ -0,0 +1,38 @@
+{
+  "id": "hard015",
+  "question": "政策文本的表述精度，即政策目标是以可量化的具体数值呈现，还是以方向性、原则性的定性语言为主，可能反映政府的政策执行意志，进而影响辖区内企业的研发行为。在仪器仪表制造业中，对已出台涉及本行业地方政策的省份进行内容分析：凡政策正文中含有具体产业发展数值目标（如产业规模达X亿元、增长X%、新建X家/X座等可核查数值）的省份，归为量化目标组；政策正文仅涵盖定性方向、原则性要求或门槛条件而不包含上述产业发展数值目标的省份，归为定性目标组。基于2022年数据，请分别计算两组省份内仪器仪表制造业有效企业（研发投入占比数据非空且数值在0%到100%之间）的研发投入占比均值，并给出量化目标组均值减去定性目标组均值的差值（单位：百分点）。",
+  "guidelines": "依次回答含量化目标省份的研发投入占比均值、仅含定性目标省份的研发投入占比均值、两者差值（量化组均值减去定性组均值）。均值和差值均保留2位小数，以百分点表示。如[12.53, 8.21, 4.32]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    14.07,
+    9.67,
+    4.4
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选涉及行业包含'仪器仪表制造业'且政策类型为地方政策的记录，共得到39条，涉及湖南省、山东省、四川省、上海市、福建省、江西省、重庆市、广东省、辽宁省、安徽省、天津市、黑龙江省、陕西省、广西壮族自治区、海南省、宁夏回族自治区、甘肃省、河北省、湖北省、新疆维吾尔自治区等省份。",
+    "筛选出与仪器仪表制造业企业分布重叠（即既有政策又有企业）的省份，共10个：湖南省（3家企业）、山东省（1家）、四川省（3家）、上海市（7家）、福建省（3家）、江西省（2家）、广东省（12家）、安徽省（4家）、河北省（2家）、湖北省（2家）。",
+    "从policy_resource.csv中读取上述10个省份共27条仪器仪表相关地方政策全文，对政策正文内容进行深度语义分析，区分量化目标与定性目标。27条政策对应id为：7、12、42、75、78、89、116、153、154、175、189、253、274、276、329、370、372、375、385、386、409、448、523、562、563、590、594。",
+    "经过对政策全文的逐条深度分析，含量化目标的省份共7个：上海市（id=75、385、386、448、562）、广东省（id=153、154、329、563、590）、安徽省（id=175、594）、湖北省（id=523）、江西省（id=89）、四川省（id=42、116、253、375）、湖南省（id=7、189、276、372）。",
+    "仅含定性目标的省份共3个：山东省（id=12、274、370）、福建省（id=78）、河北省（id=409）。",
+    "从company_profile.csv筛选industry='仪器仪表制造业'的企业，按省份分组，从company_operation_status.csv获取研发投入占比数据，筛选有效企业（研发投入占比非空且0%<占比<=100%）。含量化目标组7省共33家有效企业，研发投入占比之和=464.40，均值=464.40/33=14.07%；仅含定性目标组3省共6家有效企业，研发投入占比之和=58.01，均值=58.01/6=9.67%。",
+    "差值=量化目标组均值-定性目标组均值=14.07-9.67=4.40个百分点。"
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "仪器仪表相关地方政策总数（条）": 39,
+    "有企业的目标省份数（个）": 10,
+    "含量化目标省份数（个）": 7,
+    "仅含定性目标省份数（个）": 3,
+    "量化目标组有效企业数（家）": 33,
+    "定性目标组有效企业数（家）": 6,
+    "量化目标组研发投入占比之和（%）": 464.4,
+    "定性目标组研发投入占比之和（%）": 58.01,
+    "量化目标组研发投入占比均值（%）": 14.07,
+    "定性目标组研发投入占比均值（%）": 9.67,
+    "差值（量化组-定性组，百分点）": 4.4
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/medium001.json b/assets/qa_gold/hypothesis_verification/medium001.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f78a530302cb0183aca9d1ed81ea89cd9bacaa1
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/medium001.json
@@ -0,0 +1,44 @@
+{
+  "id": "medium001",
+  "question": "Using 2022 data, to verify the hypothesis that policy has a positive effect on corporate R&D: among listed semiconductor firms, divide them by whether their registered province has ever issued a local industrial policy whose name or covered-industry field contains \"semiconductor\" or \"integrated circuit\" (either condition suffices). What is the difference in mean R&D investment ratio between the policy-covered group and the non-covered group (difference = policy-province group mean minus non-policy-province group mean), in percentage points?",
+  "guidelines": "Answer format: a numeric value (two decimal places, in percentage points). A positive value means the policy-province group is higher; a negative value means the non-policy-province group is higher. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -0.73,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "From policy_release_status.csv, filter policies with policyClassification=\"local policy\" and (policy name contains \"semiconductor\" or \"integrated circuit\", or industry field contains \"semiconductor\"), yielding 35 relevant policies. Extract unique province values (excluding \"nationwide\"), giving 15 policy provinces: Shanghai, Yunnan, Sichuan, Anhui, Shandong, Guangdong, Xinjiang Uygur Autonomous Region, Jiangxi, Henan, Zhejiang, Hainan, Hunan, Fujian, Chongqing, Shaanxi.",
+    "From company_profile.csv, filter firms whose industry is semiconductor, extract company name, bmCode, and province, yielding 172 semiconductor firms.",
+    "From company_operation_status.csv, join 2022 data by bmCode for these firms, extract R&D investment ratio; 172 firms successfully matched.",
+    "Exclude 3 firms with null R&D investment ratio and 0 firms with R&D ratio > 100% (outliers); 169 valid firms remain.",
+    "Split firms by whether their province is in the policy-province list: policy group (110 firms) and non-policy group (59 firms).",
+    "Policy group mean R&D investment ratio = Σ(firm R&D ratio) / firm count = 10.8880%; non-policy group mean = Σ(firm R&D ratio) / firm count = 11.6136%.",
+    "Difference = policy group mean − non-policy group mean = 10.8880 − 11.6136 = −0.73 percentage points."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Policy province list": [
+      "Shanghai",
+      "Yunnan",
+      "Sichuan",
+      "Anhui",
+      "Shandong",
+      "Guangdong",
+      "Xinjiang Uygur Autonomous Region",
+      "Jiangxi",
+      "Henan",
+      "Zhejiang",
+      "Hainan",
+      "Hunan",
+      "Fujian",
+      "Chongqing",
+      "Shaanxi"
+    ],
+    "Policy group firm count": 110,
+    "Non-policy group firm count": 59,
+    "Policy group mean R&D investment ratio (%)": 10.888,
+    "Non-policy group mean R&D investment ratio (%)": 11.6136
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/medium002.json b/assets/qa_gold/hypothesis_verification/medium002.json
new file mode 100644
index 0000000000000000000000000000000000000000..b9d24e9dd8d157e53ec880d0e09fa15e52acfa1c
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/medium002.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium002",
+  "question": "In 2022, in the pharmaceutical manufacturing industry, what is the difference in Pearson correlation coefficient between private enterprises and state-owned enterprises(State-owned enterprises include central SOEs, local SOEs, SOEs under research institutes, and other SOEs.) with respect to R&D investment amount and cumulative number of granted Chinese invention patents?",
+  "guidelines": "Answer format: the difference value (four decimal places). Difference = private enterprise correlation − state-owned enterprise correlation. A positive value means private enterprises have stronger correlation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": -0.1452,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "From company_profile.csv, filter firms whose industry is pharmaceutical manufacturing, extract company name, bmCode, and ownership; 449 pharmaceutical manufacturing firms found.",
+    "Split firms by ownership into state-owned group (central SOEs, local SOEs, SOEs under research institutes, other SOEs), 67 firms, and private enterprise group, 346 firms.",
+    "From company_operation_status.csv, join 2022 data by bmCode for these firms, extract R&D investment amount and cumulative number of granted Chinese invention patents. State-owned group: 67 firms matched; private group: 346 firms matched.",
+    "State-owned group: exclude firms with null R&D amount or null cumulative invention patents; 67 valid firms remain. Pearson correlation between R&D amount and cumulative invention patents: r = 0.7170.",
+    "Private group: exclude firms with null R&D amount or null cumulative invention patents; 309 valid firms remain. Pearson correlation between R&D amount and cumulative invention patents: r = 0.5719.",
+    "Difference = private group correlation − state-owned group correlation = 0.5719 − 0.7170 = −0.1452.",
+    "Output the difference: −0.1452."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "State-owned valid count": 67,
+    "Private valid count": 309,
+    "State-owned Pearson correlation": 0.717,
+    "Private Pearson correlation": 0.5719,
+    "Difference": -0.1452
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/medium003.json b/assets/qa_gold/hypothesis_verification/medium003.json
new file mode 100644
index 0000000000000000000000000000000000000000..6415b6eceb443820e163e266e4b929caa7d67a64
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/medium003.json
@@ -0,0 +1,49 @@
+{
+  "id": "medium003",
+  "question": "To verify the hypothesis that policy density has a positive effect on average corporate profitability, using 2022 automotive manufacturing as an example, calculate the Spearman rank correlation coefficient between each province's policy density indicator (number of policies whose industry field contains \"automotive\") and average profitability (total operating profit amount / total operating revenue amount × 100%). Provinces without automotive manufacturing operating data or with zero operating revenue are excluded from the calculation; valid provinces with no policy records are assigned a policy count of 0.",
+  "guidelines": "Answer format: numeric value (rounded to 4 decimal places). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": -0.0023,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "Filter from policy_release_status.csv all policy records where the industry field contains \"automotive\", finding 69 related policies. After excluding national policies where province is null or \"National\", 53 remain. Group by province and count policy quantity per province, with 20 provinces having related policies.",
+    "Filter from regional_industry_status.csv all provincial records with industry=\"Automotive Manufacturing\", extract province, total operating profit amount, and total operating revenue amount fields, finding data for 34 provinces.",
+    "Filter out provinces where total operating revenue amount is null or zero, leaving 14 valid provinces. Calculate each province's average operating profit margin = total operating profit amount / total operating revenue amount × 100%, with profitability range from -4.4704% to 8.1755%.",
+    "Join policy count data with provincial profitability data by province; provinces without corresponding policies are assigned a policy count of 0. After joining, total valid provinces is 14, of which 11 have policies.",
+    "Calculate the Spearman rank correlation coefficient between policy count and average operating profit margin: rho = -0.0023, p-value = 0.993880.",
+    "Output the Spearman rank correlation coefficient value as -0.0023."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Total automotive-related policies": 69,
+    "Provinces with policies": 20,
+    "Total valid provinces": 14,
+    "Sample province data": {
+      "Guangdong Province": {
+        "Policy count": 10,
+        "Average operating profit margin (%)": 2.9496
+      },
+      "Shanghai": {
+        "Policy count": 7,
+        "Average operating profit margin (%)": 0.582
+      },
+      "Hunan Province": {
+        "Policy count": 5,
+        "Average operating profit margin (%)": 7.6412
+      },
+      "Sichuan Province": {
+        "Policy count": 4,
+        "Average operating profit margin (%)": 8.1755
+      },
+      "Shandong Province": {
+        "Policy count": 3,
+        "Average operating profit margin (%)": 3.402
+      }
+    },
+    "Spearman correlation coefficient": -0.0023,
+    "p-value": 0.99388
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/medium004.json b/assets/qa_gold/hypothesis_verification/medium004.json
new file mode 100644
index 0000000000000000000000000000000000000000..5633b121d19f397cc9706799a66942805e572a16
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/medium004.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium004",
+  "question": "In 2022, to verify the hypothesis on the effect of debt ratio on R&D investment ratio in Raw Chemical Materials and Chemical Products Manufacturing, calculate the difference in the mean R&D investment ratio (in percentage points) between the high-debt group (asset-liability ratio above the national industry median) and the low-debt group (asset-liability ratio below the national industry median).",
+  "guidelines": "Answer format: numeric value (rounded to 2 decimal places, unit: percentage points). A positive value indicates the high-debt group is higher. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": -0.61,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "Filter from national_industry_status.csv for industry=\"Raw Chemical Materials and Chemical Products Manufacturing\", extract the asset-liability ratio median field. The national median asset-liability ratio for this industry is 36.815.",
+    "Filter from company_profile.csv all enterprises with industry=\"Raw Chemical Materials and Chemical Products Manufacturing\", extract enterprise name and bmCode fields, finding 364 enterprises.",
+    "Join with company_operation_status.csv by bmCode to obtain 2022 data for these enterprises, extract asset-liability ratio and R&D investment ratio fields, obtaining data for 364 enterprises.",
+    "Filter out enterprises where asset-liability ratio or R&D investment ratio is null; exclude 0 anomalous enterprises with R&D investment ratio above 100%, leaving 349 valid enterprises.",
+    "Group by comparing asset-liability ratio with the national median of 36.815: asset-liability ratio above median is the high-debt group (175 enterprises); below median is the low-debt group (174 enterprises); 0 enterprises equal to median are excluded from grouping.",
+    "Calculate high-debt group average R&D investment ratio = Σ(enterprise R&D investment ratio) / number of enterprises = 3.6595%; low-debt group average R&D investment ratio = Σ(enterprise R&D investment ratio) / number of enterprises = 4.2648%.",
+    "Calculate difference = high-debt group average R&D investment ratio - low-debt group average R&D investment ratio = 3.6595 - 4.2648 = -0.61 percentage points."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "National asset-liability ratio median": 36.815,
+    "High-debt group enterprises": 175,
+    "Low-debt group enterprises": 174,
+    "High-debt group average R&D investment ratio (%)": 3.6595,
+    "Low-debt group average R&D investment ratio (%)": 4.2648
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/medium005.json b/assets/qa_gold/hypothesis_verification/medium005.json
new file mode 100644
index 0000000000000000000000000000000000000000..42542b3677d3eab058a12cfe336f4aa7f41980c2
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/medium005.json
@@ -0,0 +1,33 @@
+{
+  "id": "medium005",
+  "question": "In 2022, to verify the hypothesis that there is a clear relationship between enterprise total assets and invention patent count in the Consumer Electronics and Electrical Industry, we focus only on provinces with high R&D density (provinces where the mean R&D investment ratio in provincial industry aggregate data exceeds the corresponding national industry aggregate mean). Among all listed enterprises in this industry within these provinces, what is the Pearson correlation coefficient between enterprise total asset scale and annual China invention patent applications?",
+  "guidelines": "Answer format: numeric value (rounded to 4 decimal places). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 0.8294,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "Filter from national_industry_status.csv for industry=\"Consumer Electronics and Electrical Industry\", extract the mean R&D investment ratio field. The national mean R&D investment ratio for Consumer Electronics and Electrical Industry is 7.77502994011976.",
+    "Filter from regional_industry_status.csv for industry=\"Consumer Electronics and Electrical Industry\", extract province and mean R&D investment ratio fields. 16 provinces have data.",
+    "Filter provinces where mean R&D investment ratio exceeds the national benchmark of 7.77502994011976, obtaining 4 R&D-intensive provinces: Guangdong Province, Beijing, Shanghai, Henan Province.",
+    "Filter from company_profile.csv enterprises with industry=\"Consumer Electronics and Electrical Industry\" and province in the R&D-intensive province list, finding 180 enterprises.",
+    "Join with company_operation_status.csv for 2022 data of these enterprises, extract total assets and annual China invention patent applications fields, obtaining data for 180 enterprises.",
+    "Filter out records where total assets or annual China invention patent applications is null, leaving 153 valid enterprises. Mean total assets is 24143218637.14 CNY, mean annual China invention patent applications is 197.82.",
+    "Calculate the Pearson correlation coefficient between total assets and annual China invention patent applications: r = 0.8294."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "National mean R&D investment ratio": 7.77502994011976,
+    "R&D-intensive province list": [
+      "Guangdong Province",
+      "Beijing",
+      "Shanghai",
+      "Henan Province"
+    ],
+    "Valid enterprises": 153,
+    "Mean total assets (CNY)": 24143218637.14,
+    "Mean annual China invention patent applications": 197.82
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/medium006.json b/assets/qa_gold/hypothesis_verification/medium006.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce04bdf342dee1907c37cbc47fdd96851391d0df
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/medium006.json
@@ -0,0 +1,27 @@
+{
+  "id": "medium006",
+  "question": "In 2022, there was a view that government subsidies and revenue are clearly correlated among large-asset-scale enterprises in the food and beverage industry. Therefore, researchers sampled the top one-third of large enterprises ranked (rounded down) by total assets as the research subject. What is the Pearson correlation coefficient between government subsidy amount and year-over-year revenue growth rate?",
+  "guidelines": "Answer format: numerical value (retain 4 decimal places). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": -0.1237,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "Filter companies with industry=\"food and beverage\" from company_profile.csv, extract company name and bmCode fields, and found 247 food and beverage companies.",
+    "From company_operation_status.csv, associate 2022 data for these companies based on bmCode, extract total assets, government incentive funds, subsidies, and year-over-year revenue growth rate fields, and associate data for 247 companies.",
+    "Filter out companies with empty total assets, leaving 247 companies, sorted in descending order by total assets.",
+    "Calculate total number of companies 247, take the top 82 companies (rounded down) as the large enterprise group.",
+    "In the large enterprise group, filter out companies with empty government incentive funds, subsidies, or year-over-year revenue growth rate, leaving 80 valid enterprises.",
+    "Calculate the Pearson correlation coefficient between government incentive funds and subsidies and year-over-year revenue growth rate = -0.1237."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Total number of food and beverage enterprises": 247,
+    "Number of large enterprise group": 82,
+    "Number of valid large enterprises": 80,
+    "Average government subsidy": 137397163.62,
+    "Average revenue growth rate": 14.8526
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/medium007.json b/assets/qa_gold/hypothesis_verification/medium007.json
new file mode 100644
index 0000000000000000000000000000000000000000..14deef298656604ab8ece701f6898ba4b0b512ca
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/medium007.json
@@ -0,0 +1,51 @@
+{
+  "id": "medium007",
+  "question": "In the 2022 data, to verify the hypothesis that private enterprises in the communications transmission equipment industry benefit more from policies than state-owned enterprises, we focus on communications transmission equipment enterprises in provinces that have issued local policies involving the \"communications\"-related industry. Enterprises are grouped by ownership type (private enterprises vs. state-owned enterprises, where state-owned enterprises include only central state-owned enterprises and local state-owned enterprises). Within each group, the per capita revenue (total revenue / total number of employees) is calculated using the weighted consolidation method. Finally, return the specific difference between per capita revenue of the private enterprise group and that of the state-owned enterprise group.",
+  "guidelines": "Answer format: numerical value (retain 2 decimal places, unit: yuan/person). A positive value indicates private enterprises are higher. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 245418.07,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "Filter policy records from policy_release_status.csv where policyClassification=\"local policy\" and the industry field contains \"communications\", finding 48 relevant policies. Extract the deduplicated province field (excluding \"national\"), obtaining 18 provinces with policies: Shanghai, Yunnan, Beijing, Sichuan, Anhui, Shandong, Guangdong, Xinjiang Uygur Autonomous Region, Jiangxi, Henan, Hainan, Hubei, Hunan, Fujian, Guizhou, Liaoning, Chongqing, Shaanxi.",
+    "Filter from company_profile.csv companies whose industry contains \"communications transmission equipment\" and province is among the policy provinces. Extract company name, bmCode, and ownership fields, finding 88 companies.",
+    "Divide enterprises into the private enterprise group (ownership=\"private enterprise\", 61 companies) and the state-owned enterprise group (ownership is \"central state-owned enterprise\" or \"local state-owned enterprise\", 21 companies).",
+    "From company_operation_status.csv, associate 2022 data based on bmCode, extract revenue amount and total employee fields. Private enterprises: 61 associated; state-owned enterprises: 21 associated.",
+    "Filter out records with empty revenue amount or total employees, or zero total employees. Valid private enterprises: 61; valid state-owned enterprises: 20.",
+    "Calculate per capita revenue of private enterprise group = total revenue of private enterprises / total employees of private enterprises = 1009833229861.02 / 471176 = 2143218.73 yuan/person; calculate per capita revenue of state-owned enterprise group = total revenue of state-owned enterprises / total employees of state-owned enterprises = 227578562511.11 / 119917 = 1897800.67 yuan/person.",
+    "Calculate difference = per capita revenue of private enterprises - per capita revenue of state-owned enterprises = 2143218.73 - 1897800.67 = 245418.07 yuan/person."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "List of provinces with policies": [
+      "Shanghai",
+      "Yunnan",
+      "Beijing",
+      "Sichuan",
+      "Anhui",
+      "Shandong",
+      "Guangdong",
+      "Xinjiang Uygur Autonomous Region",
+      "Jiangxi",
+      "Henan",
+      "Hainan",
+      "Hubei",
+      "Hunan",
+      "Fujian",
+      "Guizhou",
+      "Liaoning",
+      "Chongqing",
+      "Shaanxi"
+    ],
+    "Number of private enterprises": 61,
+    "Number of state-owned enterprises": 20,
+    "Total revenue of private enterprises": 1009833229861.02,
+    "Total employees of private enterprises": 471176,
+    "Total revenue of state-owned enterprises": 227578562511.11,
+    "Total employees of state-owned enterprises": 119917,
+    "Per capita revenue of private enterprises": 2143218.73,
+    "Per capita revenue of state-owned enterprises": 1897800.67
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/medium008.json b/assets/qa_gold/hypothesis_verification/medium008.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0ae1f63ecb2badf455bcb388ae8fc898b719f62
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/medium008.json
@@ -0,0 +1,29 @@
+{
+  "id": "medium008",
+  "question": "Using the specialized equipment manufacturing industry in 2022 as the research subject, some researchers believe that enterprise size may confound the relationship between listing history length and cumulative patent accumulation. To test this hypothesis: First, among all valid enterprises (with non-null cumulative China patent applications), calculate the Pearson correlation coefficient r1 between listing years (derived by subtracting listing year from 2022) and cumulative China patent applications; second, restrict the sample to the large enterprise subset whose total assets exceed the industry median total assets in the national industry aggregate data, then calculate the Pearson correlation coefficient r2 for the same pair of variables; finally, report the specific value of the difference r2 − r1.",
+  "guidelines": "Answer format: the difference between the two correlation coefficients (retain 4 decimal places). Difference = large enterprise correlation coefficient − all enterprises correlation coefficient. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 0.0093,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "Filter industry=\"specialized equipment manufacturing\" from national_industry_status.csv, extract the median total assets field. National median total assets for specialized equipment manufacturing: 2619344581.0.",
+    "Filter all enterprises with industry=\"specialized equipment manufacturing\" from company_profile.csv, extract company name, bmCode, and listingDate fields, finding 447 companies.",
+    "Filter out 0 companies with empty listingDate, 447 companies remain. Calculate listing years = 2022 - listing year based on listingDate, listing years range from 0 to 34 years.",
+    "From company_operation_status.csv, associate 2022 data for these companies based on bmCode, extract total assets and cumulative China patent applications fields, successfully associating 447 companies.",
+    "Filter out 20 companies with empty cumulative China patent applications, 427 valid enterprises remain for the full sample. Calculate Pearson correlation coefficient r1 between listing years and cumulative China patent applications = 0.3377.",
+    "Further filter large enterprises with total assets exceeding the national industry median (2619344581.0), 217 companies in total.",
+    "Calculate Pearson correlation coefficient r2 between listing years and cumulative China patent applications for the large enterprise group = 0.3471.",
+    "Calculate difference = r2 - r1 = 0.3471 - 0.3377 = 0.0093."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "National median total assets": 2619344581.0,
+    "Number of all valid enterprises": 427,
+    "r1 value": 0.3377,
+    "Number of large enterprises": 217,
+    "r2 value": 0.3471
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/medium009.json b/assets/qa_gold/hypothesis_verification/medium009.json
new file mode 100644
index 0000000000000000000000000000000000000000..56f31cdde9721e7a5a305df836895e3a1f45d6db
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/medium009.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium009",
+  "question": "In the pharmaceutical manufacturing industry in 2022, verify whether the 'policy-innovation paradox' exists (i.e., the phenomenon where the strength of local pharmaceutical innovation support policies is negatively associated with innovation output). Among provinces that have issued pharmaceutical support policies, use the median number of policy entries as the threshold for support strength, and use the national average invention patent grants per province for pharmaceutical manufacturing as the innovation output benchmark. Count the number of provinces where policy support strength exceeds the median but total invention patent grants are below the national average.",
+  "guidelines": "Answer format: numerical value (integer). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 3,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "Filter from policy_release_status.csv: policyClassification=\"local policy\", publishDate in 2022, industry contains \"pharmaceutical manufacturing\", province is non-empty and not \"national\" (全国). Aggregate policy entry counts by province to obtain 21 provinces that have issued pharmaceutical support policies.",
+    "Take the median of policy entry counts across those 21 provinces, yielding 2.0; define \"support strength above the median\" as a strictly greater policy count than 2.",
+    "From regional_industry_status.csv, filter rows where industry=\"pharmaceutical manufacturing\". Treat empty values in 「年度中国发明专利授权数合计」 (annual total China invention patent grants) as undisclosed: compute the national average invention patent grants per province for pharmaceutical manufacturing using only provincial records with non-empty values, yielding 15 provinces with disclosed data, total 3243, provincial average = 3243 / 15 = 216.20.",
+    "Link policy provinces with patent data: a province participates in the paradox judgment only if it has a disclosed annual total for invention patent grants; provinces without disclosed patent data are excluded from the comparison.",
+    "Filter provinces that simultaneously satisfy: policy entry count > 2.0 and disclosed patent total < 216.20, yielding: Shanghai (policy count = 11, patents = 210), Henan (policy count = 3, patents = 102), Sichuan (policy count = 4, patents = 112).",
+    "The final number of qualifying provinces is 3."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Number of provinces with valid disclosure": 15,
+    "Median policy entry count": 2.0,
+    "National average invention patent grants per province (pharmaceutical manufacturing)": 216.2,
+    "Number of paradox provinces": 3
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/medium010.json b/assets/qa_gold/hypothesis_verification/medium010.json
new file mode 100644
index 0000000000000000000000000000000000000000..b82c88b6cfd34ebf5d57aae4f9ae7bf04f7fb0e9
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/medium010.json
@@ -0,0 +1,30 @@
+{
+  "id": "medium010",
+  "question": "In 2022, based on the 'high input low output' hypothesis in Romer's endogenous growth theory, verify whether state-owned enterprises in the semiconductor industry exhibit the dual anomaly of 'large asset scale but low operating profit margin' and 'high R&D investment but low patent conversion efficiency'. The hypothesis states: when state-owned enterprises rank high in the industry in both asset scale (total assets) and R&D investment (amount), their operating profit margin (operating profit / revenue) and patent conversion efficiency (cumulative China invention patent grants / R&D investment × 100 million) should rank low in the industry. Please count the number of enterprises that simultaneously satisfy the following conditions (state-owned enterprises include central state-owned enterprises, local state-owned enterprises, state-owned enterprises (research institutes), and state-owned enterprises (other)): ① total assets > median total assets of industry-wide state-owned enterprises; ② operating profit margin < median operating profit margin of industry-wide state-owned enterprises; ③ R&D investment amount > median R&D investment amount of industry-wide state-owned enterprises; ④ patent conversion efficiency < median patent conversion efficiency of industry-wide state-owned enterprises.",
+  "guidelines": "Answer format: numerical value (integer, unit: enterprises). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 3,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "Filter from company_profile.csv enterprises with industry=\"semiconductor\" and ownership as state-owned enterprise types (central state-owned enterprise, local state-owned enterprise, state-owned enterprise (research institute), state-owned enterprise (other)), finding 32 state-owned enterprises.",
+    "From company_operation_status.csv, associate 2022 data for these enterprises, extract total assets, operating profit amount, revenue amount, R&D investment amount, and cumulative China invention patent grants fields. Data associated for 32 enterprises.",
+    "Filter out enterprises with empty or zero revenue amount, 32 valid enterprises remain. Calculate operating profit margin = operating profit amount / revenue amount × 100%.",
+    "Filter out enterprises with empty or zero R&D investment amount, 32 valid enterprises remain. Calculate patent conversion efficiency = cumulative China invention patent grants / R&D investment amount × 100 million.",
+    "Calculate median for each indicator: median total assets = 5937801202.30 yuan, median operating profit margin = 11.9667%, median R&D investment amount = 245125759.25 yuan, median patent conversion efficiency = 50.7635 grants per 100 million yuan.",
+    "Filter enterprises satisfying all four conditions: total assets > median (16), operating profit margin < median (16), R&D investment amount > median (16), patent conversion efficiency < median (15). Enterprises satisfying all four conditions: 3.",
+    "Count dual anomaly enterprises: 3."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Total number of state-owned enterprises": 32,
+    "Number of valid enterprises": 32,
+    "Median total assets (yuan)": 5937801202.3,
+    "Median operating profit margin (%)": 11.9667,
+    "Median R&D investment amount (yuan)": 245125759.25,
+    "Median patent conversion efficiency (grants per 100 million yuan)": 50.7635,
+    "Number of enterprises satisfying conditions": 3
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/medium011.json b/assets/qa_gold/hypothesis_verification/medium011.json
new file mode 100644
index 0000000000000000000000000000000000000000..db90ac5fddd3e3da5f48d7ce712b1ccc64f06bb9
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/medium011.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium011",
+  "question": "In the 2022 provincial-level data for the automotive manufacturing industry, some provinces exhibit a dual structural contradiction: first, the number of automotive manufacturing enterprises in the province exceeds the average number of enterprises across all provinces with automotive manufacturing, but the province's total automotive manufacturing revenue is below the average revenue of all provinces; second, the province's average R&D investment ratio is higher than the mean of this indicator across provinces, but the average profit margin (measured as total operating profit / total revenue × 100%) is lower than the mean of this profit margin indicator across all provinces. Please count the number of provinces in 2022 that simultaneously meet both of the above contradiction conditions.",
+  "guidelines": "Answer format: numerical value (integer). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "Filter from regional_industry_status.csv all provincial records with industry=\"automotive manufacturing\" and total number of enterprises > 0. Extract province, total number of enterprises, total revenue amount, total operating profit amount, and average R&D investment ratio fields. Found 14 valid provinces.",
+    "Calculate the mean total number of enterprises across all valid provinces = 14.0000, and mean total revenue amount = 297307507961.18 yuan.",
+    "Calculate each province's average profit margin = total operating profit amount / total revenue amount × 100%, and calculate the mean of average profit margin across all valid provinces = 3.2804%.",
+    "Calculate the mean of average R&D investment ratio across all valid provinces = 6.1612.",
+    "Filter provinces satisfying all four conditions: total number of enterprises > 14.0000, total revenue amount < 297307507961.18, average R&D investment ratio > 6.1612, and average profit margin < 3.2804%. Provinces satisfying conditions: none.",
+    "Count of provinces satisfying conditions: 0."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Total number of valid provinces": 14,
+    "Mean total number of enterprises": 14.0,
+    "Mean revenue (yuan)": 297307507961.18,
+    "Mean profit margin (%)": 3.2804,
+    "Mean R&D ratio": 6.1612,
+    "List of provinces satisfying conditions": "none"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/medium012.json b/assets/qa_gold/hypothesis_verification/medium012.json
new file mode 100644
index 0000000000000000000000000000000000000000..0ff8a0a3d2b3bb5487fd89daaa3c05ec97972bbf
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/medium012.json
@@ -0,0 +1,27 @@
+{
+  "id": "medium012",
+  "question": "In 2022, based on the 'structural efficiency paradox' hypothesis in industrial economics (i.e., during enterprise technology transformation, the anomalous phenomenon may occur where revenue growth coexists with workforce reduction, and reduced R&D investment coexists with increased innovation output), verify whether the dual paradox exists in the chemical fiber manufacturing industry. The hypothesis states: when enterprises are in the technology upgrading stage, revenue grows but automation replaces labor leading to fewer employees; meanwhile, R&D investment ratio is below the industry median but patent grants are above the industry median. Please count the number of valid enterprises that simultaneously exhibit both paradox characteristics.",
+  "guidelines": "Answer format: numerical value (integer, unit: enterprises). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 2,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "Filter all enterprises with industry=\"chemical fiber manufacturing\" from company_profile.csv, extract company name and bmCode fields, finding 34 enterprises.",
+    "From company_operation_status.csv, associate 2022 data for these enterprises based on bmCode, extract four indicators: year-over-year revenue growth rate, year-over-year employee growth rate, R&D investment ratio, and annual China patent grants. Successfully associated 34 enterprises. Filter out 4 enterprises with any of the four indicators empty, obtaining 30 valid enterprises.",
+    "Calculate median R&D investment ratio of valid enterprises = 3.355, median annual China patent grants = 21.0.",
+    "Filter first paradox: enterprises with year-over-year revenue growth rate > 0 and year-over-year employee growth rate < 0, 3 enterprises in total.",
+    "Among first paradox enterprises, filter second paradox: enterprises with R&D investment ratio < industry median (3.355) and annual China patent grants > industry median (21.0), 2 enterprises in total.",
+    "Number of enterprises simultaneously satisfying both paradox conditions: 2."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Total number of valid enterprises": 30,
+    "Median R&D investment ratio": 3.355,
+    "Median annual patent grants": 21.0,
+    "Number of first paradox enterprises": 3,
+    "Number of dual paradox enterprises": 2
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/medium013.json b/assets/qa_gold/hypothesis_verification/medium013.json
new file mode 100644
index 0000000000000000000000000000000000000000..85c352f88b608812b732878fe89b29300250f7c4
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/medium013.json
@@ -0,0 +1,29 @@
+{
+  "id": "medium013",
+  "question": "In 2022, in the information transmission, software and information technology services industry, researchers aim to characterize a type of dual anomaly enterprises with high valuation and high R&D investment, yet weak profitability and insufficient patent influence. The specific criteria are: valid enterprises must have complete data for all four indicators—market cap, revenue (non-zero), R&D investment ratio, and cumulative total patent citations; net profit margin is calculated as net profit amount divided by revenue amount times 100%; among all valid enterprises, using each indicator's median as the threshold, filter enterprises where market cap exceeds the median and net profit margin is below the median, while R&D investment ratio exceeds the median and cumulative patent citations are below the median. What is the proportion of such enterprises among all valid enterprises?",
+  "guidelines": "Answer format: proportional value (retain 2 decimal places, unit: %). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 3.49,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "Filter all enterprises with industry=\"information transmission, software and information technology services\" from company_profile.csv, finding 644 enterprises.",
+    "From company_operation_status.csv, associate 2022 data for these enterprises based on bmCode, extract market cap, net profit amount, revenue amount, R&D investment ratio, and cumulative total patent citations fields. Successfully associated 644 enterprises.",
+    "Filter out 100 enterprises with any of market cap, revenue amount, R&D investment ratio, or cumulative total patent citations empty, or with zero revenue. 544 valid enterprises remain.",
+    "Calculate net profit margin = net profit amount / revenue amount × 100%. Net profit margin range: -1857.0154% to 80.7125%.",
+    "Calculate median for each of the four indicators: median market cap = 59.5, median net profit margin = 3.2905%, median R&D investment ratio = 12.155000000000001, median cumulative total patent citations = 363.0.",
+    "Filter dual anomaly enterprises satisfying all four conditions (market cap > median, net profit margin < median, R&D investment ratio > median, cumulative patent citations < median), 19 enterprises in total.",
+    "Calculate proportion = 19 / 544 × 100% = 3.49%."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Total number of valid enterprises": 544,
+    "Median market cap": 59.5,
+    "Median net profit margin (%)": 3.2905,
+    "Median R&D investment ratio": 12.155,
+    "Median cumulative patent citations": 363.0,
+    "Number of anomaly enterprises": 19
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/hypothesis_verification/medium014.json b/assets/qa_gold/hypothesis_verification/medium014.json
new file mode 100644
index 0000000000000000000000000000000000000000..49ac8bc5a3ce708df61189bfc359cf82dcffd37c
--- /dev/null
+++ b/assets/qa_gold/hypothesis_verification/medium014.json
@@ -0,0 +1,29 @@
+{
+  "id": "medium014",
+  "question": "In 2022, based on the 'leverage-growth coupling effect' hypothesis in the context of industrial upgrading (i.e., there is a non-linear relationship between corporate financial leverage and operating growth, potentially showing a reverse coupling pattern of high leverage with high growth or low leverage with negative growth), in provinces covered by local policies for the rubber and plastic products industry, verify whether listed enterprises in this industry exhibit structural leverage-growth association. Specifically: among valid enterprises, using the median asset-liability ratio as the financial leverage threshold and the sign of revenue growth rate as the growth direction indicator, count the number of coupling enterprises satisfying 'high leverage (asset-liability ratio > median) and high growth (growth rate > 0)' (A) and coupling enterprises satisfying 'low leverage (asset-liability ratio < median) and negative growth' (B), and calculate the difference A − B (integer). What is the specific value of this difference?",
+  "guidelines": "Answer format: numerical value (integer). A positive value indicates that 'high leverage high growth' coupling is more prominent; a negative value indicates that 'low leverage negative growth' coupling is more prominent. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 4,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  },
+  "steps": [
+    "Filter from policy_release_status.csv policy records where policyClassification=\"local policy\" and industry contains \"rubber\" or \"plastic\", finding 17 relevant policies. Extract deduplicated province field (excluding \"national\"), obtaining 12 policy provinces: Shanghai, Yunnan, Sichuan, Anhui, Shandong, Guangxi Zhuang Autonomous Region, Xinjiang Uygur Autonomous Region, Hebei, Hainan, Hunan, Liaoning, Shaanxi.",
+    "Filter from company_profile.csv enterprises with industry=\"rubber and plastic products\" and province in policy provinces. Extract company name, bmCode, and province fields, finding 30 enterprises.",
+    "From company_operation_status.csv, associate 2022 data for these enterprises based on bmCode, extract asset-liability ratio and year-over-year revenue growth rate fields. Successfully associated 30 enterprises.",
+    "Filter out 0 enterprises with empty asset-liability ratio or year-over-year revenue growth rate. 30 valid enterprises remain (year-over-year revenue growth rate > 0 indicates high growth, < 0 indicates negative growth).",
+    "Calculate median asset-liability ratio of valid enterprises = 36.875.",
+    "Count 'high leverage high growth' coupling enterprises: asset-liability ratio > 36.875 and year-over-year revenue growth rate > 0, 13 enterprises in total.",
+    "Count 'low leverage negative growth' coupling enterprises: asset-liability ratio < 36.875 and year-over-year revenue growth rate < 0, 9 enterprises in total (those equal to median are not counted in either group).",
+    "Calculate coupling effect strength difference = 13 - 9 = 4."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "Number of policy provinces": 12,
+    "Total number of valid enterprises": 30,
+    "Median asset-liability ratio": 36.875,
+    "Number of high leverage high growth coupling enterprises": 13,
+    "Number of low leverage negative growth coupling enterprises": 9
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/hard001.json b/assets/qa_gold/industry_planning/hard001.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d46ef95648ad0a176583884804748d4f9694e79
--- /dev/null
+++ b/assets/qa_gold/industry_planning/hard001.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard001",
+  "question": "基于2022年的数据，现对中国大陆半导体行业各省（相关企业数量>=5）进行政策分化情景推演：凡已落地集成电路产业发展促进类专项政策的省份，其辖内半导体企业可维持现有研发扩张节奏（以省内各企业研发投入同比增减幅的中位数为准）；而尚未出台上述专项政策的省份，受政策缺位影响，预计其研发增速将压缩至原有水平的一半。在此分化情景下，以3年复合增长方式推算至2025年，请问届时研发投入规模居于各省首位的是哪个省份？该省预测研发投入总额约为多少亿元？",
+  "guidelines": "依次回答省份名称和预计研发投入总额。金额以亿元为单位，保留2位小数。如[\"浙江省\", 312.75]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "上海市",
+    508.26
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "从policy_resource.csv筛选政策使用行业包含\"集成电路\"或\"半导体\"的政策，找到44条相关政策，均为地方政策，涉及广东省（10条）、上海市（5条）、安徽省（4条）、四川省（3条）、山东省（2条）、重庆市（2条）、福建省（1条）、海南省（1条）、河南省（1条）、湖南省（1条）、江西省（1条）、陕西省（1条）、新疆维吾尔自治区（1条）、云南省（1条）、浙江省（1条）。",
+    "从policy_resource.csv中读取这44条政策的全文内容，逐一分析其是否包含集成电路产业发展促进的实质性内容。经分析确认，出台了集成电路产业发展促进专项政策：广东省（id=80、153、249、605）、上海市（id=201、398、461）、安徽省（id=4、301）、四川省（id=387）、山东省（id=284）、浙江省（id=125）。",
+    "从company_profile.csv筛选行业=\"半导体业\"的企业，共172家。排除港澳台地区后剩余160家，按省份统计企业数，筛选企业数>=5的省份共6个：广东省54家、上海市27家、江苏省23家、浙江省13家、北京市10家、湖北省5家。",
+    "从company_operation_status.csv提取上述6个省份半导体业企业的研发投入金额和研发投入同比增减幅，计算各省研发投入合计和增速中位数：上海市220.03亿元/32.19%、广东省140.27亿元/8.87%、江苏省56.12亿元/21.06%、北京市46.75亿元/20.13%、浙江省22.14亿元/18.33%、湖北省19.30亿元/39.54%。",
+    "根据假设条件，有专项政策的省份（上海市、广东省、浙江省）按原增速中位数增长，无专项政策的省份（江苏省、北京市、湖北省）增速减半。安徽省虽有政策但企业数不足5家，不参与计算。",
+    "以3年复合增长方式计算各省2025年研发投入预测：上海市=220.03×(1+32.19%)^3=508.26亿元、广东省=140.27×(1+8.87%)^3=180.98亿元、江苏省=56.12×(1+10.53%)^3=75.78亿元、北京市=46.75×(1+10.07%)^3=62.34亿元、浙江省=22.14×(1+18.33%)^3=36.69亿元、湖北省=19.30×(1+19.77%)^3=33.16亿元。",
+    "上海市以508.26亿元排名第一，预计2025年研发投入总额最高。"
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "集成电路/半导体相关政策总数（条）": 4,
+    "经政策全文分析确认的省份数（个）": 4,
+    "半导体业企业总数（家）": 172,
+    "大陆半导体业企业数（家）": 160,
+    "企业数>=5的省份数（个）": 6,
+    "上海市半导体企业数（家）": 27,
+    "上海市2022研发投入合计（亿元）": 220.03,
+    "上海市研发增速中位数（%）": 32.19,
+    "上海市2025预测研发投入（亿元）": 508.26
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/hard002.json b/assets/qa_gold/industry_planning/hard002.json
new file mode 100644
index 0000000000000000000000000000000000000000..2d4e92e0cb6eae5ace3f2bd8f27036b04388aaa6
--- /dev/null
+++ b/assets/qa_gold/industry_planning/hard002.json
@@ -0,0 +1,38 @@
+{
+  "id": "hard002",
+  "question": "以2022年化学原料和化学制品制造业的实际数据为起点，构建如下政策差异化情景：若某省（仅纳入中国大陆化学原料和化学制品制造业存续企业数量不低于8家的省份，港澳台不在统计范围内）已发布新材料领域的产业相关发展政策且涉及具体目标和量化指标，则该省化工企业得以按照各自当前研发增速（取省内企业研发投入同比增减幅的中位数）持续推进研发扩张；反之，凡无此类专项政策的省份，其研发增速将以现有水平的一半计算。在3年复合增长模型下展望至2025年，试问：哪个省份的研发投入省际排名跃升幅度最为显著？该省届时预估的研发投入总规模是多少亿元？",
+  "guidelines": "依次回答省份名称和预计研发投入总额。研发投入总额以亿元为单位，保留2位小数。如[\"湖北省\", 72.31]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "湖南省",
+    25.13
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "从policy_resource.csv筛选industry字段包含\"化学原料和化学制品制造业\"的政策记录，找到60条化工相关政策，其中地方政策48条。",
+    "从policy_resource.csv中读取这48条地方政策的全文内容，分析哪些政策属于新材料产业专项发展政策（包含具体的新材料产业发展目标和量化指标），筛选出3个省份出台了此类专项政策：湖南省（id=491）、安徽省（id=177）、广东省（id=605）、江西省（id=380）、四川省（id=526）、贵州省（id=276）、河南省（id=87）、云南省（id=128）。",
+    "从company_profile.csv筛选行业=\"化学原料和化学制品制造业\"的企业，排除港澳台地区后共357家，按省份统计企业数量，筛选企业数>=8的省份，得到13个符合条件的省份。",
+    "从company_operation_status.csv提取这13个省份化工企业的研发投入金额和研发投入同比增减幅，计算各省2022年研发投入总额和增速中位数。广东省36.66亿元/9.28%、江苏省55.06亿元/9.44%、上海市37.78亿元/-1.19%、浙江省74.99亿元/13.14%、山东省130.14亿元/24.79%、四川省30.06亿元/18.20%、安徽省18.53亿元/12.78%、湖南省34.64亿元/34.64%、河南省18.32亿元/-8.38%、河北省11.07亿元/52.95%、辽宁省96.99亿元/15.98%、湖北省没有相关数据。",
+    "按照假设条件设定各省有效增速：有新材料专项政策的省份（湖南省、广东省、四川省、河南省）维持当前增速中位数，无政策省份增速减半。广东省9.28%、江苏省4.72%、上海市-0.595%、浙江省6.57%、山东省12.395%、四川省9.1%、安徽省6.3875%、湖南省34.64%、河南省-8.38%、河北省26.4725%、辽宁省7.99%",
+    "使用3年复合增长公式计算2025年研发投入总额：2025年研发投入 = 2022年研发投入 × (1 + 有效增速/100)^3。湖南省：10.29 × (1+34.64/100)^3 = 25.13亿元（排名从第10升至第7）。",
+    "比较各省排名变化：湖南省上升3位（第10→第7），上海市下降3位（第4→第7），河北省上升2位（第10→第8）。排名上升幅度最大的省份为湖南省，上升3位，2025年预计研发投入总额25.13元。"
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "化工相关政策总数（条）": 60,
+    "化工地方政策数（条）": 48,
+    "新材料专项政策省份数（个）": 3,
+    "中国大陆化工企业总数（家）": 357,
+    "符合条件省份数（企业>=8）": 13,
+    "湖南省化工企业数（家）": 19,
+    "湖南省2022年研发投入总额（亿元）": 18.32,
+    "湖南省研发增速中位数（%）": 24.64,
+    "湖南省2025年预计研发投入（亿元）": 25.13,
+    "四川省2022年排名": 10,
+    "四川省2025年排名": 7,
+    "四川省排名上升（位）": 3
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/hard003.json b/assets/qa_gold/industry_planning/hard003.json
new file mode 100644
index 0000000000000000000000000000000000000000..b27acf9f1fa6fea9be2f33724a83645fbc8067af
--- /dev/null
+++ b/assets/qa_gold/industry_planning/hard003.json
@@ -0,0 +1,38 @@
+{
+  "id": "hard003",
+  "question": "针对中国大陆医药制造业，以2022年各省数据为基准，现设定如下政策效应假设：已颁布生物医药产业发展促进相关政策的省份（仅计入中国大陆医药制造业企业数量不少于8家的省级行政区，不含港澳台地区），其辖区内医药企业在未来三年内营业收入年增速可在原有基础上额外叠加5个百分点（原增速以该省全部医药企业营业收入同比增减幅的中位数衡量）；而那些尚未出台此类促进政策的省份，因缺乏政策催化，营业收入增速将较现有水平收缩20%。在上述差异化情景下，对各省营业收入以3年复合增长方式推算至2025年，请问：哪个省份在这轮重新洗牌后实现了最大幅度的营收排名晋升？对应的2025年预计营业收入总量为多少亿元？",
+  "guidelines": "依次回答省份名称和预计营业收入总额。营业收入总额以亿元为单位，保留2位小数。如[\"湖北省\", 425.18]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "河南省",
+    350.45
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "从policy_resource.csv筛选正文包含\"生物\"，适用行业包含\"医药制造业\"的地方政策，找到55条生物医药产业发展相关地方政策，涉及9个省份：上海市（11条）、广东省（5条）、云南省（5条）、河南省（3条）、黑龙江省（3条）、山东省（2条）、天津市（2条）、浙江省（2条）、安徽省（1条）、甘肃省（1条）、广西壮族自治区（1条）、湖南省（1条）、吉林省（1条）、江苏省（1条）、江西省（1条）、四川省（1条）、新疆维吾尔族自治区（1条）。",
+    "从policy_resource.csv中读取这15条政策的全文内容，涉及生物医药产业发展促进相关政策省份有上海市（id=139、398、461、590、397、449、472、495）、广东省（id=92、153、605、325）、云南省（id=141、400）、河南省（id=559）、黑龙江（id=210、211、560）、山东省（id=181、284、446、天津市（id=447）、浙江省（id=393、445）、安徽省（id=444）、广西壮族自治区（id=273）、江苏省（id=443）、江西省（id=89）、四川省（id=387）、新疆维吾尔族自治区（id=556）",
+    "从company_profile.csv筛选industry=\"医药制造业\"的企业，关联company_operation_status.csv获取营业收入金额和营业收入同比增减幅，排除港澳台后得到420家有完整数据的企业。",
+    "按省份统计企业数，筛选企业数不少于8家的省份，得到14个符合条件的省份：北京市（55家）、广东省（51家）、江苏省（53家）、上海市（54家）、浙江省（45家）、山东省（22家）、四川省（15家）、湖北省（13家）、湖南省（11家）、吉林省（11家）、河南省（9家）、天津市（8家）、重庆市（8家）、福建省（8家）。",
+    "计算各省份2022年营业收入合计和营业收入同比增减幅中位数，确定2022年排名：北京市（4367.17亿，第1名）、广东省（3321.38亿，第2名）、上海市（1600.61亿，第3名）、浙江省（1423.10亿，第4名）、江苏省（1081.40亿，第5名）、山东省（960.48亿，第6名）、四川省（363.81亿，第7名）、湖南省（235.97亿，第8名）、吉林省（234.38亿，第9名）、河南省（188.79亿，第10名），其他省份缺失数据。",
+    "按假设条件调整增速：有政策省份增速额外加5个百分点（如河南省从17.90%调整为22.90%，广东省从9.93%调整为14.93%），无政策省份增速衰减20%（如北京市从3.64%调整为2.91%，山东省从7.53%调整为6.02%）。",
+    "以调整后增速进行3年复合增长预测各省2025年营收：广东省5042.83亿（第1名）、北京市4760.47亿（第2名）、上海市2637.73亿（第3名）、浙江省2137.93亿（第4名）、江苏省1802.98亿（第5名）、山东省1368.47亿（第6名）、四川市462.34亿（第7名）、河南省350.45亿（第8名）、湖南省296.34亿（第9名）、吉林省204.88亿（第10名）。",
+    "比较排名变化：河南省从第10名上升至第8名，排名上升2位，为上升幅度最大的省份"
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "生物医药/生物经济地方政策总数（条）": 55,
+    "已颁布生物医药产业发展促进相关政策的省份数（个）": 14,
+    "医药制造业有效数据企业数（家）": 420,
+    "企业数>=8的省份数（个）": 14,
+    "河南省2022年营收合计（亿元）": 188.79,
+    "河南省营收增速中位数（%）": 17.9,
+    "河南省调整后增速（%）": 22.9,
+    "河南省2025年预计营收（亿元）": 350.45,
+    "河南省2022年排名": 10,
+    "河南省2025年排名": 8,
+    "排名上升幅度（位）": 2
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/hard004.json b/assets/qa_gold/industry_planning/hard004.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ff470d0ced56347644115206e0052e4663d79d8
--- /dev/null
+++ b/assets/qa_gold/industry_planning/hard004.json
@@ -0,0 +1,37 @@
+{
+  "id": "hard004",
+  "question": "以2022年数据为起点，对中国大陆汽车制造业各省（统计范围限定为汽车制造业在册企业不少于5家的中国大陆省份，不含港澳台）进行如下情景模拟：已出台地方性新能源汽车及智能汽车产业发展专项政策的省份，其汽车制造业企业营业收入年增速将在当前中位增速基础上再叠加3个百分点（当前增速以各省企业营业收入同比增减幅中位数为准）；未出台此类专项政策的省份则呈现增长动力不足的局面，营业收入（增速取各省企业营业收入同比增减幅的中位数））增速将萎缩至原有水平的70%。按3年复合增长推算至2025年，请找出：在拥有政策支持的省份中，哪个省份将凭借政策加持超越其2022年时排名本高于自身的某个无政策省份，从而实现排名反超？该省届时的预计营业收入总量是多少亿元？",
+  "guidelines": "依次回答省份名称和预计营业收入总额。营业收入总额以亿元为单位，保留2位小数。如[\"广东省\", 5230.41]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "北京市",
+    4850.79
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "从policy_resource.csv筛选政策名称包含\"新能源汽车\"或\"燃料电池汽车\"或\"汽车\"的政策记录，找到13条新能源汽车相关地方政策。",
+    "从policy_resource.csv中读取这12条政策的全文内容，分析各地方政策的具体产业促进措施。涉及省份共8个，分别是北京市（id=71）、广东省（id=58、157、334、518）、海南省（id=212）、江苏省（id=601）、山东省（id=599）、上海市（id=499）、四川省（id=525）、重庆市（id=261、371）。",
+    "从company_profile.csv筛选行业=\"汽车制造业\"且省份不含港澳台的企业，按省份统计企业数量，筛选企业数>=5的省份，得到14个符合条件的省份，分别是浙江省、江苏省、广东省、上海市、山东省、北京市、湖北省、安徽省、河北省、河南省、四川省、重庆市、福建省、吉林省。",
+    "从company_operation_status.csv提取这14个省份汽车制造业企业的营业收入金额和营业收入同比增减幅，按省份计算营业收入合计和增速中位数。广东省（12407.19亿元，增速18.030%）、上海市（10315.04亿元，增速14.505%）、山东省（5249.64亿元，增速1.320%）、浙江省（3892.24亿元，增速10.120%）、北京市（3415.96亿元，增速9.400%）、河北省（3138.58亿元，增速3.295%）、江苏省（1032.08亿元，增速9.4%）、吉林省（765.19亿元，增速-2.79）、安徽省（504.95亿元，增速-5.735）、河南省（335.81亿元、增速-6.105），其余省份没有数据披露。",
+    "根据假设条件调整增速：有政策省份增速=原增速+3个百分点，无政策省份增速=原增速×0.7。",
+    "按3年复合增长预测2025年营收：北京市=3415.96×(1+12.400/100)^3=4850.79亿元，浙江省=3892.24×(1+7.084/100)^3=4779.40亿元。北京市2025年营收（4850.79亿）超过浙江省（4779.40亿），排名从第5升至第4，是唯一一个超越原排名更高的无政策省份的有政策省份。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "新能源汽车相关政策总数（条）": 12,
+    "地方性政策数（条）": 9,
+    "有政策省份数（个）": 7,
+    "符合条件省份数（企业>=5）": 14,
+    "北京市2022年营收合计（亿元）": 3415.96,
+    "北京市营收增速中位数（%）": 9.4,
+    "北京市调整后增速（%）": 12.4,
+    "浙江省2022年营收合计（亿元）": 3892.24,
+    "浙江省营收增速中位数（%）": 10.12,
+    "浙江省调整后增速（%）": 7.084,
+    "北京市2025年预计营收（亿元）": 4850.79,
+    "浙江省2025年预计营收（亿元）": 4779.4
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/hard005.json b/assets/qa_gold/industry_planning/hard005.json
new file mode 100644
index 0000000000000000000000000000000000000000..5cf5080d8e2ee6299183b74e19d76fa68f8d5b09
--- /dev/null
+++ b/assets/qa_gold/industry_planning/hard005.json
@@ -0,0 +1,34 @@
+{
+  "id": "hard005",
+  "question": "云南省于2022年正式发布了绿色铝产业发展三年行动计划，其中载明了2024年绿色铝全产业链产值的量化目标。若以该省金属冶炼和压延加工业所有上市企业（仅覆盖中国大陆范围内的金属冶炼和压延加工业企业，不含港澳台数据））2022年的实际营业收入为基数，并假设各企业按自身现有增速（取全部上市企业营业收入同比增减幅的中位数作为统一测算基准）持续保持增长，经2年复合增长推算至2024年，请问：届时这些上市企业的营收汇总值与政策文件明确设定的产业链产值目标之间还存在多大缺口（以亿元计）？若要在2年内完全弥合上述缺口，在现有增速基础上还需年均额外拉升多少个百分点？",
+  "guidelines": "依次回答缺口金额和额外增速。缺口金额以亿元为单位，保留2位小数；额外增速以百分点为单位，保留2位小数。如[280.50, 4.12]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    356.36,
+    5.85
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "从policy_resource.csv筛选industry字段包含\"金属冶炼\"且province为\"云南省\"的政策记录，找到4条云南省金属冶炼和压延加工业相关政策。",
+    "从policy_resource.csv中读取这4条政策的全文内容，分析政策中关于产业发展的具体目标。在\"云南省绿色铝产业发展三年行动（2022—2024年）\"中明确提出行动目标：\"绿色铝产业链产值力争达到3500亿元左右\"，该目标年限为2024年。",
+    "从company_profile.csv筛选industry=\"金属冶炼和压延加工业\"且province=\"云南省\"的企业，找到7家上市企业。",
+    "从company_operation_status.csv提取这7家企业的营业收入金额和营业收入同比增减幅。7家企业2022年营业收入总额为279203336972.01元（约2792.03亿元），营业收入同比增减幅中位数为6.11%。",
+    "按2年复合增长预测2024年营业收入：2792.03×(1+6.11%)^2 = 3143.64亿元。与政策目标3500亿元的缺口 = 3500 - 3143.64 = 356.36亿元。",
+    "计算弥补缺口所需的年增速：设所需年增速为r，则2792.03×(1+r)^2 = 3500，解得r = √(3500/2792.03) - 1 = 11.96%。因此需要在当前6.11%的增速基础上额外提升 11.96 - 6.11 = 5.85个百分点。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "云南省金属冶炼相关政策数（条）": 4,
+    "政策产值目标（亿元）": 3500,
+    "云南省金属冶炼企业数（家）": 7,
+    "2022年营业收入总额（亿元）": 2792.03,
+    "营业收入同比增减幅中位数（%）": 6.11,
+    "2024年预测营业收入（亿元）": 3143.64,
+    "产值缺口（亿元）": 356.36,
+    "达标所需年增速（%）": 11.96,
+    "额外增速（百分点）": 5.85
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/hard006.json b/assets/qa_gold/industry_planning/hard006.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c3834b8f623498f36f5c3d2f843faae2e8d4241
--- /dev/null
+++ b/assets/qa_gold/industry_planning/hard006.json
@@ -0,0 +1,38 @@
+{
+  "id": "hard006",
+  "question": "聚焦2022年中国大陆非金属矿物制品业，考察以下政策差异化情景对各省（非金属矿物制品业企业数不少于5家的中国大陆省级行政区，不纳入港澳台数据）研发格局的重塑效果：凡已发布专门涉及建材行业绿色转型内容的省级碳达峰或节能减排专项实施方案的省份，其辖内非金属矿物制品业企业可保持现有研发投入增速不变（增速以该省各企业研发投入同比增减幅的中位数为准）；而尚未落地此类省级专项方案的省份，其研发扩张动能将打折，增速降至当前水平的50%。以上述差异化增速进行3年复合增长测算，预测各省2025年研发投入规模并进行重新排序，请问：从2022年到2025年，哪个省份实现了最大幅度的研发投入排名跃升？该省2025年研发投入总额约为多少亿元？",
+  "guidelines": "依次回答省份名称和预计研发投入总额。总额以亿元为单位，保留2位小数。如[\"湖北省\", 20.04]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "湖南省",
+    31.15
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "从policy_resource.csv筛选industry字段包含\"非金属矿物制品业\"的地方政策记录，找到30条相关政策。",
+    "从policy_resource.csv中读取这30条政策的全文内容，筛选其中属于地方政策且政策名称含\"实施方案\"或\"工作方案\"的30条地方政策，分析涉及碳达峰或节能减排，并对建材行业（水泥、玻璃、陶瓷等）有实质性内容的政策共条。",
+    "经过对政策内容的分析，筛选出7条符合条件的地方政策，涉及6个省份：湖南省（id=492）。四川省（id=116）、安徽省（id=504），贵州（id=388），江西省(id=106)，辽宁省（id=161）",
+    "从company_profile.csv筛选行业=\"非金属矿物制品业\"且不含港澳台的中国大陆企业，共125家。按省份统计企业数，筛选企业数不少于5家的省份，得到11个符合条件的省份。",
+    "从company_operation_status.csv提取这11个省份非金属矿物制品业企业的研发投入金额和研发投入同比增减幅，计算各省2022年研发投入总额和增速中位数。其中有政策的省份为安徽省（41.42亿元，52.64%）、湖南省（8.34亿元，55.19%）、四川省（1.12亿元，32.24%），无政策的省份包括广东省（31.93亿元，-1.73%）、北京市（24.94亿元，18.88%）、浙江省（23.51亿元，12.13%）等。",
+    "按假设条件调整增速：有政策省份维持当前增速，无政策省份增速减半。以3年复合增长计算2025年研发投入总额。例如湖南省：8.34×(1+55.19%)^3=8.34×3.7376=31.15亿元。",
+    "计算2022年和2025年各省排名变化：湖南省从第9名上升至第3名，排名上升6位，为上升幅度最大的省份。"
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "非金属矿物制品业相关政策总数（条）": 39,
+    "碳达峰/节能减排建材专项地方政策数（条）": 7,
+    "涉及省份数（个）": 6,
+    "非金属矿物制品业大陆企业数（家）": 125,
+    "符合条件省份数（企业>=5家）": 11,
+    "湖南省企业数（家）": 5,
+    "湖南省2022年研发投入总额（亿元）": 8.34,
+    "湖南省研发投入增速中位数（%）": 55.19,
+    "湖南省2022年研发投入排名": 9,
+    "湖南省2025年预计研发投入总额（亿元）": 31.15,
+    "湖南省2025年研发投入排名": 3,
+    "湖南省排名上升幅度（位）": 6
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/hard007.json b/assets/qa_gold/industry_planning/hard007.json
new file mode 100644
index 0000000000000000000000000000000000000000..834b6696fb0596e9dfcc3b5970c81b600678ef1c
--- /dev/null
+++ b/assets/qa_gold/industry_planning/hard007.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard007",
+  "question": "就2022年中国大陆软件和信息技术服务业的人才规模格局而言，若设定如下政策导向假设（仅含软件和信息技术服务业存续企业数量达到5家及以上门槛的中国大陆省份，港澳台数据不纳入计算）：凡已正式出台软件产业高质量发展专项政策的省份，其企业员工总量可沿现有轨道持续扩张，年增速以该省各企业雇员同比增减幅的中位数+13%为准；而对于尚未出台此类专项政策的省份，因政策引领欠缺导致人才吸附力不足，员工扩张速度将缩减至现有增速的一半。在这一情景设定下以3年复合增长推算至2025年，哪个省份在员工总量省际排名中实现了最大幅度的正向位次变动？请一并报告该省届时预计的从业人员总规模。",
+  "guidelines": "依次回答省份名称和预计员工总人数。员工总人数保留2位小数，单位为万人。如[\"四川省\", 2.58]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "安徽省",
+    3.49
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "从policy_resource.csv筛选industry字段包含'软件和信息技术服务业'的政策，共找到206条相关政策，其中地方政策142条。",
+    "从policy_resource.csv中读取全文与\"软件\"行业相关的5条地方政策全文，经过对政策内容的分析，提取出台了软件产业高质量发展专项政策且在本题统计范围内的省份：上海市（id=201、360、397）、广东省（id=153、249、516、540、589、605）、安徽省（id=4、177、67）、山东省（id=175、435、489、600）、海南省（id=315）、福建省（id=104）、湖北省（id=260、598）、云南省（id=252）、重庆市（id=96）、河南省（id=331）、山西省（id=460）",
+    "从company_profile.csv筛选industry为信息传输、软件和信息技术服务业且省份不含港澳台的企业，共620家，按省份统计企业数，筛选企业数不少于5家的省份共12个：北京市（197家）、广东省（125家）、上海市（73家）、浙江省（59家）、江苏省（43家）、福建省（27家）、四川省（17家）、山东省（17家）、湖北省（9家）、湖南省（7家）、吉林省（6家）、安徽省（5家）。",
+    "从company_operation_status.csv获取上述12个省份各企业的雇员总数和雇员同比增减幅，按省份汇总雇员总数合计与雇员同比增减幅中位数。各省雇员总数（人）和中位数增速（%）如下：广东省（361555人、-0.385%）、北京市（3836316人、-0.15%）、江苏省（96880人、1.45%）、上海市（28077人、-1.3%）、浙江省（647855人、-4.21%）、山东省（36446人、6.17%）、四川省（31651人、3.68%）、安徽省（28505人、-6.02%）、湖南省（9413人、0.75%）、吉林省（12133人、3.745）、其他省份缺失数据。",
+    "确定各省适用增速：有政策省份（上海市、广东省、山东省、安徽省）共4个，使用原始中位数增速+13%，无政策省份增速减半。以3年复合增长计算2025年预测雇员总数：2025年雇员 = 2022年雇员 × (1 + 适用增速)^3。有政策省份：广东省361555×(1-((-0.385)+13)/100)^3≈516372人；上海市280077×(1-(0.012+13)/100)^3≈391384人；山东省280077×(1-(6.17+13)/100)^3≈61681人；安徽省28505×(1-(-6.02+13)/100)^3≈34900。无政策省份以减半增速计算。",
+    "对比各省2022年排名和2025年预测排名：2022年安徽省雇员总数28505人，排名第8位；2025年预测34900人，超越四川省（32533人），上升至第7位，排名提升1位，为所有省份中排名上升幅度最大的省份。安徽省2025年预计员工总人数34900人人，约合3.49万人。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "软件相关地方政策总数（条）": 142,
+    "地方政策涉及软件高质量发展的省份数量": 11,
+    "有政策且企业数>=5的省份数（个）": 4,
+    "企业数>=5的省份总数（个）": 12,
+    "安徽省2022年雇员总数（人）": 28505,
+    "安徽省2022年雇员同比增减幅中位数（%）": -6.02,
+    "安徽省2025年预测雇员（人）": 34900,
+    "安徽省2022年排名（位）": 8,
+    "安徽省2025年排名（位）": 7,
+    "四川省2025年预测雇员（人）": 12827.41
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/hard008.json b/assets/qa_gold/industry_planning/hard008.json
new file mode 100644
index 0000000000000000000000000000000000000000..d1a7952983846bf3c29e6a045514d2bae822aae9
--- /dev/null
+++ b/assets/qa_gold/industry_planning/hard008.json
@@ -0,0 +1,33 @@
+{
+  "id": "hard008",
+  "question": "以2022年为基期，针对中国大陆通用设备制造业设计双情景对比测算框架（通用设备制造业在册企业数量不少于5家的中国大陆省份，港澳台不纳入）。情景一（政策分化情景）：已出台制造业高质量发展省级专项文件的省份，其通用设备制造业企业总资产按各自当前增速+6%（以省内各企业营业收入同比增减幅的中位数作为总资产增速的替代指标）持续扩张，无政策省份则以减半增速计算；情景二（全量减半基准情景）：所有省份不论有无政策，一律按当前增速的一半推算总资产增长。以3年复合增长分别推算两种情景下各省2025年总资产，并以两情景之差作为\"政策带来的额外总资产增量\"，请问：在出台了上述专项政策且符合最低企业数量门槛的省份中，哪个省份从该政策中撬动的额外总资产增量（额外增量=政策分化情景下2025年总资产 - 全量减半情景下2025年总资产）最为可观？该增量具体为多少亿元？",
+  "guidelines": "依次回答省份名称和额外总资产增量。额外增量保留2位小数，单位为亿元。如[\"湖南省\", 520.38]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "广东省",
+    140.38
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "从policy_resource.csv筛选政策类型为地方政策且政策名称包含制造业的政策，共找到7条候选政策。",
+    "从policy_resource.csv中读取上述7条政策全文内容，深入分析各政策的涉及制造业高质量发展的省份有：福建省（id=78、）、上海市（id=75、398）、广东省（id=154）、湖南省（id=184）。",
+    "从company_profile.csv筛选行业为通用设备制造业的企业，排除港澳台，共得到212家中国大陆通用设备制造业上市企业；按省份统计企业数量，筛选企业数不少于5家的省份，得到11个合规省份：浙江省（62家）、江苏省（42家）、广东省（20家）、山东省（17家）、上海市（12家）、四川省（10家）、辽宁省（7家）、安徽省（6家）、湖北省（6家）、福建省（5家）、湖南省（5家）。",
+    "从company_operation_status.csv提取11个合规省份通用设备制造业企业的总资产和营业收入同比增减幅，按省份计算总资产合计（亿元）和营业收入同比增减幅中位数（%），分别为广东省总资产合计1536.03亿元、中位数增速-5.745%；江苏省总资产合计1557.78亿元、中位数增速0.635%；上海市总资产合计7001.17亿元、中位数增速-10.94%；浙江省3177.50亿元、中位数增速1.18%；山东省633.65亿元、中位数增速9.19%；四川省2658.05亿元、中位数增速15.255%；安徽省167.97亿元、中位数增速-2.095%；湖南省222.13亿元、-10.57%；河北省57.30亿元、中位数增速66.665%；辽宁省374.25亿元、中位数增速10.12%；其他省份缺失数据。",
+    "对3个有政策省份分别计算两种情景下2025年总资产：情景A（有政策，原速增长）= 2022年总资产×(1+原增速/100)^3；情景B（全部减半）= 2022年总资产×(1+原增速/200)^3；额外增量=情景A-情景B。广东省：1547.81亿元-1407.43亿元=140.38亿元",
+    "三个有政策省份中，四川省以额外总资产增量755.66亿元排名第一，为从该政策中获得边际贡献最大的省份。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "制造业高质量发展专项政策涉及省份数（个）": 4,
+    "通用设备制造业合规省份总数（个）": 11,
+    "满足企业数不少于5家条件的有政策省份数（个）": 3,
+    "广东省2022年总资产合计（亿元）": 1547.81,
+    "广东省营业收入增速中位数（%）": -5.745,
+    "广东省2025年总资产（有政策情景，亿元）": 1547.81,
+    "广东省2025年总资产（全部减半情景，亿元）": 1407.43,
+    "广东省额外总资产增量（亿元）": 140.38
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/hard009.json b/assets/qa_gold/industry_planning/hard009.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc4be902e1457a9986187d637d1ca2371f630995
--- /dev/null
+++ b/assets/qa_gold/industry_planning/hard009.json
@@ -0,0 +1,36 @@
+{
+  "id": "hard009",
+  "question": "以2022年铁路、船舶、航空航天和其他运输设备制造业的省级（限定为铁路、船舶、航空航天和其他运输设备制造业辖内企业数量达到3家以上的中国大陆省级行政区，港澳台数据不纳入）数据为基准，模拟如下政策分化对行业格局的冲击：已出台船舶与海洋工程装备产业专项发展政策的省份，其企业营业收入增速在未来三年将在现有中位水平上额外叠加5个百分点；而没有落地此类专项政策的省份，受制于政策真空，营业收入增速将萎缩至现有水平的70%（增速以各省企业营业收入同比增减幅中位数为准）。按各省调整后的增速进行3年复合增长推算，对比2022年与2025年的省际营收排名变动，请从无政策省份中找出：哪个省份享受政策红利而获得最大幅度的排名上升？它一共上升了几位？该省2025年的预测营业收入总量为多少亿元？",
+  "guidelines": "依次回答省份名称、排名下降名数（整数）、该省2025年预计营业收入总额。营业收入总额保留2位小数，单位为亿元。如[\"广东省\", 2, 512.34]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "山东省",
+    2,
+    697.09
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "从policy_resource.csv筛选涉及行业包含「船舶」或「海洋工程」关键词的政策，共找到46条相关政策，其中地方政策36条。",
+    "从policy_resource.csv中读取46条相关政策的全文内容，对每条政策进行深度分析，判断其是否专门针对船舶与海洋工程装备制造产业。经分析，绝大多数政策属于泛行业或节能减排政策，仅有少数政策以船舶与海洋工程装备为核心支持方向：上海市（id=75、139、397）、河南省（id=87）、广东省（id=153、303）、山东省（id=181、284）。",
+    "从company_profile.csv中筛选行业为铁路、船舶、航空航天和其他运输设备制造业筛选有效省份，共得到8个有效省份：北京市（22家）、江苏省（14家）、广东省（9家）、浙江省（6家）、四川省（6家）、上海市（3家）、山东省（3家）、湖南省（3家）。",
+    "从regional_industry_status.csv筛选行业为铁路、船舶、航空航天和其他运输设备制造业、省份为中国大陆、上述省份营业收入金额合计及营业收入同比增减幅中位数数据为：广东省（307.73亿元，-1.44%）、北京市（7049.18亿元，-1.055%）、江苏省（561.84亿元，7.775%），上海市（620.95亿元，-0.3%），浙江省（264.27亿元，27.62%），山东省（431.65亿元，22.08%）、四川省（87.02亿元，10.23%）、湖南省（377.54亿元，19.23）。",
+    "按政策情景计算各省实际增速：有政策省份（上海市、广东省、山东省）在原增速基础上额外提升5个百分点；无政策省份增速衰减至原增速的70%。以2022年营业收入为基数，按3年复合增长计算各省2025年预计营业收入：广东省（330.03亿元）、北京市（6930.69亿元）、江苏省（633.91亿元），上海市（680.69亿元），浙江省（393.95亿元），山东省（697.09亿元）、四川省（101.85亿元）、湖南省（502.64亿元）",
+    "对比各省排名变化：山东省从第4名上升到第2名，上升2位，2025年预计营收697.09亿元。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "涉及船舶或海洋工程行业的政策总数（条）": 46,
+    "地方政策数（条）": 36,
+    "出台专项政策的省份数（个）": 4,
+    "有效省份总数（个）": 8,
+    "无政策省份总数（个）": 5,
+    "山东省2022年预计营收（亿元）": 431.65,
+    "山东省2025年预计营收（亿元）": 697.09,
+    "山东省2022年排名（名）": 4,
+    "山东省2025年排名（名）": 2,
+    "山东省排名上市幅度（名）": 2
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/hard010.json b/assets/qa_gold/industry_planning/hard010.json
new file mode 100644
index 0000000000000000000000000000000000000000..1408636ae41d94208226877660f99673a1b3a478
--- /dev/null
+++ b/assets/qa_gold/industry_planning/hard010.json
@@ -0,0 +1,34 @@
+{
+  "id": "hard010",
+  "question": "在铁路、船舶、航空航天和其他运输设备制造业（即广义轨道交通装备制造业）领域，以2022年为基期，构建以下轨道交通政策激励传导模型（轨道交通装备制造相关企业存续数量不低于3家的中国大陆省级行政区，港澳台不计入）：若某省已明确将轨道交通装备产业列入重点打造产业集群目录并配套专项支持措施，则该省企业营业收入的年增速可在现有中位水平上再叠加3个百分点；而未落地此类产业集群专项支持政策的省份，其营业收入增速将在原有基础上收缩30%（增速取省内各企业营业收入同比增减幅中位数）。以上述情景增速进行3年复合增长预测，并计算各有政策省份从2022年到2025年的营业收入绝对增量（定义为：2025年预测营业收入 − 2022年实际营业收入），请问：哪个获得政策支持的省份营业收入的绝对增量最为可观？该省这一增量数值为多少亿元？",
+  "guidelines": "依次回答省份名称和营业收入绝对增量。绝对增量保留2位小数。如[\"上海市\", 185.42]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "山东省",
+    413.38
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选涉及行业包含「铁路、船舶、航空航天和其他运输设备制造业」的政策，共找到46条，其中地方政策36条、部委政策10条。",
+    "从policy_resource.csv中读取上述36条地方政策的全文内容，逐一分析政策正文是否明确将轨道交通装备列为重点打造的产业集群并给予专项支持。经过深度内容分析，共6个省份有相关内容：上海市（id=75、139、397）、云南省（id=152）、广东省（id=605）、湖南省（id=194、417）、山东省（id=284）、四川省（id=387）。",
+    "从company_profile.csv筛选行业为「铁路、船舶、航空航天和其他运输设备制造业」的大陆企业（排除港澳台），共96家，按省份统计企业数量，筛选企业数不少于3家的省份，共12个：北京市（22家）、江苏省（14家）、广东省（9家）、陕西省（6家）、四川省（6家）、浙江省（6家）、黑龙江省（4家）、山东省（3家）、湖北省（3家）、湖南省（3家）、贵州省（3家）、上海市（3家）。",
+    "从company_operation_status.csv获取上述12个省份的铁路运输设备制造业企业营业收入金额和营业收入同比增减幅，计算各省营业收入合计（2022年基准值）和增速中位数。广东省：营业收入合计307.73亿元，增速中位数-1.44%；北京市：营业收入合计7049.18亿元，增速中位数-1.055%；江苏省：营业收入合计561.84亿元，增速中位数7.775%；上海市：营业收入合计620.95亿元，增速中位数-1.055%；浙江省：营业收入合计264.27亿元，增速中位数-0.3%；山东市：营业收入合计431.65亿元，增速中位数-22.08%；四川省：营业收入合计87.02亿元，增速中位数10.23%；湖南省：营业收入合计377.54亿元，增速中位数19.23%。",
+    "对有政策省份，按有效增速 = 增速中位数 + 3% 计算，3年复合增长预测2025年营业收入。山东省＝431.65*（1+（22.08+3）/100）=844.69亿元",
+    "比较有政策省份的营业收入绝对增量：山东省844.69亿元 - 431.65亿元=413.38亿元，山东省的营业收入绝对增量最大。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "铁路运输设备制造业相关政策总数（条）": 46,
+    "地方政策数量（条）": 36,
+    "有政策省份数（个）": 6,
+    "满足企业数>=3条件的省份数（个）": 12,
+    "山东省2022年营业收入合计（亿元）": 431.65,
+    "山东省增速中位数（%）": 22.08,
+    "山东省调整后增速（%）": 25.08,
+    "山东省2025年预测营业收入（亿元）": 844.69,
+    "山东省营业收入绝对增量（亿元）": 413.38
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/hard011.json b/assets/qa_gold/industry_planning/hard011.json
new file mode 100644
index 0000000000000000000000000000000000000000..81d46819dd741dae7cefeeae6c69570506c6c84e
--- /dev/null
+++ b/assets/qa_gold/industry_planning/hard011.json
@@ -0,0 +1,31 @@
+{
+  "id": "hard011",
+  "question": "在铁路、船舶、航空航天和其他运输设备制造业中，以2022年各省（港澳台地区数据不计入本题统计范围）实际数据为起点，设定如下情景假设：已颁布航空航天产业发展专项支持政策的省份，其行业内企业能够维持现有营业收入增速（以各省企业营业收入同比增减幅的中位数衡量）持续扩张；而未出台此类专项支持政策的省份，由于缺乏政策引导，营业收入增速将被动压缩至当前水平衰减40%。在以上差异化条件下按3年复合增长推算至2025年，请重点关注2022年营业收入总量尚未跨越100亿元门槛的有政策省份——在这一子集中，哪个省份将在政策扶持下率先实现百亿营收的历史性突破？该省届时的预测营业收入总额为多少亿元？",
+  "guidelines": "依次回答省份名称和2025年预计营业收入总额。营业收入总额保留2位小数，单位为亿元。如[\"江西省\", 89.37]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "四川省",
+    116.56
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选policyClassification为地方政策且industry包含铁路、船舶、航空航天和其他运输设备制造业的记录，得到36条地方政策。",
+    "从policy_resource.csv中读取上述36条政策的全文内容，逐条分析政策正文中是否含有针对航空航天产业的专项支持措施。经分析共有个6省份有相关政策，分别是：上海市（id=75、139）、广东省（id=153）、湖北省（id=172）、湖南省（id=194）、江西省（id=89）、四川省（id=387）。",
+    "从company_profile.csv筛选industry为铁路、船舶、航空航天和其他运输设备制造业且省份不含港澳台的企业，共得到96家企业，分布在21个省份。北京市（22家）、江苏省（14家）、广东省（9家）、陕西省（6家）、四川省（6家）、浙江省（6家）、黑龙江省（4家）、山东省（3家）、湖北省（3家）、湖南省（3家）、贵州省（3家）、上海市（3家）、安徽省（2家）、河北省（2家）、河南省（2家）、江西省（2家）、重庆市（2家）、吉林省（1家）、内蒙古自治区（1家）、山西省（1家）、天津市（1家）",
+    "从company_operation_status.csv提取上述96家企业的营业收入金额和营业收入同比增减幅，按省份汇总，计算各省营业收入总额和增速中位数。有政策省份中，2022年营业收入总额低于100亿元的共2个：四川省（87.02亿元，中位增速10.23%）、安徽省（12.16亿，中位增速18.66）、吉林省（4.23亿元，中位增速-3.50%）。无政策省份低于100亿元的有：安徽省（14.16亿元）、山西省（12.45亿元）、吉林省（4.23亿元）。",
+    "对2022年营业收入低于100亿元的有政策省份按3年复合增长计算2025年预计营业收入：四川省：87.024112亿元 × (1 + 10.23/100)^3 = 116.5572亿元，超过100亿元门槛（2024年即已突破）;安徽省：14.1649亿元 × (1 + 18.66*0.6/100)^3 = 19.4752亿元，未超过100亿元门槛;吉林省：4.2264亿元 × (1 + 3.42*0.6/100)^3 = 4.4919亿元，未超过100亿元门槛。",
+    "结论：在2022年营业收入低于100亿元的有政策省份中，四川省将首次突破100亿元门槛，2025年预计营业收入总额为116.56亿元。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "分析地方政策数量（条）": 36,
+    "认定为有航空航天专项政策的省份数（个）": 4,
+    "有政策且2022年营业收入低于100亿元的省份数（个）": 3,
+    "四川省2022年营业收入合计（亿元）": 87.02,
+    "四川省营业收入中位增速（%）": 10.23,
+    "四川省2025年预计营业收入（亿元）": 116.56
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/hard012.json b/assets/qa_gold/industry_planning/hard012.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb363db25dacf198c7eb3bef7f140c601e4cff56
--- /dev/null
+++ b/assets/qa_gold/industry_planning/hard012.json
@@ -0,0 +1,34 @@
+{
+  "id": "hard012",
+  "question": "2022年，在中国大陆医疗仪器设备及器械制造业中（仅统计仪器仪表制造业企业不少于3家的中国大陆省份，不含港澳台），假设出台了医疗器械产业高端化（适用于医药制造和仪表仪器）发展专项政策的省份，其企业净利润增速在未来3年额外提升5个百分点，而未出台此类政策的省份净利润增速衰减20%，到2025年净利润总额排名上升幅度最大的省份是哪个（增速使用各省份企业净利润同比增减幅的中位数；增长方式为3年复合增长）？其预计净利润总额是多少亿元？",
+  "guidelines": "依次回答省份名称和预计净利润总额。净利润总额保留2位小数。如[\"安徽省\", 6.92]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "北京市",
+    4.34
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "从regional_industry_status.csv筛选行业为仪器仪表制造业、省份为中国大陆（排除台湾省、香港特别行政区、澳门特别行政区）且企业总数不少于3家、净利润金额合计和净利润同比增减幅中位数均非空的省份，共得到9个符合条件的省份：浙江省（20家）、江苏省（14家）、广东省（12家）、上海市（7家）、北京市（8家）、河南省（4家）、安徽省（4家）、湖南省（3家）、四川省（3家）。",
+    "从policy_release_status.csv筛选上述9个省份的仪器仪表制造业及医药制造业相关地方政策，共38条；",
+    "从policy_resource.csv中读取上述重点政策的全文内容，共4个省份出台了医疗器械产业高端化相关政策，分别是：上海市（id=75、590）、广东省（id=92、153、341）、江苏省（id=443）、安徽省（id=444）",
+    "根据政策情景计算各省份调整后净利润增速：有政策省份的调整增速 = 净利润同比增减幅中位数 + 5个百分点；无政策省份的调整增速 = 净利润同比增减幅中位数 × 0.8（衰减20%）。各省调整后增速：浙江省 -0.155%×0.8 = -0.124%；江苏省 3.615%+5 = 8.615%；广东省 -9.585%+5 = -4.585%；湖南省 7.670%×0.8 = 6.136%；河南省 -37.150%×0.8 = -29.720%；安徽省 10.675%+5 = 15.675%；上海市 -45.100%+5 = -40.100%；北京市 20.745%×0.8 = 16.596%；四川省 -3.500%×0.8 = -2.800%。",
+    "以2022年净利润金额合计为基数，按各省调整后增速进行3年复合增长，计算2025年预计净利润总额：浙江省 33.5927亿×(1-0.00124)³ = 33.47亿；江苏省 15.6321亿×(1+0.08615)³ = 20.03亿；广东省 8.5411亿×(1-0.04585)³ = 7.42亿；湖南省 6.6674亿×(1+0.06136)³ = 7.97亿；河南省 6.0233亿×(1-0.2972)³ = 2.09亿；安徽省 4.4203亿×(1+0.15675)³ = 6.84亿；上海市 4.0677亿×(1-0.401)³ = 0.87亿；北京市 2.7353亿×(1+0.16596)³ = 4.34亿；四川省 1.2727亿×(1-0.028)³ = 1.17亿。",
+    "按2025年预计净利润总额排名：第1浙江省（33.47亿）、第2江苏省（20.03亿）、第3湖南省（7.97亿）、第4广东省（7.42亿）、第5安徽省（6.84亿）、第6北京市（4.34亿）、第7河南省（2.09亿）、第8四川省（1.17亿）、第9上海市（0.87亿）。与2022年排名对比，北京市从第8名升至第6名，上升2位，为排名上升幅度最大的省份；其2025年预计净利润总额为4.34亿元。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "符合条件省份数（个）": 9,
+    "出台专项政策省份数（个）": 4,
+    "北京市2022净利润合计（亿元）": 2.74,
+    "北京市净利润增速中位数（%）": 20.745,
+    "北京市调整后增速（%）": 16.596,
+    "北京市2022排名（名）": 8,
+    "北京市2025预计净利润（亿元）": 4.34,
+    "北京市2025排名（名）": 6,
+    "北京市排名上升幅度（位）": 2
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/hard013.json b/assets/qa_gold/industry_planning/hard013.json
new file mode 100644
index 0000000000000000000000000000000000000000..be7dc5d3ead641d5cd991fd622f24fe24ef7a294
--- /dev/null
+++ b/assets/qa_gold/industry_planning/hard013.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard013",
+  "question": "假设针对金属冶炼和压延加工业出台了新材料产业发展专项扶持政策的省份（仅统计2022年净利润数据完整的中国大陆省份，不含港澳台），其金属冶炼和压延加工业上市企业净利润按2022年的增速（各省企业净利润同比增减幅中位数）持续增长，而未出台此类政策的省份净利润增速减半，在3年复合增长模型下，到2025年有政策省份净利润总和与无政策省份净利润总和的比值是多少？相比2022年该比值变化了多少（提升或下降）？",
+  "guidelines": "依次回答2025年的比值和比值变化量。比值保留2位小数，变化量保留2位小数并注明提升或下降。如[1.85, \"提升0.93\"]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    3.61,
+    "提升1.79"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "从policy_resource.csv筛选industry字段包含\"金属冶炼和压延加工业\"的政策记录，共找到21条相关政策，其中地方政策19条。",
+    "从policy_resource.csv中读取上述19条政策的全文内容，进行深度分析，出台推进新材料产业扶持相关政策的省份有10个：上海市（id=75）、河南省（id=87）、内蒙古自治区（id=101）、云南省（id=128、152）、湖南省（id=194、492）、山东省（id=284）、广西壮族自治区（id=273）、贵州省（id=276）、四川省（id=526、387）、新疆维吾尔自治区（id=556）。",
+    "从regional_industry_status.csv筛选行业=\"金属冶炼和压延加工业\"的记录，剔除港澳台，进一步筛选净利润金额合计数据完整且企业总数大于0的省份，共得到14个有效省份：广东省、北京市、江苏省、上海市、浙江省、山东省、四川省、安徽省、湖南省、河南省、河北省、辽宁省、吉林省、新疆维吾尔自治区、山西省。注：云南省和贵州省虽有专项政策但净利润数据缺失，不纳入计算。有效数据范围内的政策省份为四川省和河南省，无政策省份为其余13个省份。",
+    "读取2022年各省净利润金额合计和净利润同比增减幅中位数。2022年有政策/无政策比值 = 1108.93 / 610.96 = 1.82。",
+    "按情景假设计算各省2025年预计净利润（3年复合增长）：有政策省份保持原增速，四川省预计净利润 = 650.92 × (1 + 46.61%)³ = 2051.25亿元；河南省预计净利润 = 91.79 × (1 + (-8.17%))³ = 71.08亿元；湖南省预计净利润 = 76.12 × (1 + （-56.83%)）³ = 61.23亿元；上海市预计净利润 = 142.58 × (1 + （-31.27）%)³ = 46.29亿元；山东省预计净利润 = 14.75 × (1 + 9.10%)³ = 19.16亿元；四川省预计净利润 = 64.61 × (1 + 46.61%)³ = 205.13亿元；；有政策省份2025年合计 = 2366.32亿元。",
+    "无政策省份各省增速减半后计算2025年预计净利润，13省合计 = 654.81亿元。",
+    "计算2025年比值 = 2366.32 / 654.81 = 3.61；比值变化量 = 3.61 - 1.82 = 1.79，比值提升1.51。"
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "有效数据省份总数（个）": 14,
+    "有效数据范围内政策省份数（个）": 5,
+    "有政策省份2022年净利润合计（亿元）": 1108.93,
+    "无政策省份2022年净利润合计（亿元）": 610.96,
+    "2022年比值（有政策/无政策）": 1.82,
+    "有政策省份2025年净利润预计合计（亿元）": 2366.32,
+    "无政策省份2025年净利润预计合计（亿元）": 654.81,
+    "2025年比值（有政策/无政策）": 3.61,
+    "比值变化量": 1.79
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/hard014.json b/assets/qa_gold/industry_planning/hard014.json
new file mode 100644
index 0000000000000000000000000000000000000000..108a8d377c2ac673035677edccb201781ba07f13
--- /dev/null
+++ b/assets/qa_gold/industry_planning/hard014.json
@@ -0,0 +1,32 @@
+{
+  "id": "hard014",
+  "question": "2022年，在中国大陆消费电子及电气业中，假设出台了电子信息产业集群培育专项政策的省份（仅统计消费电子及电气业企业不少于5家的中国大陆省份，不含港澳台），其企业净利润增速在未来3年额外提升8个百分点，而未出台此类专项政策的省份净利润增速衰减至当前水平的50%（净利润增速取各省份企业净利润同比增速的中位数；增长方式为3年复合增长），到2025年，在2022年净利润总额排名处于后半段（排名靠后一半）的有政策省份中，排名提升幅度最大的是哪个省份？其预计净利润总额是多少亿元？",
+  "guidelines": "依次回答省份名称和预计净利润总额。净利润总额保留2位小数。如[\"江西省\", 45.82]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "四川省",
+    63.45
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选policyClassification为'地方政策'、industry字段包含'消费电子'或'通信传输'或'半导体'的政策记录，得到54条涉及电子类行业的地方政策。",
+    "从policy_resource.csv中读取上述54条地方政策全文，分析涉及电子信息产业集群培育专项地方政策的省份有4个：安徽省（id=301、609）、广东省（id=23、153、249、605）、四川省（id=387）、浙江省（id=125）。",
+    "从company_profile.csv筛选industry='消费电子及电气业'且province不含港澳台的企业，共344家。统计各省企业数，筛选企业数不少于5家的省份，得到12个有效省份：广东省、山东省、浙江省、北京市、湖南省、江苏省、四川省、湖北省、安徽省、江西省、上海市、福建省。",
+    "从company_operation_status.csv提取这12个省份消费电子及电气业企业的净利润金额和净利润同比增减幅数据。计算各省净利润合计及增速中位数，按2022年净利润合计降序排名：第1广东省（1223.93亿），第2山东省（364.09亿），第3浙江省（341.43亿），第4北京市（160.74亿），第5湖南省（47.10亿），第6江苏省（46.71亿），第7四川省（31.29亿），第8湖北省（19.13亿），第9安徽省（15.80亿），第10江西省（4.83亿），第11上海市（1.09亿），第12福建省（-10.34亿）。",
+    "确定后半段（排名第7至第12名）且有政策的省份：四川省（第7名，增速中位数18.57%）和江西省（第10名，增速中位数-25.48%）。按政策情景计算：四川省调整后增速=18.57%+8%=26.57%，2025年预测净利润=31.29×(1+(18.57 + 8)/100)³=63.45亿元，2025排名升至第5名（提升2位）。"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "涉及电子类行业的地方政策总数（条）": 54,
+    "认定有政策的省份数（个）": 3,
+    "符合条件的省份总数（个）": 12,
+    "四川省2022净利润合计（亿元）": 31.29,
+    "四川省净利润增速中位数（%）": 18.57,
+    "四川省政策加持后增速（%）": 26.57,
+    "四川省2025净利润预测（亿元）": 63.45,
+    "四川省2025排名提升幅度（位）": 2
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/medium001.json b/assets/qa_gold/industry_planning/medium001.json
new file mode 100644
index 0000000000000000000000000000000000000000..2695044e73a50ddbfd9ed18243c5ec510859c721
--- /dev/null
+++ b/assets/qa_gold/industry_planning/medium001.json
@@ -0,0 +1,33 @@
+{
+  "id": "medium001",
+  "question": "Based on 2022 data, the following policy effect scenario is simulated: For provinces that have promulgated industrial policies containing the keywords \"semiconductor\" or \"integrated circuit\", policy empowerment accelerates the R&D investment expansion pace of their semiconductor enterprises to 2 times the current growth rate over the next 3 years; for provinces that have not yet issued such policies, R&D growth rate remains unchanged. Using the median year-on-year change in enterprise R&D investment as the baseline growth rate for each province, and projecting with 3-year compound growth, which province will have the highest total semiconductor industry R&D investment nationwide by 2025? What is the corresponding estimated amount?",
+  "guidelines": "Answer format: [province name, value (2 decimal places, unit: yuan)]. If relevant data cannot be found, please answer \"No relevant data found\".",
+  "answer": [
+    "Shanghai",
+    97732260069.03
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "Filter from policy_release_status.csv policy records whose name contains \"semiconductor\" or \"integrated circuit\", obtaining 4 records. Extract the list of involved provinces (deduplicated, excluding national policies), obtaining 4 provinces: Guangdong, Zhejiang, Shanghai, Anhui.",
+    "Filter from company_profile.csv all enterprise records with industry=\"semiconductor industry\", extract enterprise name, bmCode, and province fields, finding 172 enterprises.",
+    "From company_operation_status.csv, link the above enterprises by bmCode, extract R&D investment amount and year-on-year R&D investment change rate fields, filter out enterprises with either field empty, obtaining 168 valid enterprises.",
+    "Group by province, calculate total R&D investment amount and median year-on-year R&D investment change rate for each province, totaling 22 provinces.",
+    "For each province, determine whether it belongs to provinces with policy: adjusted growth rate for provinces with policy = median growth rate × 2; adjusted growth rate for provinces without policy = median growth rate. The top-ranked province Shanghai has policy support, with adjusted growth rate of 64.38%.",
+    "Calculate estimated total R&D investment for 2025 for each province = total R&D investment amount × (1 + adjusted growth rate/100)^3. Shanghai's estimated 2025 R&D investment = 22003461800.09 × (1+64.38/100)^3 = 97732260069.03 (yuan).",
+    "Sort all provinces by estimated total R&D investment for 2025 in descending order.",
+    "The top-ranked province is Shanghai, with estimated total R&D investment for 2025 of 97732260069.03 yuan."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "Provinces with policy": 4,
+    "Valid semiconductor industry enterprises": 168,
+    "Provinces involved": 22,
+    "Shanghai total R&D investment 2022 (yuan)": 22003461800.09,
+    "Shanghai adjusted growth rate (%)": 64.38,
+    "Shanghai estimated R&D investment 2025 (yuan)": 97732260069.03
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/medium002.json b/assets/qa_gold/industry_planning/medium002.json
new file mode 100644
index 0000000000000000000000000000000000000000..e35838e5382d60bfc465fd9f143aa182e1b94af4
--- /dev/null
+++ b/assets/qa_gold/industry_planning/medium002.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium002",
+  "question": "In 2022, assume that in the pharmaceutical manufacturing industry, the annual operating revenue growth rate of private enterprises is 5 percentage points higher than state-owned enterprises (including central and local state-owned enterprises) in the same province, while state-owned enterprises maintain their current growth rate unchanged (growth rate measured by the median year-on-year change in operating revenue of enterprises in the same province). By 2025, how many provinces will have private enterprise total revenue exceeding state-owned enterprise total revenue for the first time?",
+  "guidelines": "Answer format: integer (unit: count). If relevant data cannot be found, please answer \"No relevant data found\".",
+  "answer": 1,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "Filter from company_profile.csv all enterprise records with industry=\"pharmaceutical manufacturing industry\", extract enterprise name, bmCode, province, and ownership fields, finding 449 enterprises.",
+    "Divide enterprises into two groups by ownership: private enterprise group (ownership=\"private enterprise\") 346 enterprises, state-owned enterprise group (ownership=\"central state-owned enterprise\" or \"local state-owned enterprise\") 65 enterprises.",
+    "From company_operation_status.csv, link the above enterprises by bmCode, extract operating revenue amount and year-on-year operating revenue change rate fields, filter out enterprises with either field empty. Private enterprises: 331 valid records; state-owned enterprises: 64 valid records.",
+    "Group by province, calculate total operating revenue amount and median year-on-year operating revenue change rate for private and state-owned enterprises in each province respectively. Private enterprises cover 30 provinces, state-owned enterprises cover 24 provinces.",
+    "Filter provinces that have both private and state-owned enterprise data, totaling 24. Calculate adjusted growth rates: private enterprise growth = private median growth + 5; state-owned enterprise growth = state-owned median growth (unchanged).",
+    "For each valid province, project 2025 revenue: private 2025 revenue = private 2022 total × (1 + private adjusted growth/100)^3; state-owned 2025 revenue = state-owned 2022 total × (1 + state-owned growth/100)^3.",
+    "Filter provinces meeting the condition: private revenue lower than state-owned revenue in 2022, but estimated private revenue higher than state-owned revenue in 2025. Provinces meeting the condition: Heilongjiang.",
+    "Count provinces meeting the condition: 1."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "Pharmaceutical manufacturing private enterprises": 346,
+    "Pharmaceutical manufacturing state-owned enterprises": 65,
+    "Provinces with both enterprise types": 24,
+    "Provinces where private exceeded state-owned in 2025 after being lower in 2022": 1,
+    "Provinces meeting the condition": [
+      "Heilongjiang Province"
+    ]
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/medium003.json b/assets/qa_gold/industry_planning/medium003.json
new file mode 100644
index 0000000000000000000000000000000000000000..766f8633bbbb3e9475f877b003384e1eb678c3f8
--- /dev/null
+++ b/assets/qa_gold/industry_planning/medium003.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium003",
+  "question": "Using 2022 as the base period, a differentiated policy incentive is proposed for the automobile manufacturing industry: For provinces that have implemented automobile industry policies containing the keywords \"new energy\" or \"electric\", the annual operating profit growth rate of automobile manufacturing enterprises within their jurisdiction will add 10 percentage points on top of the current median growth rate, creating a policy acceleration effect; other provinces are unaffected, and operating profit growth rate continues along the current trajectory. Under this differentiated scenario, using the median year-on-year change in operating profit as the baseline growth rate for each province, and projecting to 2025 via 3-year compound growth, which province has the most prominent increase in total operating profit compared to actual 2022 levels? What is the specific increase (increase = (2025 estimated value - 2022 actual value) / 2022 actual value × 100%)?",
+  "guidelines": "Answer format: value (2 decimal places, unit: %). If relevant data cannot be found, please answer \"No relevant data found\".",
+  "answer": 859.71,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "Filter from policy_release_status.csv policy records where industry field contains \"automobile\" and policy name contains \"new energy\" or \"electric\", obtaining 10 records. Extract the list of involved provinces (deduplicated, excluding national policies), obtaining 7 provinces: Hainan, Chongqing, Guangdong, Jiangxi, Sichuan, Shandong, Jiangsu.",
+    "Filter from company_profile.csv all enterprise records with industry=\"automobile manufacturing industry\", extract enterprise name, bmCode, and province fields, finding 230 enterprises.",
+    "From company_operation_status.csv, link the above enterprises by bmCode, extract operating profit amount and year-on-year operating profit change rate fields, filter out enterprises with either field empty, obtaining 230 valid enterprises.",
+    "Group by province, calculate total operating profit amount and median year-on-year operating profit change rate for each province, filter provinces with total operating profit amount greater than 0, totaling 22 provinces.",
+    "For each valid province, determine whether it belongs to provinces with policy: adjusted growth rate for provinces with policy = median growth rate + 10; adjusted growth rate for provinces without policy = median growth rate.",
+    "Calculate estimated total operating profit for 2025 for each province = total operating profit amount × (1 + adjusted growth rate/100)^3. Guangxi Zhuang Autonomous Region's estimated 2025 operating profit = 117492412.97 × (1+112.51/100)^3 = 1127581487.29 (yuan).",
+    "Calculate growth rate for each province = (2025 estimated value - 2022 actual value) / 2022 actual value × 100. Guangxi Zhuang Autonomous Region's growth rate = (1127581487.29 - 117492412.97) / 117492412.97 × 100 = 859.71%.",
+    "Sort by growth rate in descending order. The province with the largest growth rate is Guangxi Zhuang Autonomous Region, at 859.71%."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "Provinces with policy": 7,
+    "Valid automobile manufacturing enterprises": 230,
+    "Provinces with total operating profit > 0": 22,
+    "Guangxi total operating profit 2022 (yuan)": 117492412.97,
+    "Guangxi adjusted growth rate (%)": 112.51,
+    "Guangxi estimated operating profit 2025 (yuan)": 1127581487.29,
+    "Guangxi growth rate (%)": 859.71
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/medium004.json b/assets/qa_gold/industry_planning/medium004.json
new file mode 100644
index 0000000000000000000000000000000000000000..705ccaf73ad112e858151af92ea4ddd9cc75bfa0
--- /dev/null
+++ b/assets/qa_gold/industry_planning/medium004.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium004",
+  "question": "In 2022, in the communication transmission equipment industry, assuming that enterprises with R&D investment ratio below the national industry median will be eliminated from the market within 3 years, and only enterprises with R&D investment ratio not lower than the national median will be retained, what is the proportion for the province with the highest ratio of remaining enterprises' operating revenue after elimination to operating revenue before elimination?",
+  "guidelines": "Answer format: Value (2 decimal places, unit: %). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 100.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "Filter from national_industry_status.csv where industry=\"communication transmission equipment industry\", extract the R&D investment ratio median of 9.92 as the national benchmark.",
+    "Filter from company_profile.csv all enterprise records with industry=\"communication transmission equipment industry\", extract enterprise name, bmCode, and province fields, finding 120 enterprises.",
+    "From company_operation_status.csv, link the above enterprises by bmCode, extract operating revenue amount and R&D investment ratio fields, linking 120 enterprises.",
+    "Group by province, calculate total operating revenue amount for each province before elimination (all enterprises, including those with empty R&D investment ratio), totaling 19 provinces.",
+    "Filter enterprises with non-empty R&D investment ratio and not lower than the national median of 9.92 (enterprises with empty R&D investment ratio are deemed below median and automatically eliminated), group by province to calculate total operating revenue of remaining enterprises after elimination. 13 provinces have surviving enterprises after elimination.",
+    "For each valid province, calculate revenue retention ratio = total revenue after elimination / total revenue before elimination × 100. Anhui Province's revenue retention ratio = 2727186878.01 / 2727186878.01 × 100 = 100.00%.",
+    "Sort by revenue retention ratio in descending order. Anhui Province ranks highest with a ratio of 100.00%."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "National R&D investment ratio median": 9.92,
+    "Total communication transmission equipment industry enterprises": 120,
+    "Anhui Province total revenue before elimination (yuan)": 2727186878.01,
+    "Anhui Province total revenue after elimination (yuan)": 2727186878.01,
+    "Anhui Province revenue retention ratio (%)": 100.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/medium005.json b/assets/qa_gold/industry_planning/medium005.json
new file mode 100644
index 0000000000000000000000000000000000000000..54b1d4e2b12aeba574f95f95c33e3ac2f9cb04c9
--- /dev/null
+++ b/assets/qa_gold/industry_planning/medium005.json
@@ -0,0 +1,32 @@
+{
+  "id": "medium005",
+  "question": "In 2022, regarding the strategic choice for Guangdong Province's consumer electronics and electrical industry, a policy consulting agency proposed two competing development paths: the first is the \"high-end transformation route\", evaluated by the average R&D investment ratio of private enterprises, invention patent density (= total annual Chinese invention patent grants ÷ total number of enterprises), and the number of relevant industrial policies in that province; the second is the \"export-oriented route\", evaluated by per capita revenue (= total operating revenue ÷ total number of employees), average asset turnover rate (= mean operating revenue ÷ mean total assets), and total number of enterprises. Both routes use inter-provincial peer comparison ranking scores (score = (N - ranking) / (N - 1) × 100), with equal weight across dimensions to calculate the route total score. What is the difference between Guangdong Province's total score on the \"high-end transformation route\" and the \"export-oriented route\" (former minus latter)?",
+  "guidelines": "Answer format: Value (2 decimal places). A positive number indicates the high-end transformation route has a higher score, a negative number indicates the export-oriented route has a higher score. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 4.55,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "From company_profile.csv, filter industry=\"consumer electronics and electrical industry\", join with company_operation_status.csv for year=2022 on bmCode; keep only provinces with enterprise count > 0 to obtain N=23 provinces for peer comparison within the same industry.",
+    "Within each province, filter ownership=\"private enterprise\" and take the arithmetic mean of non-missing R&D investment ratios; Guangdong Province is approximately 9.2610.",
+    "From regional_industry_status.csv, filter industry=\"consumer electronics and electrical industry\", and use for each province: total annual Chinese invention patent grants ÷ enterprise count, total operating revenue ÷ total employees, mean operating revenue ÷ mean total assets, and enterprise count; Guangdong: invention patent density 96.4800, per capita revenue about 1501516.49, asset turnover about 0.8845, enterprise count 150.",
+    "From policy_release_status.csv, filter policyClassification=\"local policy\", publishDate in calendar year 2022, and industry field containing the full industry name \"consumer electronics and electrical industry\"; records with empty province are excluded from province-level counts.",
+    "For all six indicators, rank provinces in descending order among N=23 (higher is better), use minimum rank for ties, and compute score = (N - ranking) / (N - 1) × 100 for each province. Guangdong: private R&D ratio about rank 5 → about 81.82; patent density about rank 3 → about 90.91; policy count about rank 2 → about 95.45; per capita revenue about rank 6 → about 77.27; asset turnover about rank 6 → about 77.27; enterprise count rank 1 → 100.00.",
+    "High-end transformation score = (81.82 + 90.91 + 95.45) / 3 ≈ 89.39; export-oriented score = (77.27 + 77.27 + 100.00) / 3 ≈ 84.85.",
+    "Score difference = high-end transformation score - export-oriented score ≈ 4.55 (rounded to two decimal places)."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Private enterprise average R&D investment ratio score": 81.82,
+    "Invention patent density score": 90.91,
+    "Policy count score": 95.45,
+    "Per capita revenue score": 77.27,
+    "Asset turnover rate score": 77.27,
+    "Enterprise count score": 100.0,
+    "High-end transformation score": 89.39,
+    "Export-oriented score": 84.85,
+    "Score difference": 4.55
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/medium006.json b/assets/qa_gold/industry_planning/medium006.json
new file mode 100644
index 0000000000000000000000000000000000000000..d44f5aede93533e52df8eb47ca61b8782f9de84a
--- /dev/null
+++ b/assets/qa_gold/industry_planning/medium006.json
@@ -0,0 +1,33 @@
+{
+  "id": "medium006",
+  "question": "In 2022, regarding the industrial development direction of Hebei Province's metal smelting and rolling processing industry, researchers intend to compare two alternative transformation paths through a multi-dimensional scoring method. Path one \"green and low-carbon route\" covers three evaluation indicators: average enterprise R&D investment amount (reflecting technology upgrade willingness), invention patent density (= total cumulative Chinese invention patent grants ÷ total number of enterprises, weight 0.3), and count of provincial green-related policies (i.e., policy records whose name contains \"green\", \"low-carbon\", or \"energy-saving\", weight 0.4), with average R&D investment amount weight 0.3; Path two \"traditional capacity expansion route\" also includes three indicators: total assets (weight 0.4), total operating revenue (weight 0.3), and total number of enterprises (weight 0.3). Each province's score for each indicator is calculated by inter-provincial peer ranking (score = (N - ranking) / (N - 1) × 100). Please calculate the score difference between Hebei Province on the above two routes (green and low-carbon route score minus traditional capacity expansion route score).",
+  "guidelines": "Answer format: Value (2 decimal places). A positive number indicates the green and low-carbon route has a higher score, a negative number indicates the traditional capacity expansion route has a higher score. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 42.97,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter industry = \"金属冶炼和压延加工业\" (metal smelting and rolling processing industry), and extract for each province: total number of enterprises, mean R&D investment amount (研发投入金额均值), total cumulative Chinese invention patent grants, total assets, and total operating revenue. After dropping nulls for mean R&D, patent density, total assets, and operating revenue, there are 16 valid provinces for each of these four indicators.",
+    "Average enterprise R&D investment uses the table field mean R&D investment amount (non-null values participate in ranking). Hebei Province: 3062942709.59 yuan, rank 2 among 16 provinces (ties take the best rank), score = (16 - 2) / (16 - 1) × 100 = 93.33.",
+    "Invention patent density = total cumulative Chinese invention patent grants ÷ total number of enterprises (enterprises > 0 and both fields non-null). Hebei Province: 793 ÷ 1 = 793.0000, rank 2 among 16, score = 93.33.",
+    "From policy_release_status.csv, filter records with publishDate in calendar year 2022 (parsed as day/month/year), policyClassification = \"地方政策\", province non-empty and not \"全国\", industry field containing \"金属冶炼和压延加工业\", and policy name containing at least one of \"绿色\", \"低碳\", or \"节能\"; count by province. Eight records matched, across six provinces; Hebei Province: 0 records.",
+    "Provincial green-related policy count: count on the 34 provinces that appear in the industry table (provinces with no match are 0), rank within this indicator; Hebei has 0, tied with 27 other provinces with competitive rank 7, score = (34 - 7) / (34 - 1) × 100 = 81.82.",
+    "Traditional capacity expansion: total assets and operating revenue are ranked among the 16 provinces with non-null values; Hebei ranks 6th and 11th, scores 66.67 and 33.33 respectively. Total number of enterprises is ranked among all 34 provinces; Hebei has 1 enterprise, rank 24 (tied with six provinces), score = 30.30.",
+    "Green and low-carbon route score = 93.33 × 0.3 + 93.33 × 0.3 + 81.82 × 0.4 = 88.73; traditional capacity expansion route score = 66.67 × 0.4 + 33.33 × 0.3 + 30.30 × 0.3 = 45.76.",
+    "Score difference (green and low-carbon − traditional capacity expansion) = 88.73 − 45.76 = 42.97."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "Average enterprise R&D investment score": 93.33,
+    "Invention patent density score": 93.33,
+    "Green policy count score": 81.82,
+    "Total assets score": 66.67,
+    "Operating revenue score": 33.33,
+    "Enterprise count score": 30.3,
+    "Green and low-carbon route score": 88.73,
+    "Traditional capacity expansion route score": 45.76,
+    "Score difference": 42.97
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/medium007.json b/assets/qa_gold/industry_planning/medium007.json
new file mode 100644
index 0000000000000000000000000000000000000000..4d44ada2852a77f16501815c4a110812c1bba954
--- /dev/null
+++ b/assets/qa_gold/industry_planning/medium007.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium007",
+  "question": "In the 2022 data for the chemical raw materials and chemical products manufacturing industry, first identify the enterprise with the largest total assets in the entire industry and determine its registration province; then take that province as the research object, and use the four-indicator equal-weight scoring method (each indicator score = (N - inter-provincial ranking) / (N - 1) × 100) to compare two industrial strategy routes: the \"R&D-driven route\" comprehensively evaluates four indicators: total private enterprise R&D investment, total state-owned enterprise R&D investment (central state-owned + local state-owned + other state-owned enterprises + state-owned enterprises (research institutes)), regional R&D intensity (mean R&D investment ratio), and count of relevant R&D policies (policy name contains \"R&D\", \"innovation\", or \"technology/science\"); the \"scale expansion route\" comprehensively evaluates four indicators: total assets, total operating revenue, total number of enterprises, and total government subsidies. Which route has the higher comprehensive score for that province?",
+  "guidelines": "Answer format: \"R&D-driven route\" or \"Scale expansion route\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "Scale expansion route",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "Filter from company_profile.csv enterprises with industry=\"chemical raw materials and chemical products manufacturing industry\", extract enterprise name, bmCode, province, and ownership fields, finding 364 enterprises.",
+    "From company_operation_status.csv, link enterprises by bmCode and restrict year=2022, extract the total assets field, filter out empty values and sort in descending order. The enterprise with the highest total assets is Hengyi Changhua Fine Chemical Company (total assets 200843104094.19 yuan), located in Shandong Province.",
+    "From company_operation_status.csv, extract the R&D investment amount field for enterprises in this industry, filter out empty values, then aggregate by province and ownership; Shandong Province private enterprise total R&D investment is 6239314301.75 yuan, and state-owned enterprises (central state-owned + local state-owned + other state-owned enterprises + state-owned enterprises (research institutes)) total R&D investment is 6670389432.99 yuan.",
+    "Filter from regional_industry_status.csv where industry=\"chemical raw materials and chemical products manufacturing industry\" and province is non-empty and not \"national aggregate\", yielding 34 province rows; extract mean R&D investment ratio, total assets, total operating revenue, enterprise count, and total government reward and subsidy per province. For each of the eight indicators, ranking uses only provinces with non-missing values for that field; effective province count N is computed separately per indicator (e.g. R&D intensity N=16, enterprise count N=34).",
+    "Filter from policy_release_status.csv records whose policy name contains \"R&D\" or \"innovation\" or \"technology/science\" and whose province is non-empty and not \"national aggregate\", group by province to count; Shandong has 7 records (93 policy records in the full database satisfying this rule).",
+    "Rank each of the eight indicators in descending order by value (ties share the same rank and subsequent ranks skip accordingly), and compute Shandong's score per indicator = (N - rank) / (N - 1) × 100 (if N < 2, that indicator score is 100). Private R&D score = 96.30, state-owned R&D score = 100.00, R&D intensity score = 40.00, policy count score = 91.67, total assets score = 100.00, operating revenue score = 100.00, enterprise count score = 93.94, total subsidy score = 93.33.",
+    "R&D-driven score = (96.30 + 100.00 + 40.00 + 91.67) × 0.25 = 81.99; scale expansion score = (100.00 + 100.00 + 93.94 + 93.33) × 0.25 = 96.82.",
+    "R&D-driven score 81.99 < scale expansion score 96.82, therefore the scale expansion route has the higher score."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "Enterprise with highest total assets": "Hengyi Changhua Fine Chemical Company",
+    "Target province": "Shandong Province",
+    "Private R&D investment score": 96.3,
+    "State-owned R&D investment score": 100.0,
+    "R&D intensity score": 40.0,
+    "R&D policy score": 91.67,
+    "Total assets score": 100.0,
+    "Operating revenue score": 100.0,
+    "Enterprise count score": 93.94,
+    "Total subsidy score": 93.33,
+    "R&D-driven route score": 81.99,
+    "Scale expansion route score": 96.82
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/medium008.json b/assets/qa_gold/industry_planning/medium008.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0c80ffe3e2b40291046d8ae13fb1ee81eb92765
--- /dev/null
+++ b/assets/qa_gold/industry_planning/medium008.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium008",
+  "question": "In 2022, in the food and beverage industry, for the province where the enterprise with the most cumulative Chinese invention patent grants is located, if that province chooses the \"brand upgrade route\" (evaluating market-cap-to-revenue ratio, profit margin, per capita market cap, with weights of 35%, 35%, 30% respectively) versus the \"industrial chain extension route\" (evaluating total number of enterprises, revenue scale, upstream-downstream enterprise diversity, with weights of 40%, 30%, 30% respectively), which route has the higher score?",
+  "guidelines": "Answer format: \"Brand upgrade route\" or \"Industrial chain extension route\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "Industrial chain extension route",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "Step 1: Filter from company_profile.csv enterprises with industry=\"food and beverage industry\", extract enterprise name, bmCode, and province fields, finding 247 enterprises.",
+    "Step 2: From company_operation_status.csv, link enterprises by bmCode, extract cumulative Chinese invention patent grants field, filter out empty values, sort in descending order. The enterprise with the most cumulative Chinese invention patent grants is Qingqing Jinyin Food Company (644 grants), located in Beijing.",
+    "Step 3: Filter from regional_industry_status.csv where industry=\"food and beverage industry\", extract total company market cap, total operating revenue, total operating profit, total number of employees, and total number of enterprises for each province, totaling 30 provinces.",
+    "Step 4: Calculate each province's market-cap-to-revenue ratio = total company market cap / total operating revenue, Beijing: 0.0000; profit margin = total operating profit / total operating revenue, Beijing: 0.0748; per capita market cap = total company market cap / total number of employees, Beijing: 0.02.",
+    "Step 5: From company_profile.csv, count the number of ownership types (distinct count) for food and beverage industry enterprises by province. Beijing: 4 types.",
+    "Step 6: Sort the six indicators in descending order respectively, calculate Beijing's ranking score for each indicator = (N - ranking) / (N - 1) × 100. Market-cap-to-revenue ratio score = 26.67, profit margin score = 33.33, per capita market cap score = 13.33, enterprise count score = 75.86, revenue scale score = 82.76, diversity score = 93.10.",
+    "Step 7: Calculate brand upgrade score = 26.67 × 0.35 + 33.33 × 0.35 + 13.33 × 0.3 = 25.00; industrial chain extension score = 75.86 × 0.4 + 82.76 × 0.3 + 93.10 × 0.3 = 83.10.",
+    "Step 8: Brand upgrade score 25.00 < industrial chain extension score 83.10, therefore the industrial chain extension route has the higher score."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "Enterprise with most patents": "Qingqing Jinyin Food Company",
+    "Target province": "Beijing",
+    "Market-cap-to-revenue ratio score": 26.67,
+    "Profit margin score": 33.33,
+    "Per capita market cap score": 13.33,
+    "Enterprise count score": 75.86,
+    "Revenue scale score": 82.76,
+    "Diversity score": 93.1,
+    "Brand upgrade route score": 25.0,
+    "Industrial chain extension route score": 83.1
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/medium009.json b/assets/qa_gold/industry_planning/medium009.json
new file mode 100644
index 0000000000000000000000000000000000000000..60dfe0713c8adc460fc901b2876257990ff5ae63
--- /dev/null
+++ b/assets/qa_gold/industry_planning/medium009.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium009",
+  "question": "Among the enterprises in the textile, footwear and apparel industry in 2022, find the enterprise with the largest R&D personnel scale; its province is the analysis object. For that province's textile, footwear and apparel industry, use the following two route scoring systems to determine which development route has the advantage—the \"automation upgrade route\" scores by three indicators: R&D investment intensity (= total R&D investment amount / total operating revenue, weight 0.4), capitalized R&D investment ratio (= total capitalized R&D investment / total R&D investment amount, weight 0.3), and count of equipment manufacturing policies (policy name contains \"equipment\" or \"intelligent manufacturing\", weight 0.3); the \"brand overseas expansion route\" scores by cumulative PCT patent applications (weight 0.4), per capita revenue (weight 0.3), and count of export-related policies (policy name contains \"export\", \"foreign trade\", or \"international\", weight 0.3); both routes use inter-provincial ranking scores for each indicator (score = (N - ranking) / (N - 1) × 100). Which route has the higher score for that province?",
+  "guidelines": "Answer format: \"Automation upgrade route\" or \"Brand overseas expansion route\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "Automation upgrade route",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "Filter from company_profile.csv enterprises with industry=\"textile, footwear and apparel industry\", extract enterprise name, bmCode, and province fields, finding 177 enterprises.",
+    "From company_operation_status.csv, link enterprises by bmCode, filter year=2022, extract R&D personnel count, filter out empty values, sort in descending order. The enterprise with the most R&D personnel is Anbu Shangchang Brand Company (2,826 people), located in Guangdong Province.",
+    "Filter from regional_industry_status.csv where industry=\"textile, footwear and apparel industry\", extract for each province total R&D investment amount, total operating revenue, total capitalized R&D investment, total cumulative PCT patent applications, and total number of employees, for 34 provincial-level regions.",
+    "Calculate each province's R&D investment intensity = total R&D investment amount / total operating revenue, Guangdong: 0.017511; capitalized R&D investment ratio = total capitalized R&D investment / total R&D investment amount, Guangdong: 0.004225; per capita revenue = total operating revenue / total number of employees, Guangdong: 269886.81. (Provinces with null values or a zero denominator are excluded from ranking for that indicator.)",
+    "From policy_release_status.csv, count by policy name: equipment manufacturing policies whose name contains \"equipment\" or \"intelligent manufacturing\"; export-related policies whose name contains \"export\", \"foreign trade\", or \"international\". Policies with province=\"全国\" count once for each of the 34 provincial samples; other policies are assigned to the corresponding province.",
+    "For each of the six indicators, rank provinces in descending order within the valid set for that indicator, with ties taking the best rank; Guangdong's ranking score = (N - ranking) / (N - 1) × 100. R&D intensity score = 30.77, capitalized R&D ratio score = 91.67, equipment policy score = 90.91, PCT patent score = 86.67, per capita revenue score = 0.00, export policy score = 90.91.",
+    "Calculate automation upgrade score = 30.77 × 0.4 + 91.67 × 0.3 + 90.91 × 0.3 = 67.08; brand overseas expansion score = 86.67 × 0.4 + 0.00 × 0.3 + 90.91 × 0.3 = 61.94.",
+    "Automation upgrade score 67.08 > brand overseas expansion score 61.94, therefore the automation upgrade route has the higher score."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "Enterprise with most R&D personnel": "Anbu Shangchang Brand Company",
+    "Target province": "Guangdong Province",
+    "R&D investment intensity score": 30.77,
+    "Capitalized R&D ratio score": 91.67,
+    "Equipment policy score": 90.91,
+    "PCT patent score": 86.67,
+    "Per capita revenue score": 0.0,
+    "Export policy score": 90.91,
+    "Automation upgrade route score": 67.08,
+    "Brand overseas expansion route score": 61.94
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/medium010.json b/assets/qa_gold/industry_planning/medium010.json
new file mode 100644
index 0000000000000000000000000000000000000000..660d70403712281bef9c333fac404e04ad11c9d1
--- /dev/null
+++ b/assets/qa_gold/industry_planning/medium010.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium010",
+  "question": "Sort the semiconductor industry in 2022 by operating revenue from high to low, identify the top 10 enterprises by revenue scale, and count the provinces to which these leading enterprises belong; the province with the highest frequency is the research target. Next, conduct a national horizontal comparison of that province based on four industrial competitiveness dimensions—dimension one is enterprise agglomeration (total number of semiconductor industry enterprises per province), dimension two is innovation activity (= total annual Chinese invention patent grants ÷ total number of enterprises), dimension three is policy support (= count of relevant policies whose name contains \"semiconductor\" or \"integrated circuit\"), dimension four is industry scale (= total operating revenue). Among the national provincial rankings for each dimension, in how many dimensions did that province rank in the top 3?",
+  "guidelines": "Answer format: Integer. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 3,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "Filter from company_profile.csv enterprises with industry=\"semiconductor industry\", extract enterprise name, bmCode, and province fields, finding 172 enterprises.",
+    "From company_operation_status.csv, link enterprises by bmCode, extract operating revenue amount field; 172 enterprises after filtering empty values. Sort in descending order and take the top 10 enterprises.",
+    "Count provincial distribution of top 10 enterprises: Shanghai (4), Taiwan (3), Guangdong (2), Jiangsu (1). Shanghai has the highest frequency with 4 enterprises.",
+    "Filter from regional_industry_status.csv where industry=\"semiconductor industry\", extract total number of enterprises, total annual Chinese invention patent grants, and total operating revenue for each province, totaling 22 provinces.",
+    "Calculate each province's innovation activity = total annual Chinese invention patent grants / total number of enterprises (exclude provinces with 0 enterprises). Shanghai's innovation activity = 35.1852.",
+    "Filter from policy_release_status.csv policies whose name contains \"semiconductor\" or \"integrated circuit\", 4 records total. Group by province to count relevant policy count. Shanghai has 1 relevant policy.",
+    "Sort the four indicators in descending order respectively. Shanghai's rankings: enterprise agglomeration 2nd (27 enterprises), innovation activity 4th (35.1852), policy support 2nd (1 record), industry scale 2nd (247438863786.89 yuan).",
+    "Count the number of dimensions in which Shanghai ranks ≤ 3 among the four dimensions. 3 dimensions rank in the national top 3."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "Province with most top 10 enterprises by revenue": "Shanghai",
+    "Shanghai enterprise agglomeration (enterprise count)": 27,
+    "Shanghai enterprise agglomeration ranking": 2,
+    "Shanghai innovation activity": 35.1852,
+    "Shanghai innovation activity ranking": 4,
+    "Shanghai policy support (policy count)": 1,
+    "Shanghai policy support ranking": 2,
+    "Shanghai industry scale (total operating revenue)": 247438863786.89,
+    "Shanghai industry scale ranking": 2,
+    "Number of dimensions in top 3": 3
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/medium011.json b/assets/qa_gold/industry_planning/medium011.json
new file mode 100644
index 0000000000000000000000000000000000000000..e3af5d1343b947b1b947ea5dffdbcb31528c3203
--- /dev/null
+++ b/assets/qa_gold/industry_planning/medium011.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium011",
+  "question": "In the 2022 pharmaceutical manufacturing industry data, find the listed enterprise with the highest total Chinese patent grants in that year; after identifying that enterprise's province, construct a four-dimensional evaluation framework around \"innovation ecosystem\": R&D intensity (= total R&D investment amount ÷ total operating revenue), patent conversion efficiency (= total cumulative Chinese invention patent grants ÷ total cumulative Chinese invention patent applications), industry scale (= total operating revenue), and policy support (= count of relevant policies whose name contains \"pharmaceutical\" or \"biotechnology\"). For each dimension, identify the top 5 provinces nationwide and calculate their mean. Is that province's overall innovation ecosystem level above or below the average of the national top 5 across these four dimensions?",
+  "guidelines": "Answer format: \"Above average\" or \"Below average\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "Below average",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "Filter from company_profile.csv enterprises with industry=\"pharmaceutical manufacturing industry\", extract enterprise name, bmCode, and province fields, finding 449 enterprises.",
+    "From company_operation_status.csv, link enterprises by bmCode, extract annual Chinese patent grants field; after filtering empty values, sort in descending order. The enterprise with the most annual Chinese patent grants is Puge Ruijian Biopharmaceutical Company (337 grants), located in Henan Province.",
+    "Filter from regional_industry_status.csv where industry=\"pharmaceutical manufacturing industry\", extract total R&D investment amount, total operating revenue, total cumulative Chinese invention patent grants, and total cumulative Chinese invention patent applications for each province, totaling 34 provinces.",
+    "Calculate each province's R&D intensity = total R&D investment amount / total operating revenue, patent conversion efficiency = total cumulative Chinese invention patent grants / total cumulative Chinese invention patent applications (exclude provinces with zero denominator), 15 valid provinces.",
+    "Filter from policy_release_status.csv policies whose name contains \"pharmaceutical\" or \"biotechnology\", 21 records total. Group by province to count relevant policy count.",
+    "Sort the four indicators in descending order respectively, take the top 5 provinces for each, and calculate averages: R&D intensity top 5 average = 0.182465, patent conversion efficiency top 5 average = 0.522743, industry scale top 5 average = 237558091257.40, policy support top 5 average = 1.80.",
+    "Henan Province's values for the four indicators: R&D intensity = 0.075724, patent conversion efficiency = 0.311526, industry scale = 18878671803.01, policy support = 2. Compared with top 5 averages, 1 dimension is above average.",
+    "Comprehensive score 1 < 2, output \"Below average\"."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "Enterprise with most annual Chinese patent grants": "Puge Ruijian Biopharmaceutical Company",
+    "Province of that enterprise": "Henan Province",
+    "Annual Chinese patent grants": 337,
+    "Henan Province R&D intensity": 0.075724,
+    "R&D intensity top 5 average": 0.182465,
+    "Henan Province patent conversion efficiency": 0.311526,
+    "Patent conversion efficiency top 5 average": 0.522743,
+    "Henan Province industry scale": 18878671803.01,
+    "Industry scale top 5 average": 237558091257.4,
+    "Henan Province policy support": 2,
+    "Policy support top 5 average": 1.8,
+    "Number of dimensions above average": 1
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/medium012.json b/assets/qa_gold/industry_planning/medium012.json
new file mode 100644
index 0000000000000000000000000000000000000000..ee4c707996818cf9ab74462241b517099fac7f67
--- /dev/null
+++ b/assets/qa_gold/industry_planning/medium012.json
@@ -0,0 +1,33 @@
+{
+  "id": "medium012",
+  "question": "In 2022, in the railway, ship, aerospace and other transport equipment manufacturing industry, take the province with the highest concentration of the top 5 enterprises by total assets as the research object. What is that province's comprehensive ranking (ranking based on the arithmetic mean of the three dimension rankings) across three high-end manufacturing dimensions: technology intensity (technology intensity = total R&D personnel / total employees), capital intensity (capital intensity = total assets / total employees), and policy concentration (policy concentration = count of relevant policies)?",
+  "guidelines": "Answer format: Integer (unit: rank). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 4,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "Filter from company_profile.csv enterprises with industry=\"railway, ship, aerospace and other transport equipment manufacturing industry\", extract enterprise name, bmCode, and province fields, finding 99 enterprises.",
+    "From company_operation_status.csv, link enterprises by bmCode, extract total assets field; 99 enterprises after filtering empty values. Sort in descending order and take the top 5 enterprises.",
+    "Count provincial distribution of top 5 enterprises: Beijing (4), Shanghai (1). Beijing has the highest concentration with 4 enterprises.",
+    "Filter from regional_industry_status.csv where industry=\"railway, ship, aerospace and other transport equipment manufacturing industry\", extract total R&D personnel, total employees, and total assets for each province, totaling 34 provinces.",
+    "Calculate each province's technology intensity = total R&D personnel / total employees, capital intensity = total assets / total employees (exclude provinces with 0 employees), 14 valid provinces.",
+    "Filter from policy_release_status.csv policies whose involved industry contains \"aviation\" or \"aerospace\" or \"ship\", 46 records total. Group by province to count relevant policy count (excluding \"national\" level).",
+    "Sort the three indicators in descending order respectively. Beijing's rankings: technology intensity 4th, capital intensity 4th, policy concentration 9th. Average ranking = 5.67.",
+    "Sort all provinces by average ranking in ascending order. Beijing's comprehensive ranking is 4th."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "Province with most top 5 enterprises by total assets": "Beijing",
+    "Beijing technology intensity": 0.210283,
+    "Beijing technology intensity ranking": 4,
+    "Beijing capital intensity": 3488225.06,
+    "Beijing capital intensity ranking": 4,
+    "Beijing policy concentration (policy count)": 0,
+    "Beijing policy concentration ranking": 9,
+    "Beijing average ranking": 5.67,
+    "Comprehensive ranking": 4
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/medium013.json b/assets/qa_gold/industry_planning/medium013.json
new file mode 100644
index 0000000000000000000000000000000000000000..9932e53a9d47604703ca25e5709bc510e5c8d723
--- /dev/null
+++ b/assets/qa_gold/industry_planning/medium013.json
@@ -0,0 +1,41 @@
+{
+  "id": "medium013",
+  "question": "In the 2022 automotive manufacturing industry, take the top 10 provinces by profitability (measured by net profit) as the candidate set, and construct a three-dimensional comprehensive scoring model to identify the province with the best industrial development quality: Dimension one \"industrial chain completeness\" (weight 0.3) comprehensively assesses the number of enterprise ownership types (ownership diversity) and the interquartile range of total assets of enterprises within the province (scale diversity); dimension two \"technological capability\" (weight 0.4) comprehensively assesses mean R&D investment ratio (R&D intensity) and total annual Chinese invention patent grants divided by total number of enterprises (patent density); dimension three \"market performance\" (weight 0.3) comprehensively assesses total operating revenue (revenue scale) and total operating profit divided by total operating revenue (profit margin). Each dimension's score is represented by the average ranking of its sub-indicators among the candidate provinces (comprehensive score = weighted average of each dimension's mean ranking). Which province has the highest comprehensive score (i.e., the best overall performance)?",
+  "guidelines": "Answer format: Province name. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "Guangdong Province",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter records with industry=\"automotive manufacturing industry\"; for each province, after excluding empty values in the total net profit field, sort in descending order by that field and take the top 10 provinces industry-wide as the candidate set.",
+    "From company_profile.csv, filter enterprises with industry=\"automotive manufacturing industry\", extract bmCode and province; join company_operation_status.csv on bmCode with year=2022 and extract total assets; for each province, after excluding empty total assets, compute the interquartile range Q3-Q1 as scale diversity.",
+    "For candidate provinces, on the corresponding rows in regional_industry_status.csv compute ownership diversity: among the eight enterprise-count fields for Sino-foreign joint ventures, central state-owned, state-owned (other), state-owned (research institutes), local state-owned, foreign-funded, private, and collective, count how many types have a value greater than zero.",
+    "From regional_industry_status.csv candidate rows, extract mean R&D investment ratio (R&D intensity), total annual Chinese invention patent grants, total number of enterprises, total operating revenue, and total operating profit; patent density = total annual Chinese invention patent grants / total number of enterprises, profit margin = total operating profit / total operating revenue.",
+    "For the six sub-indicators (ownership diversity, total-assets IQR, R&D intensity, patent density, total operating revenue, profit margin), rank candidate provinces separately for each indicator using only provinces with valid values for that indicator, in descending order by value; tied values share the same rank and the next rank is skipped; larger values rank higher (better).",
+    "Industrial chain completeness dimension score = (ownership diversity rank + scale diversity rank) / 2, technological capability dimension score = (R&D intensity rank + patent density rank) / 2, market performance dimension score = (revenue scale rank + profit margin rank) / 2 (if only one sub-indicator is valid in a dimension, use only that rank in the average).",
+    "Comprehensive score = industrial chain completeness dimension score × 0.3 + technological capability dimension score × 0.4 + market performance dimension score × 0.3; this is a weighted average of mean ranks, so a smaller value indicates better overall performance.",
+    "Guangdong Province: industrial chain completeness dimension score 2.5, technological capability 2.0, market performance 4.0; comprehensive score = 0.3 × 2.5 + 0.4 × 2 + 0.3 × 4 = 2.75, the smallest among the 10 candidate provinces, hence the best overall performance (rank 1)."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "Number of candidate provinces": 10,
+    "Candidate province list": [
+      "Guangdong",
+      "Hebei",
+      "Shandong",
+      "Zhejiang",
+      "Beijing",
+      "Jiangsu",
+      "Shanghai",
+      "Jilin",
+      "Henan",
+      "Sichuan"
+    ],
+    "Guangdong industrial chain completeness ranking": 2.5,
+    "Guangdong technological capability ranking": 2.0,
+    "Guangdong market performance ranking": 4.0,
+    "Guangdong comprehensive score": 2.75
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/industry_planning/medium014.json b/assets/qa_gold/industry_planning/medium014.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a373c936fcfeaed083d853e5ff60291e488c6af
--- /dev/null
+++ b/assets/qa_gold/industry_planning/medium014.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium014",
+  "question": "In the 2022 general equipment manufacturing industry provincial data, find the province with the highest total government reward and subsidy amount; after identifying that province, conduct a comprehensive rating of its industrial competitiveness across two strategic dimensions: the \"industrial upgrade capability\" dimension combines the inter-provincial mean ranking of three sub-indicators—mean year-on-year change in R&D investment (R&D investment growth rate), year-on-year growth rate of annual Chinese patent applications (patent application growth rate), and mean R&D personnel ratio (high-end talent ratio); the \"industrial foundation\" dimension combines the inter-provincial mean ranking of three sub-indicators—total number of enterprises (enterprise scale), total operating revenue (revenue scale), and the ratio of total operating revenue to total government reward and subsidy (subsidy efficiency). The overall comprehensive performance ranking is determined by the average of the two dimension rankings. Can that province's comprehensive industrial performance rank among the national top 5?",
+  "guidelines": "Answer format: \"Yes\" or \"No\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter industry=\"general equipment manufacturing industry\", extract each province's government reward funds and subsidy total fields; after dropping nulls and keeping values >0, 14 provinces remain.",
+    "Sort by subsidy total descending; the province with the highest total government subsidy is Shanghai (2466523519.29 yuan).",
+    "From the same table, extract each province's mean year-on-year change in R&D investment, mean R&D personnel share, total number of enterprises, and total operating revenue;",
+    "Compute each province's subsidy efficiency = total operating revenue / combined government reward and subsidy (exclude zero or missing denominators); 14 provinces have computable subsidy efficiency.",
+    "For the six sub-indicators (R&D investment growth rate, patent application YoY growth rate, R&D personnel share, enterprise count, revenue scale, subsidy efficiency), rank provinces separately within each indicator's valid sample from high to low; rank 1 is best; nulls are excluded from that indicator's ranking; ties share the same rank and subsequent ranks are skipped.",
+    "Industrial upgrade capability mean of three sub-indicator ranks = (R&D growth rank + patent application growth rank + R&D personnel share rank) / 3; for Shanghai (15+1+11)/3 = 9.",
+    "Industrial foundation mean of three sub-indicator ranks = (enterprise scale rank + revenue scale rank + subsidy efficiency rank) / 3; for Shanghai (5+1+7)/3 = 4.33 (two decimal places).",
+    "Rank each province's \"industrial upgrade capability three-rank mean\" and \"industrial foundation three-rank mean\" separately in ascending order (smaller mean is better) to obtain the two dimension ranks; overall comprehensive score = (industrial upgrade dimension rank + industrial foundation dimension rank) / 2; for Shanghai (13+4)/2 = 8.5.",
+    "Among provinces that have both dimension ranks (all six sub-indicators yield both dimensional composites), sort by overall score ascending; 13 provinces qualify; Shanghai ranks 10th (tied with Henan on score), which is >5, output \"No\"."
+  ],
+  "steps_num": 9,
+  "milestone": {
+    "Province with highest government subsidy": "Shanghai",
+    "Shanghai total government subsidy (yuan)": 2466523519.29,
+    "Shanghai industrial upgrade capability ranking": 9,
+    "Shanghai industrial foundation ranking": 4.33,
+    "Shanghai comprehensive performance ranking value": 8.5,
+    "Shanghai comprehensive ranking": 10
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/hard001.json b/assets/qa_gold/international_comparison/hard001.json
new file mode 100644
index 0000000000000000000000000000000000000000..b7eaed625244b1be403f2b361c800fef90074191
--- /dev/null
+++ b/assets/qa_gold/international_comparison/hard001.json
@@ -0,0 +1,37 @@
+{
+  "id": "hard001",
+  "question": "Against the backdrop of China's strong push for semiconductor industry self-sufficiency and many regions issuing special support policies, what are the R&D expenses and operating revenue disclosed in United Microelectronics Corporation (UMC)'s 2022 annual report (in New Taiwan Dollars)? What is the R&D investment ratio calculated therefrom? In the domestic semiconductor industry, the top 10% by operating revenue are classified as leading enterprises. By how many percentage points does UMC's R&D investment ratio differ from the median R&D investment ratio of these leading enterprises?",
+  "guidelines": "Please answer in order: (1) UMC 2022 R&D expenses (hundred million NTD, 2 decimal places); (2) UMC 2022 operating revenue (hundred million NTD, 2 decimal places); (3) UMC R&D investment ratio (%, 2 decimal places); (4) Difference between UMC's R&D investment ratio and the median R&D investment ratio of domestic semiconductor industry leading enterprises (top 10% by revenue) (percentage points, 2 decimal places, negative if lower；return as an array). If relevant data cannot be found, please answer \"No relevant data found\".",
+  "answer": [
+    129.53,
+    2787.05,
+    4.65,
+    -1.16
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract financial data from United Microelectronics Corporation (UMC) 2022 Annual Report (Form 20-F) PDF: Research and development expenses NT$12,953 million (i.e., 12.953 billion NTD), Operating revenues NT$278,705 million (i.e., 278.705 billion NTD). The annual report pages 46-47 explicitly state R&D as percentage of revenue is 4.6%.",
+    "Calculate UMC's R&D investment ratio precisely: 12953 / 278705 × 100 = 4.65%.",
+    "Filter from company_profile.csv enterprises with industry='semiconductor industry', 172 enterprises total. Merge with company_operation_status.csv, filter records with non-empty R&D investment ratio, 169 valid records.",
+    "Sort by operating revenue in descending order, take top 10% (operating revenue >= 12.169 billion yuan) as leading enterprises, 17 enterprises; the remaining 152 are non-leading enterprises.",
+    "Calculate the median R&D investment ratio of leading enterprises: 5.81%.",
+    "Calculate the difference between UMC and the leading enterprise median: 4.65 - 5.81 = -1.16 percentage points, indicating UMC's R&D investment intensity is below the median of domestic semiconductor industry leading enterprises."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "UMC 2022 R&D expenses (hundred million NTD)": 129.53,
+    "UMC 2022 operating revenue (hundred million NTD)": 2787.05,
+    "UMC R&D investment ratio (%)": 4.65,
+    "Valid semiconductor industry enterprises": 169,
+    "Leading enterprise count (top 10%)": 17,
+    "Non-leading enterprise count": 152,
+    "Top 10% revenue threshold (hundred million yuan)": 121.69,
+    "Leading enterprise median R&D investment ratio (%)": 5.81,
+    "Non-leading enterprise median R&D investment ratio (%)": 7.98,
+    "UMC vs. leading median difference (percentage points)": -1.16
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/hard002.json b/assets/qa_gold/international_comparison/hard002.json
new file mode 100644
index 0000000000000000000000000000000000000000..43bfd7b8fca77634970047f21c327b2a3626d3ce
--- /dev/null
+++ b/assets/qa_gold/international_comparison/hard002.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard002",
+  "question": "In 2022, against the backdrop of the expiration of new energy vehicle purchase subsidy policies and the transition of industry support from direct subsidies to indirect incentives such as purchase tax exemption, what is the ratio of government subsidies and related income to operating revenue in Li Auto's annual report? Compared with the median government subsidy-to-revenue ratio of private enterprises and state-owned enterprises (including central and local state-owned enterprises) in the domestic automotive manufacturing industry, by how many percentage points is it higher for each?",
+  "guidelines": "Answer in order: Li Auto government subsidy-to-revenue ratio (%), percentage points above private enterprise median, percentage points above state-owned enterprise median. Retain 2 decimal places and return as an array. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": [
+    1.38,
+    0.66,
+    0.8
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from Li Auto (LI) 2022 Annual Report PDF (20-F): Consolidated Statements of Comprehensive Loss - Total revenues = RMB 45,286,816 thousand; Others, net = RMB 625,633 thousand. According to the accounting policy notes in the annual report, non-designated-purpose government subsidies (Other subsidies) are recorded as income under Others, net; this line item also includes VAT refunds of RMB 234,531 thousand and other government-related income.",
+    "Calculate Li Auto's 2022 government subsidy and related income as percentage of revenue = 625,633 / 45,286,816 × 100% = 1.38%.",
+    "Filter from company_profile.csv enterprises with industry=\"automotive manufacturing industry\", 230 enterprises total. Group by ownership: 161 private enterprises, 31 local state-owned enterprises, 17 central state-owned enterprises, 17 foreign enterprises, 4 Sino-foreign joint ventures.",
+    "Merge with company_operation_status.csv, read \"government reward and subsidy\" and \"operating revenue amount\" fields for each enterprise. After excluding records with zero revenue or missing subsidy data, 224 valid samples. Calculate each enterprise's government subsidy-to-revenue ratio = government reward and subsidy / operating revenue amount × 100%.",
+    "Calculate median by ownership: 156 private enterprises, median = 0.72%; state-owned enterprises (16 central + 31 local) = 47 total, median = 0.58%.",
+    "Calculate comparison gap: Li Auto is 1.38% - 0.72% = 0.66 percentage points above private enterprise median; 1.38% - 0.58% = 0.80 percentage points above state-owned enterprise median."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Li Auto total revenue (thousand RMB)": 45286816,
+    "Li Auto Others, net (thousand RMB)": 625633,
+    "Li Auto government subsidy-to-revenue ratio (%)": 1.38,
+    "Valid automotive manufacturing private enterprise samples": 156,
+    "Valid automotive manufacturing state-owned enterprise samples": 47,
+    "Private enterprise median subsidy-to-revenue ratio (%)": 0.72,
+    "State-owned enterprise median subsidy-to-revenue ratio (%)": 0.58,
+    "Li Auto above private median (percentage points)": 0.66,
+    "Li Auto above state-owned median (percentage points)": 0.8
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/hard003.json b/assets/qa_gold/international_comparison/hard003.json
new file mode 100644
index 0000000000000000000000000000000000000000..94cdc416bd453a313a396ac3bc7fc023908279f8
--- /dev/null
+++ b/assets/qa_gold/international_comparison/hard003.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard003",
+  "question": "2022年，在碳达峰政策驱动光伏装机需求激增、多晶硅阶段性供不应求导致价格大幅上涨的背景下，大全新能源（Daqo New Energy）年报中的净利润率（净利润÷营业收入×100%）是多少？分别与国内化学原料和化学制品制造业中民营企业和国有企业（含中央及地方国有企业）的净利润率中位数相差多少个百分点？",
+  "guidelines": "依次回答：大全新能源2022年净利润率（%）、与民营企业中位数之差（百分点）、与国有企业中位数之差（百分点）。数值保留2位小数，并返回数组。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    53.81,
+    45.48,
+    44.76
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "从大全新能源（DQ）2022年度报告PDF（20-F年报）的合并损益表（Consolidated Statements of Operations）中提取：Revenues = $4,608,350千美元；Net income = $2,479,642千美元。",
+    "计算大全新能源2022年净利润率 = 2,479,642 / 4,608,350 × 100% = 53.81%。由于计算比率，美元计价与人民币计价结果一致。",
+    "从company_profile.csv中筛选industry为\"化学原料和化学制品制造业\"的企业，共364家。按ownership分组：民营企业263家、地方国有企业62家、中央国有企业21家、国有企业（其他）1家、国有企业（院所）2家、外资企业13家、集体企业1家、中外合资经营企业1家。",
+    "关联company_operation_status.csv，读取各企业的\"净利润金额\"和\"营业收入金额\"字段，剔除营收为0或数据缺失的记录后，有效样本364家。计算每家企业的净利润率 = 净利润金额 / 营业收入金额 × 100%。",
+    "分所有制计算净利润率中位数：民营企业263家，中位数 = 8.33%；国有企业（中央国有企业21家 + 地方国有企业62家 + 国有企业（其他）1家 + 国有企业（院所）2家）共86家，中位数 = 9.05%。",
+    "综合分析：大全新能源2022年净利润率53.81%，远高于国内化工行业民营企业中位数8.33%（高45.48个百分点）和国有企业中位数9.05%（高44.76个百分点）；2022年多晶硅价格阶段性大幅上涨，大全作为多晶硅龙头企业净利润率远超传统化工行业水平。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "大全新能源Revenue(千美元)": 4608350,
+    "大全新能源Net income(千美元)": 2479642,
+    "大全新能源净利润率(%)": 53.81,
+    "化工业民营企业有效样本(家)": 263,
+    "化工业国有企业有效样本(家)": 86,
+    "民营企业净利润率中位数(%)": 8.33,
+    "国有企业净利润率中位数(%)": 9.05,
+    "大全高于民营中位数(百分点)": 45.48,
+    "大全高于国有中位数(百分点)": 44.76
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/hard004.json b/assets/qa_gold/international_comparison/hard004.json
new file mode 100644
index 0000000000000000000000000000000000000000..19a33e5c8d6a1029114f4ad5e90ae5dfc1f78150
--- /dev/null
+++ b/assets/qa_gold/international_comparison/hard004.json
@@ -0,0 +1,36 @@
+{
+  "id": "hard004",
+  "question": "In 2022, against the backdrop of intensively issued policies promoting biopharmaceutical industry cluster development and encouraging innovative drug R&D across regions, what is Zai Lab's price-to-sales ratio (market cap ÷ annual operating revenue) in multiples? How many times is its price-to-sales ratio relative to the median price-to-sales ratio of domestic pharmaceutical manufacturing industry leading enterprises (top 10% by revenue)?",
+  "guidelines": "Answer in order: (1) Zai Lab 2022 price-to-sales ratio (multiples); (2) Zai Lab's price-to-sales ratio as a multiple of the domestic pharmaceutical manufacturing industry top 10% by revenue leading enterprises' median price-to-sales ratio. Retain 2 decimal places for all values and return as an array. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": [
+    15.17,
+    7.73
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from Zai Lab (ZLAB) 2022 Annual Report (10-K): Consolidated Statements of Operations - Total revenues = $215,040 thousand, including Product revenue, net $212,672 thousand and Collaboration revenue $2,368 thousand.",
+    "Extract from annual report Consolidated Balance Sheets: As of December 31, 2022, issued and outstanding common shares = 960,219,570. Each ADS represents 10 common shares, hence ADS equivalent = 96,021,957. ZLAB ADS closing price on December 30, 2022 (last trading day of year) = $33.97. Calculate market cap = 96,021,957 × $33.97 = $3,261,865,879.",
+    "Calculate Zai Lab price-to-sales ratio = market cap / operating revenue = $3,261,865,879 / $215,040,000 = 15.17x.",
+    "Filter from company_profile.csv enterprises with industry=\"pharmaceutical manufacturing industry\", 449 enterprises total. Merge with company_operation_status.csv; after excluding records with zero operating revenue or missing/zero company market cap, 436 valid samples.",
+    "Calculate each enterprise's price-to-sales ratio = company market cap (hundred million yuan) × 10^8 / operating revenue amount (yuan). Rank by operating revenue, take top 10% (revenue ≥ 9.266 billion yuan) as leading enterprises, 44 enterprises. Leading enterprise median price-to-sales ratio = 1.96x.",
+    "Calculate Zai Lab's price-to-sales ratio as a multiple of domestic pharmaceutical manufacturing industry top 10% leading enterprise median: 15.17 / 1.96 = 7.73x."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Zai Lab annual revenue (thousand USD)": 215040,
+    "Zai Lab outstanding common shares": 960219570,
+    "ADS year-end closing price (USD)": 33.97,
+    "Zai Lab market cap (USD)": 3261865879,
+    "Zai Lab price-to-sales ratio (x)": 15.17,
+    "Valid pharmaceutical manufacturing samples": 436,
+    "Leading enterprise revenue threshold (yuan)": 9266378824,
+    "Leading enterprise count": 44,
+    "Leading enterprise median price-to-sales ratio (x)": 1.96,
+    "Industry-wide median price-to-sales ratio (x)": 4.21,
+    "Zai Lab to leading median ratio (x)": 7.73
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/hard005.json b/assets/qa_gold/international_comparison/hard005.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae788257abf8d2bba0572382a56692fba27f4cda
--- /dev/null
+++ b/assets/qa_gold/international_comparison/hard005.json
@@ -0,0 +1,37 @@
+{
+  "id": "hard005",
+  "question": "In 2022, against the backdrop of intensively issued policies in Shanghai, Hefei, Hangzhou and elsewhere promoting high-quality development of the integrated circuit industry, what is Silicon Motion Technology's per capita net profit in the annual report converted to RMB in ten thousand yuan? Compared with the median per capita net profit of private enterprises and state-owned enterprises in the domestic semiconductor industry respectively, by how many times is it higher?",
+  "guidelines": "Answer in order: Silicon Motion 2022 per capita net profit (ten thousand yuan, converted at 2022 average exchange rate 1 USD ≈ 6.73 RMB), ratio to domestic semiconductor industry private enterprise median per capita net profit (times), ratio to state-owned enterprise median per capita net profit (times). Retain 2 decimal places and return as an array. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": [
+    70.66,
+    8.23,
+    4.79
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from Silicon Motion (SIMO) 2022 Annual Report PDF (20-F): Consolidated Statements of Income - Net Income = US$172,510 thousand (i.e., $172.51 million). From Item 6 Employees section: As of December 31, 2022, the company had 1,643 employees.",
+    "Calculate Silicon Motion 2022 per capita net profit: 172,510,000 / 1,643 = 104,996.96 USD per person. Convert to RMB at 2022 average rate 6.73: 104,996.96 × 6.73 = 706,629.52 yuan = 70.66 ten thousand yuan.",
+    "Filter from company_profile.csv enterprises with industry=\"semiconductor industry\", 172 enterprises total. Group by ownership: 111 private enterprises, 20 foreign enterprises, 16 local state-owned enterprises, 12 central state-owned enterprises, 9 Sino-foreign joint ventures, 2 other state-owned, 2 research-institute state-owned.",
+    "Merge with company_operation_status.csv, read \"net profit amount\" and \"total employees\" fields, calculate each enterprise's per capita net profit = net profit amount / total employees, convert to ten thousand yuan. Exclude records with zero employees or missing data.",
+    "Calculate median per capita net profit by ownership: 111 private enterprises, median = 8.58 ten thousand yuan; state-owned enterprises (12 central + 16 local + 2 other + 2 research-institute) = 32 total, median = 14.75 ten thousand yuan.",
+    "Calculate Silicon Motion per capita net profit as multiple of domestic semiconductor industry private enterprise median: 70.66 / 8.58 = 8.23x; as multiple of state-owned enterprise median: 70.66 / 14.75 = 4.79x."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Silicon Motion Net Income (thousand USD)": 172510,
+    "Silicon Motion employee count": 1643,
+    "Silicon Motion per capita net profit (USD)": 104996.96,
+    "2022 average exchange rate (USD/RMB)": 6.73,
+    "Silicon Motion per capita net profit (ten thousand RMB)": 70.66,
+    "Semiconductor industry private enterprise sample count": 111,
+    "Semiconductor industry state-owned enterprise sample count": 32,
+    "Private enterprise median per capita net profit (ten thousand yuan)": 8.58,
+    "State-owned enterprise median per capita net profit (ten thousand yuan)": 14.75,
+    "Silicon Motion vs. private ratio (x)": 8.23,
+    "Silicon Motion vs. state-owned ratio (x)": 4.79
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/hard006.json b/assets/qa_gold/international_comparison/hard006.json
new file mode 100644
index 0000000000000000000000000000000000000000..179236e965b6d0b87af40bc2c85ec037de6161f8
--- /dev/null
+++ b/assets/qa_gold/international_comparison/hard006.json
@@ -0,0 +1,34 @@
+{
+  "id": "hard006",
+  "question": "A global technology thematic portfolio uses a \"technological moat + policy diffusion\" framework for its semiconductor sub-portfolio. For candidate companies, first calculate: 1. Technological moat gap = company advanced process revenue ratio - median R&D investment ratio of A-share semiconductor industry top 10% by revenue in 2022, where advanced process revenue ratio = (5nm + 7nm revenue) ÷ wafer revenue; 2. Policy diffusion ratio = count of China semiconductor industry policies ÷ number of provincial-level administrative regions covered by non-national policies; 3. Theme conviction score = 0.6 × technological moat gap + 0.4 × policy diffusion ratio. If technological moat gap > 40 and policy diffusion ratio > 2.5, list as core overweight with active weight = min(3.00%, theme conviction score ÷ 10); otherwise do not include in core overweight list. Using TSMC's 2022 annual report and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: technological moat gap (percentage points), policy diffusion ratio, theme conviction score, active weight, most appropriate conclusion (conclusion must specify position action and active weight). Retain 2 decimal places and return as an array. If relevant data cannot be found, please answer \"No relevant data found\".",
+  "answer": [
+    46.6,
+    2.93,
+    29.13,
+    "Core overweight, active weight 2.91%"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Web search TSMC 2022 annual report; from the process revenue disclosure section extract 2022 5nm revenue = 508,689.9, 7nm revenue = 535,153.8, wafer revenue = 1,991,855.9, all in NT$ million.",
+    "Calculate advanced process revenue ratio = (508,689.9 + 535,153.8) ÷ 1,991,855.9 × 100% = 52.4056%, rounded to 2 decimals = 52.41%.",
+    "Filter from company_profile.csv A-share companies with industry=\"semiconductor industry\", 172 total; then from company_operation_status.csv extract 2022 operating revenue amount and R&D investment ratio for these companies, retaining 169 companies with both fields valid.",
+    "Sort the 169 companies by operating revenue descending, take top 10% = 17 companies, calculate median R&D investment ratio = 5.81%.",
+    "Calculate technological moat gap per question definition = 52.41% - 5.81% = 46.60 percentage points.",
+    "Filter semiconductor industry policies from policy_release_status.csv, 44 records; after excluding province=\"national\", 15 provincial-level administrative regions covered.",
+    "Calculate policy diffusion ratio = 44 ÷ 15 = 2.9333, rounded to 2.93; then theme conviction score = 0.6 × 46.60 + 0.4 × 2.93 = 29.1320, rounded to 29.13. Since technological moat gap > 40 and policy diffusion ratio > 2.5, core overweight condition is met; active weight = min(3.00%, 29.13 ÷ 10) = 2.91%."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "TSMC advanced process revenue ratio (%)": 52.41,
+    "A-share semiconductor industry top 10% by revenue median R&D investment ratio (%)": 5.81,
+    "Technological moat gap (percentage points)": 46.6,
+    "Policy diffusion ratio": 2.93,
+    "Theme conviction score": 29.13,
+    "Active weight (%)": 2.91
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/hard007.json b/assets/qa_gold/international_comparison/hard007.json
new file mode 100644
index 0000000000000000000000000000000000000000..c7f5ebdc731daf56d8fce509157f20aa7e9b20de
--- /dev/null
+++ b/assets/qa_gold/international_comparison/hard007.json
@@ -0,0 +1,39 @@
+{
+  "id": "hard007",
+  "question": "A global electric vehicle growth portfolio uses an \"innovation offsets earnings deficit\" framework for loss-making but high-R&D complete vehicle companies. For each candidate, first calculate: 1. Innovation excess = company R&D-to-revenue ratio − median R&D-to-revenue ratio among A-share automotive manufacturing firms in the top 10% by operating revenue in 2022; 2. Profit gap = median net profit margin among those top-10%-by-revenue A-share automotive manufacturing firms − company net profit margin; 3. Policy leverage = number of China automotive manufacturing industry policies ÷ number of provincial-level administrative regions covered by non-national policies. If innovation excess > 8, profit gap < 10, and policy leverage > 1.5, tactical overweight is allowed with active weight = min(2.00%, innovation excess ÷ 5 − profit gap ÷ 10 + policy leverage ÷ 10); otherwise only the watch list applies. Using Li Auto (LI) 2022 annual report and the local database, compute and state the most appropriate conclusion.",
+  "guidelines": "Answer in order: innovation excess (percentage points), profit gap (percentage points), policy leverage, most appropriate conclusion. Retain 2 decimal places and return as an array; the conclusion must specify position action and active weight. If relevant data cannot be found, please answer \"No relevant data found\".",
+  "answer": [
+    10.19,
+    7.73,
+    3.45,
+    "Tactical overweight, active weight 1.61%"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Web search Li Auto 2022 annual report; take Total revenues 45,286,816, Research and development expenses 6,780,032, Net loss 2,032,348 (RMB thousand); R&D-to-revenue ratio = 6,780,032 ÷ 45,286,816 × 100% = 14.9713%, rounded to 14.97%; net profit margin = −2,032,348 ÷ 45,286,816 × 100% = −4.4877%, rounded to −4.49%.",
+    "In company_profile.csv, filter industry=\"automotive manufacturing industry\" and A-share exchanges (SZSE, SSE, BSE), excluding Hong Kong listings; join company_operation_status.csv on bmCode with year=2022. The profile lists 230 automotive manufacturing firms (including 7 Hong Kong–listed); the valid A-share sample is 187 firms. Among rows with non-empty R&D-to-revenue ratio, operating revenue, and net profit, sort by operating revenue descending.",
+    "Top-10% count k = max(1, ⌈10% × N⌉); for N = 187, k = 19. For these 19 firms: median R&D-to-revenue ratio is 4.78%; for each firm compute net profit margin = net profit ÷ operating revenue × 100%, median net profit margin is 3.24%.",
+    "Innovation excess = 14.97% − 4.78% = 10.19 percentage points; profit gap = 3.24% − (−4.49%) = 7.73 percentage points.",
+    "In policy_release_status.csv, count as automotive manufacturing policies those whose industry field contains \"automotive manufacturing industry\", 69 records; after excluding province=\"national\", deduplicate province on the remainder to obtain 20 provincial-level regions covered by non-national policies.",
+    "Policy leverage = 69 ÷ 20 = 3.45.",
+    "Because innovation excess 10.19 > 8, profit gap 7.73 < 10, and policy leverage 3.45 > 1.5, tactical overweight conditions are met; active weight = min(2.00%, 10.19 ÷ 5 − 7.73 ÷ 10 + 3.45 ÷ 10) = min(2.00%, 1.61%) = 1.61%."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Li Auto R&D-to-revenue ratio (%)": 14.97,
+    "A-share automotive manufacturing valid sample count (2022)": 187,
+    "Top-10%-by-revenue firm count": 19,
+    "Median R&D-to-revenue ratio of A-share automotive top 10% by revenue (%)": 4.78,
+    "Innovation excess (percentage points)": 10.19,
+    "Median net profit margin of A-share automotive top 10% by revenue (%)": 3.24,
+    "Profit gap (percentage points)": 7.73,
+    "China automotive manufacturing related policy count": 69,
+    "Non-national policy covered provincial regions (deduped)": 20,
+    "Policy leverage": 3.45,
+    "Active weight (%)": 1.61
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/hard008.json b/assets/qa_gold/international_comparison/hard008.json
new file mode 100644
index 0000000000000000000000000000000000000000..d67ef375a6e242c3dabad1316e0a146611379780
--- /dev/null
+++ b/assets/qa_gold/international_comparison/hard008.json
@@ -0,0 +1,33 @@
+{
+  "id": "hard008",
+  "question": "A global consumer defensive portfolio, when screening platform retailers, views fulfillment expenses as having quasi-fixed cost characteristics. For candidate companies, apply the following stress test: ① Fulfillment expense ratio = Fulfillment ÷ Net revenues; ② Stressed net profit margin = company net profit margin - 0.2 × fulfillment expense ratio; ③ Defense gap = stressed net profit margin - median net profit margin of A-share wholesale and retail industry top 10% by revenue in 2022. If stressed net profit margin < 0, do not include in defensive core position; if stressed net profit margin is between 0 and industry median, benchmark hold only; if above industry median, overweight is allowed. Using JD.com (JD) 2022 annual report and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: fulfillment expense ratio (%), stressed net profit margin (%), defense gap (percentage points), most appropriate conclusion. Retain 2 decimal places and return as an array; conclusion must specify position action. If relevant data cannot be found, please answer \"No relevant data found\".",
+  "answer": [
+    6.02,
+    -0.28,
+    -1.16,
+    "Do not include in defensive core position"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Web search JD.com 2022 annual report; extract Net revenues = 1,046,236, Fulfillment = 63,011, Net income = 9,691, all in RMB million.",
+    "Fulfillment expense ratio = 63,011 ÷ 1,046,236 × 100%; net profit margin = 9,691 ÷ 1,046,236 × 100%. Carry full precision through intermediate steps—do not round these ratios to two decimals before the next calculation.",
+    "Stressed net profit margin = net profit margin − 0.2 × fulfillment expense ratio, computed directly from the unrounded ratio values above; only then round the final stressed margin to two decimal places (rounding pre-rounded 0.93% and 6.02% would incorrectly yield −0.27%).",
+    "From company_profile.csv, filter industry consistent with A-share wholesale and retail (批发和零售业) where exchange is Shanghai, Shenzhen, or Beijing (exclude Hong Kong Stock Exchange listings). Merge with company_operation_status.csv for year = 2022; keep records with valid operating revenue and net profit and operating revenue > 0 (n = 185).",
+    "For each company, net profit margin = net profit amount ÷ operating revenue amount × 100%. Sort by operating revenue descending; take the top ⌈n × 10%⌉ = 19 companies; median net profit margin ≈ 0.880508%, rounded to two decimals = 0.88%.",
+    "Defense gap = stressed net profit margin − industry median (in percentage points), using full-precision values before rounding the gap to two decimals. Because stressed net profit margin < 0, JD.com must not be included in the defensive core position."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Fulfillment expense ratio (%)": 6.02,
+    "Net profit margin (%)": 0.93,
+    "Stressed net profit margin (%)": -0.28,
+    "A-share wholesale and retail top 10% by revenue median net profit margin (%)": 0.88,
+    "Defense gap (percentage points)": -1.16,
+    "A-share sample note": "exchange ∈ {Shanghai, Shenzhen, Beijing}; Hong Kong excluded; n = 185; top decile count = 19"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/hard009.json b/assets/qa_gold/international_comparison/hard009.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e335744058a12f2f8f02561d6fbd350f4a1958f
--- /dev/null
+++ b/assets/qa_gold/international_comparison/hard009.json
@@ -0,0 +1,34 @@
+{
+  "id": "hard009",
+  "question": "An active equity manager uses a \"high-profit reinvestment\" framework for the internet retail growth sub-portfolio. For candidate companies, first calculate: ① High-quality growth score = difference between company net profit margin and median net profit margin of A-share wholesale and retail industry top 10% by revenue in 2022 + 0.5 × transaction services revenue ratio + 0.5 × R&D investment ratio; ② Policy diffusion ratio = count of wholesale and retail industry policies ÷ number of provincial-level administrative regions covered by non-national policies. If high-quality growth score > 35 and policy diffusion ratio > 2.0, list as strategic overweight with active weight = min(4.00%, high-quality growth score ÷ 10); otherwise ordinary position only. Using Pinduoduo (PDD) 2022 annual report and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: high-quality growth score, policy diffusion ratio, active weight (%), most appropriate conclusion. Retain 2 decimal places and return as an array; conclusion must specify position action. If relevant data cannot be found, please answer \"No relevant data found\".",
+  "answer": [
+    37.73,
+    2.0,
+    0.0,
+    "Ordinary position"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Web search Pinduoduo 2022 annual report; extract Transaction services = 27,626,494, Total revenues = 130,557,589, Research and development expenses = 10,384,716, Net income = 31,538,062, all in RMB thousand.",
+    "Calculate transaction services revenue ratio = 27,626,494 ÷ 130,557,589 × 100% = 21.1604%, rounded to 21.16%; calculate R&D investment ratio = 10,384,716 ÷ 130,557,589 × 100% = 7.9541%, rounded to 7.95%; calculate net profit margin = 31,538,062 ÷ 130,557,589 × 100% = 24.1564%, rounded to 24.16%.",
+    "Filter from company_profile.csv A-share companies with industry=\"wholesale and retail industry\", 273 total; then from company_operation_status.csv extract 2022 operating revenue amount and net profit amount for these companies.",
+    "Among 273 companies with valid operating revenue and net profit, calculate each company's net profit margin = net profit amount ÷ operating revenue amount × 100%, then sort by operating revenue descending and take top 10% = 28 companies, median net profit margin = 0.97%; thus net profit margin difference = 24.16% - 0.97% = 23.18 percentage points.",
+    "Calculate high-quality growth score per question definition = 23.18 + 0.5 × 21.16 + 0.5 × 7.95 = 23.18 + 10.58 + 3.975 = 37.7350, rounded to 37.73.",
+    "Filter from policy_release_status.csv policies whose industry field contains wholesale and retail industry, 28 records; after excluding province=\"national\", 14 provincial-level administrative regions covered. Calculate policy diffusion ratio = 28 ÷ 14 = 2.0000, rounded to 2.00.",
+    "Since high-quality growth score > 35 but policy diffusion ratio 2.00 does not meet the > 2.0 threshold, strategic overweight condition is not met; ordinary position only, strategic overweight active weight formula not applicable (recorded as 0.00%)."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Transaction services revenue ratio (%)": 21.16,
+    "R&D investment ratio (%)": 7.95,
+    "Net profit margin difference (percentage points)": 23.18,
+    "High-quality growth score": 37.73,
+    "Policy diffusion ratio": 2.0,
+    "Active weight (%)": 0.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/hard010.json b/assets/qa_gold/international_comparison/hard010.json
new file mode 100644
index 0000000000000000000000000000000000000000..e82ea635a48485a53ec7528f84e07d26b98e2ce5
--- /dev/null
+++ b/assets/qa_gold/international_comparison/hard010.json
@@ -0,0 +1,33 @@
+{
+  "id": "hard010",
+  "question": "A private wealth global consumption themed account wishes to incorporate the feature of 'overseas cash flow hedging domestic cycle' for China optional consumption. For candidate companies, apply the following rules: ① Overseas Hedging Quality Score = the difference between the company's net profit margin and the median net profit margin of A-share wholesale and retail industry in 2022 + 0.5 × international market revenue share; ② If the score ≥ 15, list as core holding, with active weight = min(3.00%, Overseas Hedging Quality Score ÷ 8); otherwise, only satellite holding is permitted. Based on MINISO (MNSO) FY2022 annual report and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: International market revenue share (%), the difference between net profit margin and the median net profit margin of A-share wholesale and retail industry (percentage points), Overseas Hedging Quality Score, and the most appropriate conclusion. Retain 2 decimal places for numerical values and return as an array. The conclusion must include position action and active weight. Example: [\"26.20\", \"4.79\", \"17.89\", \"Core holding, active weight 2.24%\"]. If relevant data cannot be found, respond with \"Relevant data not found\".",
+  "answer": [
+    26.2,
+    4.38,
+    17.48,
+    "Core holding, active weight 2.18%"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Search the web for MINISO's FY2022 annual report; extract Revenue of 10,085,649 and Profit for the year of 639,743, both in RMB thousand; and extract international market revenue contribution of 26.2% from segment disclosure.",
+    "International market revenue share is directly disclosed in the annual report as 26.20%. Calculate net profit margin = 639,743 ÷ 10,085,649 × 100% ≈ 6.34% (two decimal places).",
+    "From national_industry_status.csv, select the row where the industry is \"批发和零售业\" (wholesale and retail) and district is \"全国\" (national). Compute the industry-level median net profit margin as median net profit amount ÷ median revenue amount × 100% = 78,000,847 ÷ 3,970,535,358 × 100% ≈ 1.96%.",
+    "Net profit margin difference vs. the industry benchmark = 6.34% − 1.96% = 4.38 percentage points (two decimal places).",
+    "Per the problem definition, Overseas Hedging Quality Score = 4.38 + 0.5 × 26.20 = 17.48.",
+    "Because the score is ≥ 15, the position qualifies as a core holding; active weight = min(3.00%, 17.48 ÷ 8) = 2.18% (two decimal places)."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "International market revenue share (%)": 26.2,
+    "Net profit margin (%)": 6.34,
+    "Median net profit margin of A-share wholesale and retail industry (%)": 1.96,
+    "Net profit margin difference (percentage points)": 4.38,
+    "Overseas Hedging Quality Score": 17.48,
+    "Active weight (%)": 2.18
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/hard011.json b/assets/qa_gold/international_comparison/hard011.json
new file mode 100644
index 0000000000000000000000000000000000000000..a48fca591ecc7bd152066f9721afef2577ab3bfd
--- /dev/null
+++ b/assets/qa_gold/international_comparison/hard011.json
@@ -0,0 +1,32 @@
+{
+  "id": "hard011",
+  "question": "When an industrial internet fund evaluation distribution platform transforms toward high value-added services, it adopts a two-step method: ? Intangible input excess = company's R&D investment ratio - median R&D investment ratio of the A-share wholesale and retail industry in 2022; ? Transformation score = net service revenue ratio + intangible input excess. Only when net service revenue ratio >= 5 and transformation score >= 5 can a company enter the platform-based overweight list; otherwise, it enters the watch list. Based on ZKH's 2022 annual report and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: Net service revenue ratio (%), intangible input excess (percentage points), transformation score, and the most appropriate conclusion. Retain 2 decimal places for numerical values and return as an array. The conclusion must clearly indicate whether to enter the platform-based overweight list or the watch list. If relevant data cannot be found, respond with \"Relevant data not found\".",
+  "answer": [
+    "2.16",
+    "2.49",
+    "4.65",
+    "Watch list"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Search the web for ZKH's 2022 annual report, extract Net service revenues of 179,508, Total revenues of 8,315,236, and Research and development expenses of 240,534, all in RMB thousand.",
+    "Calculate net service revenue ratio = 179,508 ÷ 8,315,236 × 100% = 2.1588%, rounded to 2.16% with two decimal places; calculate R&D investment ratio = 240,534 ÷ 8,315,236 × 100% = 2.8927%, rounded to 2.89% with two decimal places.",
+    "Filter A-share companies with industry=\"wholesale and retail\" from company_profile.csv, totaling 273 companies; then extract the 2022 R&D investment ratio field for these companies from company_operation_status.csv. Among the 141 companies with valid R&D investment ratio, the median is 0.40%.",
+    "According to the problem definition, calculate intangible input excess = 2.89% - 0.40% = 2.49 percentage points.",
+    "Calculate transformation score = 2.16 + 2.49 = 4.65.",
+    "Since net service revenue ratio of 2.16% is below 5%, and transformation score of 4.65 is also below 5, the platform-based overweight list conditions are not met; therefore, the conclusion is watch list."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Net service revenue ratio (%)": 2.16,
+    "R&D investment ratio (%)": 2.89,
+    "Median R&D investment ratio of A-share wholesale and retail industry (%)": 0.4,
+    "Intangible input excess (percentage points)": 2.49,
+    "Transformation score": 4.65
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/hard012.json b/assets/qa_gold/international_comparison/hard012.json
new file mode 100644
index 0000000000000000000000000000000000000000..906595cb7683f12de46d77224b4d5a3fac87aa6a
--- /dev/null
+++ b/assets/qa_gold/international_comparison/hard012.json
@@ -0,0 +1,34 @@
+{
+  "id": "hard012",
+  "question": "A real estate transformation special account evaluates companies that replace development cycles with existing property services. For candidate companies, apply the following rules: ① Transformation buffer = home renovation and furnishing revenue ratio + R&D investment ratio; ② Profit gap = median net profit margin of top 10% A-share real estate companies by revenue in 2022 - company net profit margin; ③ Net transformation score = transformation buffer - profit gap. If net transformation score ≤ 0, exclude; if 0 < net transformation score < 5, only tactical small overweight is permitted, with active weight = min(1.50%, net transformation score ÷ 10); if net transformation score ≥ 5, standard overweight is permitted. Based on KE Holdings (BEKE) 2022 annual report and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: Transformation buffer, profit gap (percentage points), net transformation score, and the most appropriate conclusion. Retain 2 decimal places for numerical values and return as an array. The conclusion must include position action and active weight. If relevant data cannot be found, respond with \"Relevant data not found\".",
+  "answer": [
+    12.51,
+    4.22,
+    8.29,
+    "Standard overweight; the rules do not specify a concrete active weight percentage"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Search the web for KE Holdings' 2022 annual report, extract Home renovation and furnishing revenue of 5,046,627, Total net revenues of 60,668,779, Research and development expenses of 2,545,549, and Net loss of 1,397,284, all in RMB thousand.",
+    "Calculate home renovation and furnishing revenue ratio = 5,046,627 ÷ 60,668,779 × 100% = 8.3148%, rounded to 8.31% with two decimal places; calculate R&D investment ratio = 2,545,549 ÷ 60,668,779 × 100% = 4.1958%, rounded to 4.20% with two decimal places; calculate net profit margin = -1,397,284 ÷ 60,668,779 × 100% = -2.3032%, rounded to -2.30% with two decimal places.",
+    "According to the problem definition, calculate transformation buffer = 8.31 + 4.20 = 12.51.",
+    "From company_profile.csv, filter A-share companies with companyType=\"沪深\" and exchange in SZSE/SSE/BSE and industry=\"房地产业\"; merge with company_operation_status.csv records where year=2022 on bmCode.",
+    "Among samples with valid operating revenue > 0 (110 companies in total), sort by operating revenue descending, take top 10% = ceil(110×10%) = 11 companies, compute net profit margin = net profit amount ÷ operating revenue amount × 100% for each; median net profit margin is 1.92%; therefore profit gap = 1.92% - (-2.30%) = 4.22 percentage points.",
+    "Calculate net transformation score = 12.51 - 4.22 = 8.29. Since net transformation score ≥ 5, KE Holdings is eligible for standard overweight; the problem only specifies active weight formula min(1.50%, net transformation score ÷ 10) for tactical small overweight, and does not specify a concrete active weight percentage for standard overweight."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Home renovation and furnishing revenue ratio (%)": 8.31,
+    "R&D investment ratio (%)": 4.2,
+    "Transformation buffer": 12.51,
+    "Median net profit margin of top 10% A-share real estate companies by revenue (%)": 1.92,
+    "Profit gap (percentage points)": 4.22,
+    "Net transformation score": 8.29,
+    "Active weight (%)": null
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/hard013.json b/assets/qa_gold/international_comparison/hard013.json
new file mode 100644
index 0000000000000000000000000000000000000000..f21a98bffd1a425d7bebf49902d5b70d75642fd9
--- /dev/null
+++ b/assets/qa_gold/international_comparison/hard013.json
@@ -0,0 +1,32 @@
+{
+  "id": "hard013",
+  "question": "A quasi-infrastructure growth portfolio allows allocation to data center operators during accounting loss periods, but requires a significant operating profit buffer. For candidate companies, apply the following rules: ① Profit conversion penalty = |net profit margin| ÷ adjusted EBITDA margin × 100; ② Policy diffusion ratio = number of data center, Eastern Data Western Computing, or computing power related policies ÷ number of provincial-level administrative regions covered by non-national policies; ③ Infrastructure capacity score = adjusted EBITDA margin - profit conversion penalty + 5 × policy diffusion ratio. If profit conversion penalty < 35 and infrastructure capacity score > 25, the company may be listed for satellite overweight 1.00%; otherwise, benchmark allocation only. Based on GDS Holdings (GDS) 2022 annual report and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: Profit conversion penalty, policy diffusion ratio, infrastructure capacity score, and the most appropriate conclusion. Retain 2 decimal places for numerical values and return as an array. The conclusion must clearly indicate position action. If relevant data cannot be found, respond with \"Relevant data not found\".",
+  "answer": [
+    29.77,
+    1.25,
+    22.08,
+    "Benchmark allocation only"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Search the web for GDS Holdings' 2022 annual report, extract Net revenue of 9,325,631 and Net loss of 1,266,118, both in RMB thousand; and read Adjusted EBITDA margin of 45.6% from the annual report.",
+    "Calculate net profit margin = −1,266,118 ÷ 9,325,631 ≈ −0.135763, i.e. about −13.58% (two decimal places for the percentage).",
+    "Per the problem definition, adjusted EBITDA margin 45.6% corresponds to decimal 0.456; profit conversion penalty = (1,266,118 ÷ 9,325,631) ÷ 0.456 × 100 ≈ 29.77 (retain two decimal places; avoid bias from rounding net profit margin before dividing).",
+    "From policy_release_status.csv, filter policies whose title or industry field mentions data centers, Eastern Data Western Computing, or computing power, totaling 5; treat non-national policies as \"local policies\". For rows with province column \"national\" but where the title or issuing authority indicates a provincial scope (Ningxia Hui Autonomous Region, Guizhou Province related plans), count the actual covered provinces; together with Shanghai Municipality and Yunnan Province this yields 4 provincial-level administrative regions; policy diffusion ratio = 5 ÷ 4 = 1.25.",
+    "Calculate infrastructure capacity score = 45.60 − 29.77 + 5 × 1.25 = 45.60 − 29.77 + 6.25 = 22.08.",
+    "Profit conversion penalty 29.77 < 35, but infrastructure capacity score 22.08 is not greater than 25, so the dual conditions for satellite overweight are not met; the most appropriate conclusion is benchmark allocation only."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Adjusted EBITDA margin (%)": 45.6,
+    "Net profit margin (%)": -13.58,
+    "Profit conversion penalty": 29.77,
+    "Policy diffusion ratio": 1.25,
+    "Infrastructure capacity score": 22.08
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/hard014.json b/assets/qa_gold/international_comparison/hard014.json
new file mode 100644
index 0000000000000000000000000000000000000000..d33aeda7083169cd5903e9d1eaad877012675464
--- /dev/null
+++ b/assets/qa_gold/international_comparison/hard014.json
@@ -0,0 +1,33 @@
+{
+  "id": "hard014",
+  "question": "A distressed-reversal fund applies a \"screen first, value second\" rule to platform retail stocks. For each candidate company, first compute: ① Survival score = technology service revenue ratio + R&D intensity - 0.1×|company net margin - median net margin of A-share wholesale and retail industry in 2022|; ② If company net margin is below -100%, trigger one-vote veto and remove directly; otherwise, only enter the watch pool when survival score ≥ 20. Based on Mogujie (MOGU) annual report for the fiscal year ended March 31, 2022 and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: technology service revenue ratio (%), R&D intensity (%), survival score, and most appropriate conclusion. Values rounded to 2 decimal places; conclusion must clearly state whether the stock is removed; return as an array. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    13.65,
+    24.49,
+    18.95,
+    "Remove"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Web search Mogujie annual report for fiscal year ended March 31, 2022: Technology service revenues = 46,077, Total revenues = 337,469, Research and development expenses = 82,641, Net loss = 642,374, all in thousands of CNY.",
+    "Calculate technology service revenue ratio = 46,077 ÷ 337,469 × 100% = 13.6537%, rounded to 13.65%; R&D intensity = 82,641 ÷ 337,469 × 100% = 24.4885%, rounded to 24.49%; net margin = -642,374 ÷ 337,469 × 100% = -190.3505%, rounded to -190.35%.",
+    "Filter A-share companies with industry=\"wholesale and retail\" from company_profile.csv, 273 companies; extract 2022 operating revenue and net profit from company_operation_status.csv, retain all 273 companies with valid data.",
+    "Compute net margin for each company: net profit ÷ operating revenue × 100%; median net margin of A-share wholesale and retail industry in 2022 = 1.55%. Absolute deviation from industry median = | -190.35% - 1.55% | = 191.90 percentage points.",
+    "Calculate survival score per definition: 13.65 + 24.49 - 0.1×191.90 = 38.14 - 19.19 = 18.95.",
+    "Since company net margin -190.35% is below -100%, one-vote veto is triggered; regardless of survival score, the conclusion is to remove."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Technology service revenue ratio(%)": 13.65,
+    "R&D intensity(%)": 24.49,
+    "Net margin(%)": -190.35,
+    "Median net margin of A-share wholesale and retail industry(%)": 1.55,
+    "Absolute deviation of net margin from median(percentage points)": 191.9,
+    "Survival score": 18.95
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium001.json b/assets/qa_gold/international_comparison/medium001.json
new file mode 100644
index 0000000000000000000000000000000000000000..00dc708453b330ea35504b47634e65904ef75651
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium001.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium001",
+  "question": "What was NetEase's R&D intensity (R&D expenses as a percentage of revenue) in 2022? Compared with the median R&D intensity of listed companies in China's information transmission, software and IT services industry, how many percentage points higher or lower is it?",
+  "guidelines": "Answer both sub-questions: 1) NetEase 2022 R&D intensity (2 decimal places, %); 2) Difference from industry median (2 decimal places, percentage points; positive value means above industry median; return as an array). If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    15.59,
+    3.77
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from NetEase 2022 annual report (20-F) consolidated income statement: total revenue = 96,495,809 thousand CNY, R&D expenses = 15,039,014 thousand CNY; R&D intensity = 15,039,014 / 96,495,809 × 100 = 15.59%.",
+    "Filter company_profile.csv for companies with industry=\"information transmission, software and information technology services\" (information transmission, software and IT services), 644 companies.",
+    "Obtain 2022 \"R&D intensity\" for these companies from company_operation_status.csv; after removing missing values, 606 valid records; median = 11.82%.",
+    "Calculate difference: 15.59% - 11.82% = 3.77 percentage points; NetEase R&D intensity is 3.77 percentage points above industry median."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "NetEase 2022 revenue(thousand CNY)": 96495809,
+    "NetEase 2022 R&D expenses(thousand CNY)": 15039014,
+    "NetEase R&D intensity(%)": 15.59,
+    "Industry median R&D intensity(%)": 11.82,
+    "Difference(percentage points)": 3.77
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium002.json b/assets/qa_gold/international_comparison/medium002.json
new file mode 100644
index 0000000000000000000000000000000000000000..32d6ba06d26f6589b9495190a988f01ae49eb4e1
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium002.json
@@ -0,0 +1,29 @@
+{
+  "id": "medium002",
+  "question": "What is the net profit margin (net profit / operating revenue × 100%) of Trip.com Group according to its 2022 annual report? What is the difference in percentage points compared with the median net profit margin of listed companies in China's transport, storage and postal services industry?",
+  "guidelines": "Answer in order: Trip.com net profit margin (%), and the difference from domestic industry median (percentage points; negative value means Trip.com is below industry median). Values rounded to 2 decimal places, returned as an array, e.g. [-3.50, -1.20]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    6.82,
+    -1.46
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from Trip.com Group (TCOM) 2022 20-F annual report, Selected Consolidated Statements of Income/(Loss) Data section: Net revenues = RMB 20,039 million, Net income = RMB 1,367 million. Net profit margin = 1,367 / 20,039 × 100% = 6.82%.",
+    "Filter company_profile.csv for companies with industry=\"transport, storage and postal services\" (transport, storage and postal services), 176 companies.",
+    "Join the 176 companies with company_operation_status.csv by bmCode, obtain net profit and operating revenue for each, compute net profit margin (net profit / operating revenue × 100%); all 176 have valid data. Median net profit margin = 8.28%.",
+    "Calculate difference between Trip.com and domestic transport industry median: 6.82% - 8.28% = -1.46 percentage points; Trip.com net profit margin is below domestic industry median."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Trip.com 2022 net revenue(RMB million)": 20039,
+    "Trip.com 2022 net profit(RMB million)": 1367,
+    "Trip.com net profit margin(%)": 6.82,
+    "Domestic transport industry company count": 176,
+    "Domestic transport industry median net profit margin(%)": 8.28,
+    "Net profit margin difference(percentage points)": -1.46
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium003.json b/assets/qa_gold/international_comparison/medium003.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c18a70caf06af19afa712883f2ccc3760a0f703
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium003.json
@@ -0,0 +1,29 @@
+{
+  "id": "medium003",
+  "question": "What is the revenue per employee (operating revenue / total employees, unit: ten thousand CNY) of Bilibili according to its 2022 annual report? Compared with the median revenue per employee of listed companies in China's \"information transmission, software and IT services\" industry, how many times the industry median is Bilibili's revenue per employee?",
+  "guidelines": "Answer in order: Bilibili revenue per employee (ten thousand CNY), and the multiple of industry median. Values rounded to 2 decimal places, e.g. [197.43, 2.16]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    197.43,
+    2.16
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from Bilibili (BILI) 2022 20-F annual report, Consolidated Results of Operations: 2022 Net revenues = RMB 21,899,167 thousand (i.e. 21.899 billion); from Item 6D. Employees, total employees as of Dec 31, 2022 = 11,092. Revenue per employee = 21,899,167,000 / 11,092 / 10,000 = 197.43 ten thousand CNY.",
+    "Filter company_profile.csv for companies with industry==\"information transmission, software and information technology services\" (information transmission, software and IT services), 644 companies.",
+    "Join the 644 companies with company_operation_status.csv by company name; obtain operating revenue and total employees for each. After excluding missing values and records with zero employees, 642 valid companies remain. Revenue per employee = operating revenue / total employees / 10000 (ten thousand CNY); median = 91.24 ten thousand CNY.",
+    "Calculate multiple of Bilibili revenue per employee vs industry median: 197.43 / 91.24 = 2.16x."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Bilibili 2022 net revenue(RMB thousand)": 21899167,
+    "Bilibili 2022 total employees": 11092,
+    "Bilibili revenue per employee(ten thousand CNY)": 197.43,
+    "Domestic IT industry valid company count": 642,
+    "Domestic IT industry median revenue per employee(ten thousand CNY)": 91.24,
+    "Multiple": 2.16
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium004.json b/assets/qa_gold/international_comparison/medium004.json
new file mode 100644
index 0000000000000000000000000000000000000000..e2fad7117896255f96ead91d9c97cae6bf0a519b
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium004.json
@@ -0,0 +1,29 @@
+{
+  "id": "medium004",
+  "question": "What is the net profit margin (net profit ÷ operating revenue × 100%) of Vipshop according to its 2022 annual report? How many percentage points higher is it compared with the median net profit margin of listed companies in China's wholesale and retail industry?",
+  "guidelines": "Answer in order: Vipshop net profit margin (%, 2 decimal places) and the difference from domestic industry median (percentage points, 2 decimal places; positive value means Vipshop is higher), e.g. [6.12, 4.57]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    6.12,
+    4.57
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from Vipshop 2022 20-F annual report PDF, Consolidated Statements of Income: 2022 Total net revenues = RMB 103,152,489 thousand, Net income = RMB 6,311,835 thousand. Net profit margin = 6,311,835 / 103,152,489 × 100% = 6.12%.",
+    "Filter company_profile.csv for companies with industry=='wholesale and retail', 273 companies.",
+    "Obtain operating revenue and net profit for the 273 companies from company_operation_status.csv; compute net profit margin (net profit ÷ revenue × 100%) for each; all 273 have valid data. Median net profit margin = 1.55%.",
+    "Calculate difference between Vipshop and domestic wholesale and retail industry median: 6.12% - 1.55% = 4.57 percentage points; Vipshop is significantly above domestic industry median."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Vipshop 2022 total net revenue(RMB thousand)": 103152489,
+    "Vipshop 2022 net profit(RMB thousand)": 6311835,
+    "Vipshop net profit margin(%)": 6.12,
+    "Domestic wholesale and retail industry company count": 273,
+    "Domestic wholesale and retail industry median net profit margin(%)": 1.55,
+    "Net profit margin difference(percentage points)": 4.57
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium005.json b/assets/qa_gold/international_comparison/medium005.json
new file mode 100644
index 0000000000000000000000000000000000000000..c6a947d663ce9efd774095233cc5845f0d942e9f
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium005.json
@@ -0,0 +1,30 @@
+{
+  "id": "medium005",
+  "question": "ZTO Express is a profit leader in China's express delivery industry. Based on ZTO Express 2022 annual report, answer in order: (1) ZTO Express 2022 net profit per employee (net profit ÷ total employees, unit: ten thousand CNY); (2) Median net profit per employee of listed companies in China's transport, storage and postal services industry (ten thousand CNY); (3) How many times the industry median is ZTO Express net profit per employee?",
+  "guidelines": "Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [26.76, 13.26, 2.02]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    26.76,
+    13.26,
+    2.02
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from ZTO Express 2022 20-F annual report PDF: 2022 consolidated net profit (Net income) = RMB 6,658,966 thousand (approx. 6.659 billion), total employees (Employees) = 24,888.",
+    "Calculate ZTO net profit per employee: 6,658,966,000 CNY / 24,888 persons / 10,000 = 26.76 ten thousand CNY per person.",
+    "Filter company_profile.csv for industry=='transport, storage and postal services' (transport, storage and postal services), 176 companies. Obtain net profit and total employees for these 176 from company_operation_status.csv; after excluding missing values, 174 valid. Net profit per employee = net profit / total employees / 10000; median = 13.26 ten thousand CNY.",
+    "Calculate multiple: 26.76 / 13.26 = 2.02x. ZTO Express net profit per employee is about 2 times the domestic transport, storage and postal services industry median."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "ZTO Express 2022 net profit(RMB thousand)": 6658966,
+    "ZTO Express 2022 total employees": 24888,
+    "ZTO Express net profit per employee(ten thousand CNY)": 26.76,
+    "Transport storage postal industry valid company count": 174,
+    "Industry median net profit per employee(ten thousand CNY)": 13.26,
+    "ZTO vs industry median multiple": 2.02
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium006.json b/assets/qa_gold/international_comparison/medium006.json
new file mode 100644
index 0000000000000000000000000000000000000000..abc483e0d7e19139c7cc5fe4777ebc4c8317cece
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium006.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium006",
+  "question": "For MINISO (MNSO) FY2022 (ended June 30, 2022), answer in order: (1) What is the total asset turnover ratio? (2) What is the median total asset turnover ratio of listed companies in China's wholesale and retail industry in 2022? (3) What is the difference between the two (positive value means industry median is higher)?",
+  "guidelines": "Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [0.89, 1.02, 0.13]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    0.89,
+    1.02,
+    0.13
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from MINISO (MNSO) FY2022 annual report (20-F): Revenue = RMB 10,085,649 thousand, Total assets = RMB 11,281,788 thousand. Total asset turnover = 10,085,649 / 11,281,788 = 0.89.",
+    "Filter company_profile.csv for industry=\"wholesale and retail\", 273 companies.",
+    "Obtain 2022 operating revenue and total assets for the 273 companies from company_operation_status.csv; after excluding invalid data, compute total asset turnover = operating revenue / total assets for each company.",
+    "Median total asset turnover of domestic wholesale and retail industry = 1.02.",
+    "Calculate difference: 1.02 - 0.89 = 0.13; domestic wholesale and retail industry median is 0.13 higher than MINISO."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "MINISO operating revenue(RMB thousand)": 10085649,
+    "MINISO total assets(RMB thousand)": 11281788,
+    "MINISO total asset turnover": 0.89,
+    "Wholesale and retail industry company count": 273,
+    "Wholesale and retail industry median total asset turnover": 1.02,
+    "Difference": 0.13
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium007.json b/assets/qa_gold/international_comparison/medium007.json
new file mode 100644
index 0000000000000000000000000000000000000000..091d7b0b42211d820fc03175396ba3ad14ffe5f2
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium007.json
@@ -0,0 +1,30 @@
+{
+  "id": "medium007",
+  "question": "H World Group (HTHT) is one of China's leading hotel groups. Based on H World Group 2022 annual report, answer in order: (1) Revenue per employee (total operating revenue ÷ total employees, unit: ten thousand CNY); (2) Median revenue per employee of listed companies in China's accommodation and catering industry (ten thousand CNY); (3) How many times the industry median is H World Group's revenue per employee?",
+  "guidelines": "Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [56.96, 41.29, 1.38]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    56.96,
+    41.29,
+    1.38
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from H World Group 2022 20-F annual report: 2022 Total revenues = RMB 13,862 million (approx. 13.862 billion), total employees (Employees) = 24,335.",
+    "Calculate H World Group revenue per employee: 13,862,000,000 CNY / 24,335 persons / 10,000 = 56.96 ten thousand CNY per person.",
+    "Filter company_profile.csv for industry=\"accommodation and catering\", 35 companies. Obtain operating revenue and total employees from company_operation_status.csv; after excluding invalid data, 32 valid. Revenue per employee = operating revenue / total employees / 10000; median = 41.29 ten thousand CNY.",
+    "Calculate multiple: 56.96 / 41.29 = 1.38x. H World Group revenue per employee is about 1.38 times the domestic accommodation and catering industry median."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "H World Group 2022 total operating revenue(RMB million)": 13862,
+    "H World Group 2022 total employees": 24335,
+    "H World Group revenue per employee(ten thousand CNY)": 56.96,
+    "Accommodation and catering industry valid company count": 32,
+    "Industry median revenue per employee(ten thousand CNY)": 41.29,
+    "H World Group vs industry median multiple": 1.38
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium008.json b/assets/qa_gold/international_comparison/medium008.json
new file mode 100644
index 0000000000000000000000000000000000000000..f867096a005d9271ff278d0c920d09b8ec08df93
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium008.json
@@ -0,0 +1,30 @@
+{
+  "id": "medium008",
+  "question": "Yum China (YUMC) is one of China's largest restaurant chains, with brands including KFC and Pizza Hut. Based on Yum China 2022 annual report, answer in order: (1) Net profit margin (net profit ÷ operating revenue, %); (2) Median net profit margin of listed companies in China's accommodation and catering industry (%); (3) How many percentage points higher is Yum China's net profit margin than that median?",
+  "guidelines": "Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [4.62, -16.17, 20.79]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    4.62,
+    -16.17,
+    20.79
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from Yum China 2022 annual report (10-K) consolidated income statement: 2022 Total revenues = 9,569 million USD, Net Income – Yum China Holdings, Inc. = 442 million USD.",
+    "Calculate Yum China net profit margin: 442 / 9,569 × 100% = 4.62%.",
+    "Filter company_profile.csv for accommodation and catering industry, 35 companies. Join with company_operation_status.csv for net profit and operating revenue; after excluding records with zero or missing revenue, compute net profit margin (net profit / operating revenue × 100%) for each; median = -16.17%.",
+    "Calculate difference: 4.62% - (-16.17%) = 20.79 percentage points. Yum China net profit margin is significantly above the domestic accommodation and catering industry median."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Yum China 2022 operating revenue(million USD)": 9569,
+    "Yum China 2022 net profit(million USD)": 442,
+    "Yum China net profit margin(%)": 4.62,
+    "Accommodation and catering industry valid company count": 35,
+    "Industry median net profit margin(%)": -16.17,
+    "Yum China above industry median(percentage points)": 20.79
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium009.json b/assets/qa_gold/international_comparison/medium009.json
new file mode 100644
index 0000000000000000000000000000000000000000..29d00b0f77f46684d3c5676ac20d2b6c3a4ce91f
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium009.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium009",
+  "question": "What is ASE Technology Holding's net profit margin in 2022? Among listed companies in China's semiconductor industry, the top 10% by operating revenue (count rounded up) are defined as industry leaders. What is ASE Technology Holding's net profit margin? How many percentage points does it differ from the median net profit margin of industry leaders?",
+  "guidelines": "Answer in order: 1) ASE Technology Holding's 2022 net profit margin (2 decimal places, %); 2) Difference between ASE's net profit margin and the median net profit margin of domestic semiconductor industry leaders (top 10% by revenue) in percentage points (2 decimal places). Return as a list. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": [
+    9.17,
+    -0.43
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "From ASE Technology Holding's 2022 annual report (Form 20-F), extract consolidated income statement data: 2022 Net revenues = 670,872,643 thousand TWD, Net profit attributable to owners of the parent = 61,501,545 thousand TWD. Net profit margin = 61,501,545 / 670,872,643 × 100 = 9.17%.",
+    "From company_profile.csv, filter industry = \"semiconductor industry\", 172 companies.",
+    "From company_operation_status.csv, obtain 2022 net profit and operating revenue for these companies; 172 valid records after excluding invalid data. Sort by operating revenue descending; take the top 10% (18 companies, rounded up) as industry leaders, remaining 154 as non-leaders.",
+    "Compute median net profit margins for industry leaders and non-leaders separately; industry leaders median = 9.60%.",
+    "Compute the difference between ASE and the industry leaders' median: 9.17% − 9.60% = −0.43 percentage points; ASE's net profit margin is 0.43 percentage points below the median of domestic semiconductor industry leaders."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "ASE Technology Holding 2022 operating revenue (thousand TWD)": 670872643,
+    "ASE Technology Holding 2022 net profit attributable to owners of the parent (thousand TWD)": 61501545,
+    "ASE Technology Holding net profit margin (%)": 9.17,
+    "Semiconductor industry total enterprise count": 172,
+    "Industry leader enterprise count (top 10%, rounded up)": 18,
+    "Industry leaders median net profit margin (%)": 9.6,
+    "Gap between ASE Technology Holding and industry leaders median (percentage points)": -0.43
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium010.json b/assets/qa_gold/international_comparison/medium010.json
new file mode 100644
index 0000000000000000000000000000000000000000..5ec2571728ed8720eda4010caf78659f68b02aa5
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium010.json
@@ -0,0 +1,29 @@
+{
+  "id": "medium010",
+  "question": "What was Futu Holdings' (FUTU) net profit per employee in 2022, in ten thousand CNY? What is the multiple of Futu's net profit per employee compared with the median net profit per employee of listed companies in China's capital market services industry (securities and diversified financials)?",
+  "guidelines": "Answer both sub-questions: 1) Futu Holdings 2022 net profit per employee (2 decimal places, ten thousand CNY; convert HKD to CNY using 2022 average rate 1 HKD ≈ 0.86 RMB); 2) Multiple of Futu net profit per employee vs domestic capital market services industry median (2 decimal places). If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    90.42,
+    5.62
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from Futu Holdings 2022 annual report (20-F) consolidated financial data: 2022 net profit attributable to shareholders = 2,926,944 thousand HKD, total employees = 2,784.",
+    "Convert net profit to CNY: 2,926,944 × 1000 × 0.86 = 2,517,171,840 CNY; net profit per employee = 2,517,171,840 / 2,784 / 10,000 = 90.42 ten thousand CNY.",
+    "Filter company_profile.csv for capital market services with secondary industry \"securities\" and \"diversified financials\", 202 companies.",
+    "Obtain 2022 net profit and total employees for these companies from company_operation_status.csv; after excluding invalid data, 201 valid records. Compute net profit per employee for each; median = 16.09 ten thousand CNY.",
+    "Calculate multiple: 90.42 / 16.09 = 5.62x; Futu net profit per employee is about 5.62 times the domestic capital market services industry median."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Futu 2022 net profit(thousand HKD)": 2926944,
+    "Futu 2022 employee count": 2784,
+    "Futu net profit per employee(ten thousand CNY)": 90.42,
+    "Domestic capital market services median net profit per employee(ten thousand CNY)": 16.09,
+    "Multiple": 5.62
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium011.json b/assets/qa_gold/international_comparison/medium011.json
new file mode 100644
index 0000000000000000000000000000000000000000..fe7f28579f5f883b7bb3db73e5e208fb11e1b23a
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium011.json
@@ -0,0 +1,30 @@
+{
+  "id": "medium011",
+  "question": "Atour Lifestyle Group (ATAT) is a representative mid-to-high-end chain hotel brand in China. Based on Atour 2022 annual report, answer in order: (1) Asset-liability ratio (total liabilities ÷ total assets, %); (2) Median asset-liability ratio of listed companies in China's accommodation and catering industry (%); (3) How many percentage points higher is Atour's asset-liability ratio than that median?",
+  "guidelines": "Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [75.07, 63.83, 11.24]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    75.07,
+    63.83,
+    11.24
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from Atour Lifestyle Group (ATAT) 2022 annual report PDF consolidated balance sheet: 2022 year-end Total liabilities = RMB 3,574,620 thousand, Total assets = RMB 4,762,026 thousand.",
+    "Calculate Atour asset-liability ratio: 3,574,620 / 4,762,026 × 100% = 75.07%.",
+    "Filter company_profile.csv for industry=\"accommodation and catering\", 35 companies. Obtain 2022 asset-liability ratio for these 35 from company_operation_status.csv; after excluding missing values, 35 valid; median = 63.83%.",
+    "Calculate difference: 75.07% - 63.83% = 11.24 percentage points. Atour's asset-liability ratio is 11.24 percentage points above the domestic accommodation and catering industry median, mainly affected by operating lease liabilities (IFRS 16/ASC 842) recognition."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Atour 2022 total liabilities(RMB thousand)": 3574620,
+    "Atour 2022 total assets(RMB thousand)": 4762026,
+    "Atour asset-liability ratio(%)": 75.07,
+    "Accommodation and catering industry valid company count": 35,
+    "Industry median asset-liability ratio(%)": 63.83,
+    "Atour above industry median(percentage points)": 11.24
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium012.json b/assets/qa_gold/international_comparison/medium012.json
new file mode 100644
index 0000000000000000000000000000000000000000..f958bb91d88ed26a8e3d8c18d478697061d7506d
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium012.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium012",
+  "question": "Full Truck Alliance (NYSE: YMM) is a leading digital freight platform in China. Based on Full Truck Alliance 2022 annual report, answer in order: (1) R&D intensity (R&D expenses ÷ operating revenue, %); (2) Median R&D intensity of listed companies in China's transport, storage and postal services industry (%); (3) How many times the industry median is Full Truck Alliance's R&D intensity?",
+  "guidelines": "Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [13.58, 0.56, 24.25]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    13.58,
+    0.56,
+    24.25
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from Full Truck Alliance 2022 20-F annual report: 2022 Net revenues = RMB 6,733,644 thousand (approx. 6.734 billion), Research and development expenses = RMB 914,151 thousand (approx. 914 million).",
+    "Calculate Full Truck Alliance R&D intensity: 914,151 / 6,733,644 × 100% = 13.58%.",
+    "Filter company_profile.csv for industry=='transport, storage and postal services' (transport, storage and postal services), 176 companies. Obtain R&D intensity for these 176 from company_operation_status.csv; after excluding missing and invalid data, 107 companies have valid R&D intensity.",
+    "Industry median R&D intensity: median for the 107 companies = 0.56%.",
+    "Calculate multiple: 13.58 / 0.56 = 24.25x. As a digital platform company, Full Truck Alliance's R&D intensity is far above traditional transport, storage and postal services listed companies."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Full Truck Alliance 2022 operating revenue(RMB thousand)": 6733644,
+    "Full Truck Alliance 2022 R&D expenses(RMB thousand)": 914151,
+    "Full Truck Alliance R&D intensity(%)": 13.58,
+    "Transport storage postal industry valid company count": 107,
+    "Industry median R&D intensity(%)": 0.56,
+    "Full Truck Alliance vs industry median multiple": 24.25
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium013.json b/assets/qa_gold/international_comparison/medium013.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ca11f4e47611b53ee73c4115d324cc473d66b5b
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium013.json
@@ -0,0 +1,30 @@
+{
+  "id": "medium013",
+  "question": "Himax Technologies (HIMX) is a leading display driver IC design company globally. Based on Himax 2022 annual report, answer in order: (1) R&D intensity (R&D expenses as percentage of operating revenue, %); (2) Median R&D intensity of listed semiconductor companies in China (%); (3) How many times the domestic semiconductor industry median is Himax's R&D intensity?",
+  "guidelines": "Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [14.61, 7.17, 2.04]. Reasoning may be shown in the process, but final answer should include only these three items. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    14.61,
+    7.17,
+    2.04
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from Himax Technologies (HIMX) 2022 20-F annual report consolidated income statement: 2022 Total revenues = $1,201,339 thousand USD, Research and development = $175,557 thousand USD.",
+    "Calculate Himax R&D intensity: $175,557 thousand / $1,201,339 thousand × 100% = 14.61%.",
+    "Filter company_profile.csv for industry=\"semiconductor industry\" (semiconductors), 172 companies. Join with company_operation_status.csv for R&D intensity; after excluding invalid data, 169 valid. Industry median R&D intensity = 7.17%.",
+    "Calculate multiple: 14.61% / 7.17% = 2.04x. Himax R&D intensity is about 2 times the domestic semiconductor industry median."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Himax 2022 R&D expenses(thousand USD)": 175557,
+    "Himax 2022 operating revenue(thousand USD)": 1201339,
+    "Himax R&D intensity(%)": 14.61,
+    "Semiconductor industry valid company count": 169,
+    "Industry median R&D intensity(%)": 7.17,
+    "Himax vs industry median multiple": 2.04
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium014.json b/assets/qa_gold/international_comparison/medium014.json
new file mode 100644
index 0000000000000000000000000000000000000000..781821b3f36981df30610ba95192c7cc8d671f7d
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium014.json
@@ -0,0 +1,32 @@
+{
+  "id": "medium014",
+  "question": "Weibo (WB / Weibo Corporation) is a leading social media platform in China. Based on Weibo 2022 annual report, compute its net profit margin (net profit attributable to Weibo shareholders ÷ operating revenue, in percentage), and conduct peer analysis against listed companies in China's information transmission, software and IT services industry. Rank by operating revenue; top 10% are defined as industry leaders. Answer in order: Weibo's net profit margin, and the difference (in percentage points) between Weibo's net profit margin and the median net profit margin of industry leaders.",
+  "guidelines": "Answer in order: Weibo 2022 net profit margin (%), and the difference between Weibo and industry leaders (top 10% by revenue) median net profit margin (percentage points). Values rounded to 2 decimal places, e.g. [4.66, 1.21]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    4.66,
+    1.21
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from Weibo (WB) 2022 annual report (20-F) consolidated income statement: 2022 Net income attributable to Weibo's shareholders = 85,555 thousand USD, Total net revenues = 1,836,332 thousand USD.",
+    "Calculate Weibo net profit margin: 85,555 / 1,836,332 × 100% = 4.66%.",
+    "Filter company_profile.csv for industry=\"information transmission, software and information technology services\" (information transmission, software and IT services), 644 companies. Join with company_operation_status.csv for 2022 net profit and operating revenue; after excluding invalid data (zero or missing revenue), 644 valid. Compute net profit margin for each.",
+    "Rank by operating revenue descending; top 10% (65 companies) as industry leaders, remaining 579 as non-leaders. Leaders median net profit margin = 3.45%, non-leaders median = 2.25%.",
+    "Calculate difference: 4.66% - 3.45% = 1.21 percentage points. Weibo net profit margin is 1.21 percentage points above domestic IT industry leaders median."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Weibo 2022 net profit attributable to shareholders(thousand USD)": 85555,
+    "Weibo 2022 operating revenue(thousand USD)": 1836332,
+    "Weibo net profit margin(%)": 4.66,
+    "IT industry valid company count": 644,
+    "Industry leaders count": 65,
+    "Industry leaders median net profit margin(%)": 3.45,
+    "Non-leaders median net profit margin(%)": 2.25,
+    "Weibo vs leaders median difference(percentage points)": 1.21
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium015.json b/assets/qa_gold/international_comparison/medium015.json
new file mode 100644
index 0000000000000000000000000000000000000000..3399633ce161c07da7071afeaa17d36839ec3dd0
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium015.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium015",
+  "question": "Delta Electronics (2308.TW) is a leading global provider of power and thermal management solutions. Based on Delta Electronics' 2022 annual report, compute its consolidated net profit margin (net profit / operating revenue × 100%) and benchmark against listed companies in China's Electrical Machinery and Equipment Manufacturing industry: the top 10% by operating revenue are defined as industry leaders. What is Delta Electronics' net profit margin? What is the gap (in percentage points) between Delta's net profit margin and the median net profit margin of industry leaders?",
+  "guidelines": "Answer in order: Delta Electronics 2022 consolidated net profit margin (%), and the gap (percentage points) between Delta Electronics and the median net profit margin of domestic Electrical Machinery and Equipment Manufacturing industry leaders (top 10% by revenue). Round all values to 2 decimal places, e.g. [9.62, 3.30]. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": [
+    9.62,
+    3.3
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "From Delta Electronics' 2022 annual report (ROC calendar year 111), consolidated condensed comprehensive income statement: consolidated operating revenue = 384,443,308 thousand TWD, consolidated net profit for the period = 36,990,738 thousand TWD.",
+    "Compute Delta consolidated net profit margin: 36,990,738 / 384,443,308 × 100% = 9.62%.",
+    "From company_profile.csv, filter industry = \"Electrical Machinery and Equipment Manufacturing\", 320 companies. Join company_operation_status.csv for 2022 net profit and operating revenue; compute each company's net profit margin (net profit / operating revenue × 100%); all 320 records valid.",
+    "Sort by operating revenue descending; top 10% (32 companies) as industry leaders, remaining 288 as non-leaders. Median net profit margin of industry leaders = 6.32%.",
+    "Compute the gap between Delta's net profit margin and the industry leaders' median: 9.62% − 6.32% = 3.30 percentage points. Delta's net profit margin is above the domestic industry leaders' median."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Delta Electronics 2022 consolidated operating revenue (thousand TWD)": 384443308,
+    "Delta Electronics 2022 consolidated net profit (thousand TWD)": 36990738,
+    "Delta Electronics net profit margin (%)": 9.62,
+    "Electrical Machinery and Equipment Manufacturing valid enterprise count": 320,
+    "Industry leader enterprise count": 32,
+    "Industry leaders median net profit margin (%)": 6.32,
+    "Gap between Delta Electronics and industry leaders median (percentage points)": 3.3
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium016.json b/assets/qa_gold/international_comparison/medium016.json
new file mode 100644
index 0000000000000000000000000000000000000000..77e9da6937089b285c33b66bcbac76ace89b0564
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium016.json
@@ -0,0 +1,30 @@
+{
+  "id": "medium016",
+  "question": "In 2022, amid strong policy support for domestic semiconductor self-reliance in China, what was TSMC's R&D intensity (R&D expenses as percentage of revenue) in its annual report? What is the difference in percentage points compared with the median R&D intensity of listed semiconductor companies in China?",
+  "guidelines": "Answer in order: TSMC R&D intensity (%), and the difference from domestic median (percentage points; positive value means TSMC is higher). Values rounded to 2 decimal places. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    7.21,
+    0.04
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from TSMC 2022 20-F annual report PDF: 2022 Net Revenue = NT$2,263,891 million, Research and Development Expenses = NT$163,262 million. R&D intensity = 163,262 / 2,263,891 × 100% = 7.21%.",
+    "Filter company_profile.csv for industry=='semiconductor industry' (semiconductors), 172 companies.",
+    "Obtain 'R&D intensity' for the 172 semiconductor companies from company_operation_status.csv; 169 have valid data. Median = 7.17%.",
+    "Calculate difference between TSMC and domestic semiconductor industry median: 7.21% - 7.17% = 0.04 percentage points; TSMC is slightly above domestic median.",
+    "TSMC R&D intensity (7.21%) differs very little from domestic semiconductor industry median (7.17%)—only 0.04 percentage points."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "TSMC 2022 net revenue(NT$ million)": 2263891,
+    "TSMC 2022 R&D expenses(NT$ million)": 163262,
+    "TSMC R&D intensity(%)": 7.21,
+    "Domestic semiconductor industry company count": 172,
+    "Domestic semiconductor industry median R&D intensity(%)": 7.17,
+    "R&D intensity difference(percentage points)": 0.04
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium017.json b/assets/qa_gold/international_comparison/medium017.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e8740b54e312f39a595a398cde8071be2d1152e
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium017.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium017",
+  "question": "In 2022, under national policies promoting innovative drug R&D, what was BeiGene's R&D expense as a percentage of revenue in its annual report? How many percentage points does that ratio differ from the median R&D-to-revenue ratio among domestic pharmaceutical manufacturing firms that are (1) private enterprises and (2) state-owned enterprises (including centrally administered SOEs, locally administered SOEs, and institute-type SOEs), after excluding zeros and invalid data? What is the gap between the median R&D ratios of private vs. state-owned domestic pharmaceutical firms?",
+  "guidelines": "Answer in order: BeiGene's R&D-to-revenue ratio (%); gap vs. private enterprises' median (percentage points); gap vs. state-owned enterprises' median (percentage points); gap between private and state-owned medians (percentage points). Use 2 decimal places; return as a list. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": [
+    115.86,
+    108.4,
+    111.23,
+    2.83
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "From BeiGene's 2022 annual report (PDF), Consolidated Statements of Operations: Total revenues = $1,415,921 thousand USD, Research and development expense = $1,640,508 thousand USD. R&D-to-revenue ratio = 1,640,508 / 1,415,921 × 100 = 115.86%.",
+    "From company_profile.csv, filter industry = \"Pharmaceutical Manufacturing\", 449 companies. By ownership: private enterprises 346; state-owned enterprises (centrally administered SOEs 16 + locally administered SOEs 49 + institute-type SOEs 2) = 67 in total.",
+    "Join company_operation_status.csv; filter valid R&D-to-revenue ratios > 0. Private enterprises: 326 valid samples, median R&D ratio = 7.46%. State-owned enterprises: 66 valid samples, median R&D ratio = 4.63%.",
+    "Compute gaps vs. benchmarks: BeiGene vs. private median = 115.86% − 7.46% = 108.40 percentage points; BeiGene vs. state-owned median = 115.86% − 4.63% = 111.23 percentage points.",
+    "Gap between private and state-owned medians = 7.46% − 4.63% = 2.83 percentage points; private enterprises show higher R&D intensity than state-owned enterprises."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "BeiGene Total revenues (thousand USD)": 1415921,
+    "BeiGene R&D expense (thousand USD)": 1640508,
+    "BeiGene R&D-to-revenue ratio (%)": 115.86,
+    "Private enterprises valid sample count": 326,
+    "Private enterprises median R&D-to-revenue ratio (%)": 7.46,
+    "State-owned enterprises valid sample count": 66,
+    "State-owned enterprises median R&D-to-revenue ratio (%)": 4.63,
+    "BeiGene vs. private median gap (percentage points)": 108.4,
+    "BeiGene vs. state-owned median gap (percentage points)": 111.23,
+    "Private vs. state-owned median gap (percentage points)": 2.83
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium018.json b/assets/qa_gold/international_comparison/medium018.json
new file mode 100644
index 0000000000000000000000000000000000000000..01806a9481b13674edb969224c0f39bba4c0cf54
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium018.json
@@ -0,0 +1,30 @@
+{
+  "id": "medium018",
+  "question": "In 2022, amid China's new energy vehicle industry policy support, what was XPeng's net profit margin (net profit ÷ operating revenue × 100%) in its annual report? What is the difference in percentage points compared with the median net profit margin of domestic automotive manufacturing industry leaders (top 10% by revenue)?",
+  "guidelines": "Answer in order: XPeng net profit margin (%) and the difference from industry leaders median (percentage points; negative value means XPeng is lower). Values rounded to 2 decimal places, e.g. [-34.03, -37.27]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    -34.03,
+    -37.27
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from XPeng 2022 annual report (20-F) PDF: Total revenues = RMB 26,855,119 thousand (approx. 26.855 billion), Net loss = RMB 9,138,972 thousand (approx. 9.139 billion). Net profit margin = -9,138,972 / 26,855,119 × 100% = -34.03%.",
+    "Filter company_profile.csv for industry=\"automotive manufacturing\", 230 companies; obtain operating revenue and net profit from company_operation_status.csv; after excluding invalid data, 230 companies have complete financial data.",
+    "Rank by operating revenue descending; take top 10% (23 companies) as industry leaders. Compute net profit margin (net profit / operating revenue × 100%) for each leader; median = 3.24%.",
+    "Calculate difference: XPeng net profit margin (-34.03%) - leaders median (3.24%) = -37.27 percentage points."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "XPeng 2022 operating revenue(hundred million CNY)": 268.55,
+    "XPeng 2022 net loss(hundred million CNY)": 91.39,
+    "XPeng net profit margin(%)": -34.03,
+    "Automotive manufacturing total company count": 230,
+    "Industry leaders count": 23,
+    "Industry leaders median net profit margin(%)": 3.24,
+    "Difference(percentage points)": -37.27
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium019.json b/assets/qa_gold/international_comparison/medium019.json
new file mode 100644
index 0000000000000000000000000000000000000000..9750a59976285ddcc5907efa526b569bb974c7bd
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium019.json
@@ -0,0 +1,29 @@
+{
+  "id": "medium019",
+  "question": "In 2022, amid policy support for digital economy and e-commerce development in China, what was PDD Holdings' net profit margin (net profit ÷ operating revenue × 100%) in its annual report? How many percentage points higher is it compared with the median net profit margin of listed companies in China's wholesale and retail industry?",
+  "guidelines": "Answer in order: PDD Holdings net profit margin (%), and the number of percentage points above domestic median. Values rounded to 2 decimal places, e.g. [24.16, 22.60]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    24.16,
+    22.6
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from PDD Holdings 2022 20-F annual report PDF: 2022 Total Revenues = RMB 130,557,589 thousand (approx. 130.558 billion), Net Income = RMB 31,538,100 thousand (approx. 31.538 billion). Net profit margin = 31,538,100 / 130,557,589 × 100% = 24.16%.",
+    "Filter company_profile.csv for industry=='wholesale and retail', 273 companies.",
+    "Obtain net profit and operating revenue for the 273 wholesale and retail companies from company_operation_status.csv; compute net profit margin = net profit / operating revenue × 100% for each; all 273 have valid data. Industry median net profit margin = 1.55%.",
+    "Calculate difference between PDD and domestic wholesale and retail industry median: 24.16% - 1.55% = 22.60 percentage points; PDD is significantly above industry median."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "PDD 2022 total revenue(RMB thousand)": 130557589,
+    "PDD 2022 net profit(RMB thousand)": 31538100,
+    "PDD net profit margin(%)": 24.16,
+    "Wholesale and retail industry company count": 273,
+    "Wholesale and retail industry median net profit margin(%)": 1.55,
+    "Net profit margin difference(percentage points)": 22.6
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium020.json b/assets/qa_gold/international_comparison/medium020.json
new file mode 100644
index 0000000000000000000000000000000000000000..fcaedd53d66b2482ae2c0b84165fc4369c253e7e
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium020.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium020",
+  "question": "In 2022, amid national carbon peaking and clean energy development policies, what was JinkoSolar's asset-liability ratio (total liabilities ÷ total assets × 100%) in its annual report? Compared with the median asset-liability ratios among listed companies in China's Electricity, Heat, Gas and Water Supply industry—separately for state-owned enterprises (including centrally administered SOEs, locally administered SOEs, and other SOEs) and for private enterprises—how many percentage points higher is JinkoSolar's ratio in each case?",
+  "guidelines": "Answer in order: JinkoSolar's asset-liability ratio (%); percentage points above the state-owned median; percentage points above the private enterprise median. Use 2 decimal places; return as an array. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": [
+    75.15,
+    13.88,
+    22.57
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "From JinkoSolar (JKS) 2022 annual report PDF (English Form 20-F), Consolidated Balance Sheets as of December 31, 2022: Total assets = RMB 108,662,182 thousand, Total liabilities = RMB 81,658,289 thousand.",
+    "Compute JinkoSolar's 2022 asset-liability ratio = 81,658,289 / 108,662,182 × 100% = 75.15%.",
+    "From company_profile.csv, filter industry = \"Electricity, Heat, Gas and Water Supply\", 189 companies. By ownership: state-owned enterprises (centrally administered SOEs 44 + locally administered SOEs 81 + other SOEs 1) = 126; private enterprises 56.",
+    "From company_operation_status.csv, read each company's asset-liability ratio; compute group medians: state-owned median = 61.27%, private enterprise median = 52.58%.",
+    "Compute gaps vs. JinkoSolar: above state-owned median 75.15% − 61.27% = 13.88 percentage points; above private enterprise median 75.15% − 52.58% = 22.57 percentage points."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "JinkoSolar total assets (thousand RMB)": 108662182,
+    "JinkoSolar total liabilities (thousand RMB)": 81658289,
+    "JinkoSolar asset-liability ratio (%)": 75.15,
+    "State-owned enterprise count (Electricity, Heat, Gas and Water Supply)": 126,
+    "Private enterprise count (Electricity, Heat, Gas and Water Supply)": 56,
+    "State-owned enterprises median asset-liability ratio (%)": 61.27,
+    "Private enterprises median asset-liability ratio (%)": 52.58,
+    "Above state-owned median (percentage points)": 13.88,
+    "Above private enterprise median (percentage points)": 22.57
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium021.json b/assets/qa_gold/international_comparison/medium021.json
new file mode 100644
index 0000000000000000000000000000000000000000..644b53feb4727b36fb4b04056d108b842270dd6d
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium021.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium021",
+  "question": "In 2022, amid policy support for digital economy development and healthy platform economy regulation in China, what was Alibaba's revenue per employee (operating revenue ÷ total employees) in ten thousand CNY? How many times the median revenue per employee of domestic IT industry leaders (top 10% (round down) by operating revenue in information transmission, software and IT services) is it?",
+  "guidelines": "Answer in order: Alibaba revenue per employee (ten thousand CNY), and the multiple of industry leaders median revenue per employee. Values rounded to 2 decimal places, returned as an array. If relevant data cannot be found, reply \"No relevant data found\".",
+  "answer": [
+    369.31,
+    1.5
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "Extract from Alibaba (BABA) 2022 annual report PDF (20-F, fiscal year ended March 31, 2023, corresponding to calendar year 2022): Consolidated Total Revenue = RMB 868,687 million (i.e. 8,686.87 hundred million), full-time employees as of March 31, 2023 = 235,216.",
+    "Calculate Alibaba revenue per employee = 868,687,000,000 / 235,216 = 3,693,064.34 CNY = 369.31 ten thousand CNY.",
+    "Filter company_profile.csv for industry=\"information transmission, software and information technology services\" (information transmission, software and IT services), 644 companies. Join with company_operation_status.csv; exclude records with empty/zero revenue or employees; 642 valid companies.",
+    "Compute revenue per employee = operating revenue / total employees for each. Rank by operating revenue descending; top 10% (64 companies) as industry leaders. Leaders median revenue per employee = 2,456,389.49 CNY = 245.64 ten thousand CNY.",
+    "Calculate multiple: Alibaba revenue per employee / leaders median = 369.31 / 245.64 = 1.50x."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Alibaba operating revenue(million RMB)": 868687,
+    "Alibaba total employees": 235216,
+    "Alibaba revenue per employee(ten thousand CNY)": 369.31,
+    "IT industry valid company count": 642,
+    "Industry leaders count(top 10%)": 64,
+    "Industry leaders median revenue per employee(ten thousand CNY)": 245.64,
+    "Alibaba vs leaders median multiple": 1.5
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium022.json b/assets/qa_gold/international_comparison/medium022.json
new file mode 100644
index 0000000000000000000000000000000000000000..3bd22e1b8d6df6aeafa0d73ba17beb7bbf112c16
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium022.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium022",
+  "question": "In 2022, against the backdrop of intensive policy rollouts for the new energy vehicle industry, what was the proportion of R&D personnel to total employees in NIO's annual report? By how many percentage points did it differ from the median R&D personnel ratio among private enterprises and state-owned enterprises (including central and local state-owned enterprises) in China's automobile manufacturing industry?",
+  "guidelines": "Answer in order: NIO's R&D personnel ratio (%), the number of percentage points above the private enterprise median, and the number of percentage points above the state-owned enterprise median. Retain 2 decimal places for values. E.g. [37.46, 24.19, 22.02]. If the relevant data cannot be found, answer \"Relevant data not found\"",
+  "answer": [
+    37.46,
+    24.19,
+    22.02
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "From NIO's 2022 Annual Report PDF (20-F) employee section (Item 6.D Employees, As of December 31, 2022), extract: Total employees = 26,763; Product and software development = 10,025 (i.e., R&D personnel).",
+    "Calculate NIO's 2022 R&D personnel ratio = 10,025 / 26,763 × 100% = 37.46%.",
+    "From company_profile.csv, filter enterprises with industry \"Automobile Manufacturing\", totaling 230. Group by ownership: 161 private enterprises (149 with R&D personnel ratio data); 48 state-owned enterprises (17 central + 31 local, 43 with data).",
+    "From company_operation_status.csv, read the R&D personnel ratio field for the above enterprises and calculate the median for each group: Private enterprise median = 13.27%; State-owned enterprise median = 15.44%.",
+    "Calculate NIO's gap from each group: 37.46% - 13.27% = 24.19 percentage points above private enterprise median; 37.46% - 15.44% = 22.02 percentage points above state-owned enterprise median."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "NIO total employees (persons)": 26763,
+    "NIO R&D personnel (persons)": 10025,
+    "NIO R&D personnel ratio (%)": 37.46,
+    "Automobile manufacturing private enterprises (count)": 161,
+    "Automobile manufacturing private enterprises with data (count)": 149,
+    "Automobile manufacturing state-owned enterprises (count)": 48,
+    "Automobile manufacturing state-owned enterprises with data (count)": 43,
+    "Private enterprise R&D personnel ratio median (%)": 13.27,
+    "State-owned enterprise R&D personnel ratio median (%)": 15.44,
+    "Above private enterprise median (percentage points)": 24.19,
+    "Above state-owned enterprise median (percentage points)": 22.02
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium023.json b/assets/qa_gold/international_comparison/medium023.json
new file mode 100644
index 0000000000000000000000000000000000000000..7a94725300ed74fe3fc313417bee0e79ef0cc32a
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium023.json
@@ -0,0 +1,32 @@
+{
+  "id": "medium023",
+  "question": "Against the backdrop of the state promoting high-quality development of the digital economy and platform economy, what were JD.com's total revenue and total number of employees disclosed in its 2022 annual report? What is the per capita revenue in 10,000 yuan calculated therefrom? Among industries with no fewer than 30 domestic enterprises (excluding financial, real estate and diversified industries), how many times is JD.com's per capita revenue compared to the industry with the highest median per capita revenue?",
+  "guidelines": "Please answer in order: (1) JD.com's 2022 total revenue (100 million yuan, 2 decimal places); (2) JD.com's 2022 total employees (persons); (3) JD.com's per capita revenue (10,000 yuan, 2 decimal places); (4) The ratio of JD.com's per capita revenue to the industry with highest median per capita revenue (2 decimal places). Return as an array. If the relevant data cannot be found, answer \"Relevant data not found\".",
+  "answer": [
+    10462.36,
+    450679,
+    232.15,
+    0.69
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "From JD.com's 2022 Annual Report (20-F) PDF, extract financial data: Total net revenues RMB 1,046,236 million (i.e., 1046.236 billion yuan), total employees 450,679 (as of December 31, 2022).",
+    "Calculate JD.com's per capita revenue: 1046.236 billion yuan / 450,679 employees = 232.15 ten thousand yuan per person.",
+    "From company_profile.csv obtain industry classification for each enterprise, merge with company_operation_status.csv, calculate per capita revenue for each enterprise (operating revenue / total employees / 10000, unit: 10,000 yuan). Exclude financial, real estate, and diversified industries; filter industries with enterprise count >= 30.",
+    "For the 36 qualifying industries, rank by median per capita revenue in descending order; confirm the highest industry is metal smelting and rolling, median 335.86 ten thousand yuan (145 enterprises).",
+    "Calculate the ratio of JD.com's per capita revenue to this industry's median: 232.15 / 335.86 = 0.69."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "JD.com 2022 total revenue (100 million yuan)": 10462.36,
+    "JD.com 2022 total employees (persons)": 450679,
+    "JD.com per capita revenue (10,000 yuan)": 232.15,
+    "Qualifying industries count": 36,
+    "Metal smelting and rolling median per capita revenue (10,000 yuan)": 335.86,
+    "Ratio of JD.com per capita revenue to industry median": 0.69
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium024.json b/assets/qa_gold/international_comparison/medium024.json
new file mode 100644
index 0000000000000000000000000000000000000000..34aa42d3df60751441c354181db9792b979472b0
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium024.json
@@ -0,0 +1,37 @@
+{
+  "id": "medium024",
+  "question": "Against the backdrop of national policies promoting high-level AI application and supporting digital economy enterprises in strengthening hard-tech innovation, what was Baidu's (BIDU, listed on NASDAQ) R&D intensity (R&D expenses as a percentage of revenue) in 2022? Compared with A-share listed companies in China's information transmission, software, and information technology services industry, when divided into leading enterprises (revenue ≥ 90th percentile) and non-leading enterprises, by how many percentage points did Baidu's R&D intensity exceed the median R&D ratio of each group?",
+  "guidelines": "Answer in order: Baidu's R&D intensity (%), the number of percentage points above the A-share leading group (revenue ≥ sample 90th percentile) median R&D ratio, and the number of percentage points above the A-share non-leading group. Retain 2 decimal places, return as an array. If the relevant data cannot be found, answer \"Relevant data not found\".",
+  "answer": [
+    18.85,
+    12.59,
+    5.79
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "From Baidu (BIDU) 2022 Annual Report PDF (20-F, filed with SEC) Consolidated Statements of Comprehensive Income, extract: Total revenues = RMB 123,675 million (123.675 billion yuan); Research and development = RMB 23,315 million (23.315 billion yuan).",
+    "Calculate Baidu's 2022 R&D intensity = 23,315 / 123,675 × 100% = 18.85%.",
+    "From company_profile.csv, filter A-share listed companies with industry \"Information transmission, software, and information technology services\" and companyType \"Shanghai-Shenzhen\", 433 enterprises; merge with company_operation_status.csv, filter valid samples with operating revenue > 0 and non-null R&D ratio, 430 enterprises.",
+    "Calculate sample 90th percentile threshold for operating revenue ≈ 5.653 billion yuan; leading group (43 enterprises) with revenue ≥ threshold, non-leading group (387 enterprises) for the rest. Calculate median R&D ratio by group: leading group median = 6.26%, non-leading group median = 13.06%.",
+    "Calculate Baidu's difference from each group median: 18.85% - 6.26% = 12.59 percentage points; 18.85% - 13.06% = 5.79 percentage points."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Baidu total revenue (million RMB)": 123675,
+    "Baidu R&D expenses (million RMB)": 23315,
+    "Baidu R&D intensity (%)": 18.85,
+    "A-share IT industry profile count": 433,
+    "A-share IT industry valid samples": 430,
+    "Revenue 90th percentile threshold (billion yuan)": 56.53,
+    "Leading group enterprises count": 43,
+    "Non-leading group enterprises count": 387,
+    "Leading group R&D intensity median (%)": 6.26,
+    "Non-leading group R&D intensity median (%)": 13.06,
+    "Baidu above leading group median (percentage points)": 12.59,
+    "Baidu above non-leading group median (percentage points)": 5.79
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/international_comparison/medium025.json b/assets/qa_gold/international_comparison/medium025.json
new file mode 100644
index 0000000000000000000000000000000000000000..71356aa8fae7012967057b14b70b43e29aa66d03
--- /dev/null
+++ b/assets/qa_gold/international_comparison/medium025.json
@@ -0,0 +1,29 @@
+{
+  "id": "medium025",
+  "question": "Against the backdrop of intensive policy rollouts for the new energy vehicle industry, what was ZEEKR's asset turnover ratio (revenue divided by total assets) in 2022? How did it compare to the median asset turnover ratio of China's automobile manufacturing industry as a whole?",
+  "guidelines": "Answer in order: ZEEKR's asset turnover ratio (2 decimal places), and the difference from the industry median (2 decimal places; positive means ZEEKR is higher). Return as an array. If the relevant data cannot be found, answer \"Relevant data not found\".",
+  "answer": [
+    1.64,
+    1.05
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  },
+  "steps": [
+    "From ZEEKR (ZK) F-1 prospectus PDF, extract 2022 financial data: Net revenues RMB 31,899,448 thousand, Total assets RMB 19,477,316 thousand. Calculate asset turnover ratio = 31,899,448 / 19,477,316 = 1.64.",
+    "From company_profile.csv, filter enterprises with industry == \"Automobile manufacturing\", 230 enterprises.",
+    "From company_operation_status.csv, obtain 'operating revenue' and 'total assets' for the above 230 automobile manufacturing enterprises; all 230 have valid data. Calculate asset turnover ratio for each = operating revenue / total assets; industry median = 0.59.",
+    "Calculate the difference between ZEEKR and industry median: 1.64 - 0.59 = 1.05; ZEEKR is significantly higher than the industry median, reflecting its asset-light operation and rapid scaling as a new EV brand."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "ZEEKR 2022 net revenue (RMB thousand)": 31899448,
+    "ZEEKR 2022 total assets (RMB thousand)": 19477316,
+    "ZEEKR asset turnover ratio": 1.64,
+    "Automobile manufacturing enterprises count": 230,
+    "Industry asset turnover ratio median": 0.59,
+    "ZEEKR difference from industry median": 1.05
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/hard001.json b/assets/qa_gold/risk_assessment/hard001.json
new file mode 100644
index 0000000000000000000000000000000000000000..b6c399af5795aea1a63931c4b0cd032572592b04
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/hard001.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard001",
+  "question": "In 2022, in the automobile manufacturing industry, focusing on provinces with a total of 8 or more relevant enterprises, what proportion of the qualifying provinces simultaneously meet both the 'high policy dependence' risk (government subsidies as a proportion of operating profit >30%) and the 'high market concentration' risk (operating revenue CR4>60%)?",
+  "guidelines": "Answer format: percentage value (2 decimal places). E.g. 25.67. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 20.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter records with industry=\"Automobile Manufacturing\" and total enterprises>=8, extract province, total operating profit amount, total government reward and subsidy funds, and total operating revenue amount fields, finding 5 provinces.",
+    "Filter provinces with total operating profit amount>0 and all required fields non-null as provinces meeting the statistical criteria, totaling 5 provinces.",
+    "For each province meeting the statistical criteria, calculate government subsidies as a proportion of operating profit = (total government reward and subsidy funds / total operating profit amount) × 100%, and filter provinces with high policy dependence (>30%).",
+    "From company_profile.csv, filter enterprises with industry=\"Automobile Manufacturing\"; from company_operation_status.csv, obtain 2022 operating revenue amount, and group by province.",
+    "For each province meeting the statistical criteria, filter enterprises with non-null operating revenue amount, sort by operating revenue amount in descending order, calculate CR4 = (sum of top 4 enterprises' operating revenue / total operating revenue of all enterprises in that province) × 100%, and filter provinces with high market concentration (>60%).",
+    "Take the intersection of high policy dependence provinces and high market concentration provinces, and count the number of dual-risk provinces as 1.",
+    "Calculate the proportion = (number of dual-risk provinces / total number of provinces meeting statistical criteria) × 100% = (1 / 5) × 100% = 20.00%."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Total number of provinces meeting statistical criteria": 5,
+    "High policy dependence provinces": [
+      "Shanghai"
+    ],
+    "High market concentration provinces": [
+      "Guangdong",
+      "Shanghai",
+      "Zhejiang",
+      "Shandong"
+    ],
+    "Number of dual-risk provinces": 1,
+    "Proportion (%)": 20.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/hard002.json b/assets/qa_gold/risk_assessment/hard002.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e7d29392b71de07af18d9f72d495af0a2fc69ea
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/hard002.json
@@ -0,0 +1,28 @@
+{
+  "id": "hard002",
+  "question": "In 2022, in the communication transmission equipment industry, assume that enterprises with R&D investment intensity below the national median for that industry are eliminated from the market within the next three years, and only enterprises with R&D investment intensity not below the national median remain. What is the ratio of remaining enterprises' operating revenue after elimination to operating revenue before elimination for the province with the highest such ratio?",
+  "guidelines": "Answer format: numeric value (2 decimal places, unit %). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 100.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "From national_industry_status.csv, filter industry=\"Communication Transmission Equipment\", extract the median R&D investment intensity as 9.92, used as the national benchmark.",
+    "From company_profile.csv, filter industry=\"Communication Transmission Equipment\" for all enterprise records, extract enterprise name, bmCode, and province fields; 120 enterprises found.",
+    "From company_operation_status.csv, join by bmCode to the enterprises above, extract operating revenue amount and R&D investment intensity; 120 enterprises matched.",
+    "Group by province and compute each province's total operating revenue before elimination (all enterprises, including those with missing R&D intensity); 19 provinces in total.",
+    "Filter enterprises with non-null R&D intensity not below the national median 9.92 (enterprises with null R&D intensity are treated as below the median and eliminated); group by province and compute total operating revenue of surviving enterprises after elimination. 13 provinces have at least one surviving enterprise.",
+    "For each valid province, compute revenue retention ratio = (total operating revenue after elimination / total operating revenue before elimination) × 100. For Anhui Province, revenue retention ratio = 2727186878.01 / 2727186878.01 × 100 = 100.00%.",
+    "Sort by revenue retention ratio in descending order; the province with the highest ratio is Anhui Province, at 100.00%."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "National median R&D investment intensity": 9.92,
+    "Total enterprises in communication transmission equipment industry": 120,
+    "Anhui Province total operating revenue before elimination (CNY)": 2727186878.01,
+    "Anhui Province total operating revenue after elimination (CNY)": 2727186878.01,
+    "Anhui Province revenue retention ratio (%)": 100.0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/hard003.json b/assets/qa_gold/risk_assessment/hard003.json
new file mode 100644
index 0000000000000000000000000000000000000000..df68e95ae09d126c7877f511ade28bbf42ccff74
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/hard003.json
@@ -0,0 +1,38 @@
+{
+  "id": "hard003",
+  "question": "2022年，国家层面发布了数字经济发展相关规划，同时部分省份也配套出台了支持通信传输设备业的政策，形成中央-地方协同支持格局。在同时受益于上述两级政策支持、且通信传输设备业上市企业数量不低于8家的省份中（计算政府补贴依赖度时，仅纳入政府补贴金额、营业利润金额、营业收入金额三项同时有完整记录的企业，依赖度=省内企业政府补贴总额÷省内企业营业利润总额），哪个省份对政府补贴的财务依赖最为突出？进一步模拟：一旦该省所有通信传输设备企业同时遭遇50%补贴缩减，且利润等额受损，全省通信传输设备业的营业利润将萎缩多大比例？",
+  "guidelines": "依次回答省份名称、政府补贴依赖度和营业利润下降比例。依赖度和下降比例均以百分数表示，保留2位小数。如[\"广东省\", 25.33, 12.67]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "北京市",
+    19.52,
+    9.76
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"通信传输设备业\"的政策记录，找到70条相关政策，其中国务院政策1条、部委政策23条、地方政策46条。",
+    "从policy_resource.csv中读取46条地方政策全文，筛选出44条明确涉及特定省份的地方政策（排除2条标注为\"全国\"的地方政策），覆盖17个省份：广东省11条（含广东省促进工业经济平稳增长措施、深圳市战略性新兴产业发展意见、广东省数字经济工作要点等）、安徽省4条、山东省4条、上海市3条、重庆市3条、四川省3条、云南省2条、贵州省2条、湖南省2条、陕西省2条、北京市2条（含北京市数字经济全产业链开放发展行动方案等）、河南省1条、海南省1条、湖北省1条（湖北省推进北斗产业高质量发展若干措施）、新疆维吾尔自治区1条、福建省1条、江西省1条。",
+    "从company_profile.csv筛选行业=\"通信传输设备业\"的企业共120家，按省份统计企业数。拥有不少于8家企业的省份有6个：广东省38家、江苏省18家、湖北省11家、北京市10家、浙江省9家、四川省8家。",
+    "取同时满足\"有地方通信产业政策支持\"和\"企业总数>=8\"的省份交集，得到4个符合条件的省份：广东省、北京市、湖北省、四川省。江苏省和浙江省虽有足够企业但无地方通信传输设备业相关政策。",
+    "从company_operation_status.csv获取这4个省份通信传输设备业企业的政府奖励资金补贴、营业利润金额和营业收入金额数据，筛选三项数据均完整的企业。广东省36家、北京市9家、湖北省11家、四川省8家数据完整。",
+    "计算各省份的政府补贴依赖度=省内企业政府补贴总额/省内企业营业利润总额×100%。北京市：补贴总额8.63亿元/营业利润总额44.22亿元=19.52%；广东省：76.22亿元/434.77亿元=17.53%；湖北省：7.84亿元/46.27亿元=16.94%；四川省：0.61亿元/15.24亿元=4.02%。北京市补贴依赖度最高，达19.52%。",
+    "模拟北京市政府补贴削减50%的冲击：补贴削减额=8.63亿元×50%=4.32亿元，营业利润等额减少4.32亿元。营业利润下降比例=4.32/44.22×100%=9.76%。"
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "通信传输设备业相关政策总数（条）": 70,
+    "地方政策数（条）": 46,
+    "有地方政策的省份数（个）": 17,
+    "企业总数>=8的省份数（个）": 6,
+    "同时满足两项条件的省份数（个）": 4,
+    "北京市数据完整企业数（家）": 9,
+    "北京市政府补贴总额（亿元）": 8.63,
+    "北京市营业利润总额（亿元）": 44.22,
+    "北京市补贴依赖度（%）": 19.52,
+    "补贴削减额（亿元）": 4.32,
+    "营业利润下降比例（%）": 9.76
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/hard004.json b/assets/qa_gold/risk_assessment/hard004.json
new file mode 100644
index 0000000000000000000000000000000000000000..5a33a59e9be18987b6a5005f82f4e0927883a4ab
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/hard004.json
@@ -0,0 +1,28 @@
+{
+  "id": "hard004",
+  "question": "In 2022, suppose an industrial policy researcher is screening vulnerable provinces for the chemical raw materials and chemical products manufacturing industry: she defines \"government subsidy amount exceeding 5% of that province's industry total operating revenue\" as excessive subsidy dependence, and \"the single largest enterprise's operating revenue as a share of that province's industry total operating revenue above 40%\" as market structure imbalance. Only among provinces with total enterprises not less than 12 (operating profit margin = total operating profit / total operating revenue × 100%), among provinces that simultaneously trigger both alerts, what is the operating profit margin of the province with the weakest profitability?",
+  "guidelines": "Answer format: percentage value (2 decimal places). E.g. \"8.56\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "No relevant data found",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter records with industry=\"Chemical Raw Materials and Chemical Products Manufacturing\" and total enterprises>=12, extract province, total government reward and subsidy funds, total operating revenue amount, maximum operating revenue amount, and total operating profit amount fields; 7 provinces found.",
+    "Filter provinces where all required fields are non-null and total operating revenue amount>0; 7 provinces in total.",
+    "For each province, compute government subsidy to operating revenue ratio = (total government reward and subsidy funds / total operating revenue amount) × 100%, and determine high government subsidy dependence (>5%); 0 provinces.",
+    "For each province, compute largest enterprise revenue share = (maximum operating revenue amount / total operating revenue amount) × 100%, and determine high revenue concentration (>40%); 1 province.",
+    "Filter provinces that simultaneously satisfy both high government subsidy dependence and high revenue concentration (dual-risk provinces); 0 provinces.",
+    "For each dual-risk province, compute operating profit margin = (total operating profit amount / total operating revenue amount) × 100%.",
+    "Sort by operating profit margin in ascending order; the province with the lowest operating profit margin was not found; operating profit margin is 0.00%."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Number of provinces with complete required fields": 7,
+    "Number of high government subsidy dependence provinces": 0,
+    "Number of high revenue concentration provinces": 1,
+    "Number of dual-risk provinces": 0,
+    "Lowest operating profit margin (%)": 0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/hard005.json b/assets/qa_gold/risk_assessment/hard005.json
new file mode 100644
index 0000000000000000000000000000000000000000..ad69b9fc81491894023c1bc0f46ade83ab38978e
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/hard005.json
@@ -0,0 +1,26 @@
+{
+  "id": "hard005",
+  "question": "At the end of 2022, semiconductor industry analysts modeled 2023 scenarios: revenue across enterprises is projected to decline by 20%, while fixed operating costs (defined as operating revenue minus operating profit) rise by 15% from their baseline. Under these assumptions, compute the change in operating profit margin for each enterprise in Jiangsu Province, Guangdong Province, and Shanghai (original operating profit margin = operating profit / operating revenue × 100%; new operating profit margin = (new operating revenue − new operating cost) / new operating revenue × 100%; decline magnitude = original operating profit margin − new operating profit margin), then take the average decline per region—limited to enterprises with complete operating revenue and operating profit data and positive operating revenue. Among the three regions, which province/municipality has the largest average decline in enterprise operating profit margin?",
+  "guidelines": "Answer format: province or city name. E.g. \"Zhejiang Province\" or \"Beijing\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "Guangdong",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "From company_profile.csv, filter enterprise records with industry=\"Semiconductor industry\" and province in (\"Jiangsu\", \"Guangdong\", \"Shanghai\"), extract enterprise name, bmCode, and province fields; 104 enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by enterprise name and bmCode, extract operating revenue amount and operating profit amount; 104 enterprises successfully matched.",
+    "Filter enterprise records with non-null operating revenue and operating profit and operating revenue amount>0; 104 enterprises in total.",
+    "For each enterprise, compute original operating profit margin = (operating profit amount / operating revenue amount) × 100%, original operating cost = operating revenue amount − operating profit amount.",
+    "For each enterprise, compute new operating revenue = operating revenue amount × 0.8, new operating cost = original operating cost × 1.15.",
+    "For each enterprise, compute new operating profit = new operating revenue − new operating cost, new operating profit margin = (new operating profit / new operating revenue) × 100%, decline magnitude = original operating profit margin − new operating profit margin.",
+    "Group enterprises by province; for each province compute average decline magnitude, sort by average decline magnitude in descending order; Guangdong Province has the largest average decline."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Number of enterprises with complete data": 104,
+    "Province/municipality with largest average decline": "Guangdong Province",
+    "Largest average decline (percentage points)": 42.12
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/hard006.json b/assets/qa_gold/risk_assessment/hard006.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d2984848d86d0dc15e7b5add5e82b6a58d97eee
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/hard006.json
@@ -0,0 +1,28 @@
+{
+  "id": "hard006",
+  "question": "In 2022, in the comprehensive assessment of the semiconductor industry ecosystem, the scope is limited to provinces with total enterprises not less than 6. The assessment framework has three dimensions: ? \"Industry Scale Foundation\" (weight 30%): half the sum of each province's share of national semiconductor enterprise count and each province's share of national semiconductor operating revenue as the raw score for this dimension; ? \"Policy Ecosystem Density\" (weight 30%): each province's count of policies involving the Chinese terms for semiconductors, chips, or integrated circuits divided by the national total of such policies; ? \"Technological Accumulation Depth\" (weight 40%): each province's share of cumulative Chinese invention patent grants in the national total for the industry. Each dimension is min-max normalized (mapped to 0-100), then weighted to yield the comprehensive development potential index. Which province has the highest comprehensive score?",
+  "guidelines": "Answer format: province name. E.g. \"Guangdong Province\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "Shanghai",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "From national_industry_status.csv, filter records with industry=\"semiconductor industry\", and extract national benchmarks: total enterprises 172, total operating revenue amount 1494377201673.42, and total cumulative Chinese invention patent grants 54069.",
+    "From policy_release_status.csv, filter policy records whose policy name or full text involves the Chinese terms for semiconductors, chips, or integrated circuits, extract the province field, and count the number of semiconductor-related policies per province as well as the national total of such policies.",
+    "From regional_industry_status.csv, filter records with industry=\"semiconductor industry\" and total enterprises >= 6, extract the province, total enterprises, total operating revenue amount, and total cumulative Chinese invention patent grants fields; 6 provinces found.",
+    "Filter provinces where all required fields are non-null; 5 provinces in total.",
+    "For each province, compute raw industry scale foundation = (total enterprises / national total enterprises + total operating revenue amount / national total operating revenue amount) / 2.",
+    "For each province, compute raw policy ecosystem = that province's semiconductor-related policy count / national total semiconductor-related policies, and raw technological accumulation = total cumulative Chinese invention patent grants / national total cumulative Chinese invention patent grants.",
+    "Perform min-max normalization to 0–100 for industry scale foundation, policy ecosystem, and technological accumulation; for each province compute industry ecosystem resilience index = industry scale foundation score × 0.3 + policy ecosystem score × 0.3 + technological accumulation score × 0.4; sort by index in descending order; the highest-scoring province is Shanghai."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "National total enterprises": 172,
+    "National total semiconductor-related policies": 4,
+    "Number of qualifying provinces": 5,
+    "Highest-scoring province": "Shanghai",
+    "Highest score": 87.38
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/hard007.json b/assets/qa_gold/risk_assessment/hard007.json
new file mode 100644
index 0000000000000000000000000000000000000000..f18ce8e02ea6f55001cd1cabecca3d65b9aa1f24
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/hard007.json
@@ -0,0 +1,26 @@
+{
+  "id": "hard007",
+  "question": "In 2022, Guangdong Province's automobile manufacturing industry faces pressure from subsidy phase-out. A three-year phase-out path is set: compared with the actual subsidy amount in 2022, cut 20% in 2023, 40% in 2024, and 60% in 2025. For each Guangdong automobile manufacturing enterprise that has a recorded government reward/subsidy amount, its net profit loss in each year is exactly equal to that year's subsidy reduction; summing the losses for 2023–2025 gives the enterprise's cumulative net profit loss. Under this definition, how many hundred million yuan is the three-year total loss for the enterprise with the heaviest cumulative damage?",
+  "guidelines": "Answer format: a numeric value (2 decimal places). For example, \"5.67\" means a cumulative loss of 5.67 hundred million yuan. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": "20.53",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "From company_profile.csv, filter records with province = \"Guangdong Province\" and industry = \"Automobile Manufacturing\", extract enterprise name and bmCode fields, 27 enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by enterprise name and bmCode, extract the government reward/subsidy field, 27 enterprises matched.",
+    "Filter to records where government reward/subsidy is not null, 26 enterprises.",
+    "For each enterprise, compute 2023 subsidy reduction = government reward/subsidy × 0.2.",
+    "For each enterprise, compute 2024 subsidy reduction = government reward/subsidy × 0.4.",
+    "For each enterprise, compute 2025 subsidy reduction = government reward/subsidy × 0.6.",
+    "For each enterprise, compute cumulative net profit loss = 2023 subsidy reduction + 2024 subsidy reduction + 2025 subsidy reduction; sort by cumulative loss descending. The enterprise with the largest cumulative loss is Bei Qi Lu Yuan Xin Neng Yuan Qi Che Co., Ltd.; convert the amount from yuan to hundred million yuan, cumulative loss 20.53 hundred million yuan."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Enterprises with complete government subsidy data": 26,
+    "Maximum cumulative loss (yuan)": 2052595485.6,
+    "Maximum cumulative loss (hundred million yuan)": 20.53
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/hard008.json b/assets/qa_gold/risk_assessment/hard008.json
new file mode 100644
index 0000000000000000000000000000000000000000..29e42fe090c8a29dba2a14256f5a1e11f649723d
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/hard008.json
@@ -0,0 +1,25 @@
+{
+  "id": "hard008",
+  "question": "In 2022, suppose the chemical raw materials and chemical products manufacturing industry faces dual pressure in 2023: environmental protection investment must increase by an amount equivalent to 8% of operating revenue, and raw material price increases lead to a 10% increase in existing costs. If existing costs account for 75% of operating revenue, among enterprises in this industry in Jiangsu Province, how many enterprises will have negative new operating profit?",
+  "guidelines": "Answer format: integer. E.g. 15. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "From company_profile.csv, filter enterprise records with province=\"Jiangsu Province\" and industry=\"Chemical Raw Materials and Chemical Products Manufacturing\", extract enterprise name and bmCode fields; 55 enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by enterprise name and bmCode, extract operating revenue amount and operating profit amount; 55 enterprises successfully matched.",
+    "Filter enterprise records with non-null operating revenue and operating profit; 55 enterprises in total.",
+    "For each enterprise, compute original cost = operating revenue amount × 75%.",
+    "For each enterprise, compute new cost = original cost × 1.1.",
+    "For each enterprise, compute environmental protection investment increase = operating revenue amount × 8%.",
+    "For each enterprise, compute new operating profit = operating revenue amount − new cost − environmental protection investment increase; count enterprises with new operating profit < 0: 0 enterprises."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Number of enterprises with complete data": 55,
+    "Number of enterprises with negative new operating profit": 0
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/hard009.json b/assets/qa_gold/risk_assessment/hard009.json
new file mode 100644
index 0000000000000000000000000000000000000000..d9b4795e1978715013c526385189fd9be415fa51
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/hard009.json
@@ -0,0 +1,27 @@
+{
+  "id": "hard009",
+  "question": "In 2022, build a comprehensive industry-level risk resilience scoring system covering all manufacturing industries with total enterprises not less than 25 (excluding Finance, Real Estate, and Conglomerate industries). The scoring framework has three dimensions with different weights: Asset Return Efficiency (weight 40%), measured as the industry's average operating profit per enterprise divided by average total assets per enterprise; R&D Investment Intensity (weight 30%), measured as the industry's average R&D investment per enterprise divided by average operating revenue per enterprise; and Financial Desensitization to Policy Subsidies (weight 30%), measured as the industry's total operating profit divided by (total government subsidies + 1)—adding 1 to the denominator avoids division by zero when subsidies are zero. Each of the three raw indicators is min-max normalized across industries, scaled to a 0–100 score, then weighted by the above weights to yield the final score. Under this system, what is the comprehensive score of the top-ranked industry?",
+  "guidelines": "Answer format: numeric value (2 decimal places). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 65.11,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "From national_industry_status.csv, extract total enterprises for all industries; filter manufacturing industries with total enterprises>=25, excluding non-manufacturing industries such as \"Finance\", \"Real Estate\", and \"Conglomerate\"; 38 qualifying industries found.",
+    "For each qualifying industry, from national_industry_status.csv extract mean operating profit amount, mean total assets, mean R&D investment amount, mean operating revenue amount, total operating profit amount, and total government reward and subsidy funds fields.",
+    "Filter industry records where all required fields are non-null; 38 industries in total.",
+    "For each industry, compute raw financial robustness = mean operating profit amount / mean total assets, raw innovation capacity = mean R&D investment amount / mean operating revenue amount, raw policy independence = total operating profit amount / (total government reward and subsidy funds + 1).",
+    "Perform min-max normalization to 0–100 for the three dimensions (financial robustness, innovation capacity, policy independence); normalization formula = (value − min) / (max − min) × 100.",
+    "For each industry, compute comprehensive risk resilience index = financial robustness score × 0.4 + innovation capacity score × 0.3 + policy independence score × 0.3.",
+    "Sort by comprehensive risk resilience index in descending order; the top-ranked industry is Mining, with a score of 65.11."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Number of qualifying industries": 38,
+    "Number of industries with complete required fields": 38,
+    "Top-ranked industry": "Mining",
+    "Highest score": 65.11
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/hard010.json b/assets/qa_gold/risk_assessment/hard010.json
new file mode 100644
index 0000000000000000000000000000000000000000..04ae2e49e6d311b6bd3333b1257a2df3e2690a21
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/hard010.json
@@ -0,0 +1,41 @@
+{
+  "id": "hard010",
+  "question": "In 2022, to measure the resilience reserve of each province's pharmaceutical manufacturing industry under extreme external shocks, a research team established the following three-dimensional evaluation framework (including only provinces with total enterprises not less than 10): Dimension 1 \"Policy Shield Thickness\" (weight 35%): for each province, the number of local policies involving the Chinese terms for pharmaceuticals or bio-industry divided by the national total of such policies yields the raw relative policy density; Dimension 2 \"Independent Profitability Buffer\" (weight 35%): each province's total pharmaceutical manufacturing operating profit minus total government subsidies, divided by total assets, reflecting actual asset profitability excluding subsidies; Dimension 3 \"Technological Innovation Reserve Thickness\" (weight 30%): patent grants per unit of R&D investment by province, measuring R&D output efficiency. Each dimension is min-max normalized across provinces (mapped to 0-100), then weighted by the above weights to produce the comprehensive score. Among the top three ranked provinces, what is the average comprehensive score?",
+  "guidelines": "Answer format: numeric value (2 decimal places). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 69.4,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "From policy_release_status.csv, filter policy records whose policy name or full text involves the Chinese terms for pharmaceuticals or bio-industry, extract the province field, and count the number of pharmaceutical-related policies per province as well as the national total of such policies.",
+    "From regional_industry_status.csv, filter records with industry=\"Pharmaceutical Manufacturing\" and total enterprises>=10, extract province, total operating profit amount, total government reward and subsidy funds, total assets, total cumulative Chinese invention patent grants, and total R&D investment amount fields; 11 provinces found.",
+    "Filter provinces where all required fields are non-null, total assets>0, and total R&D investment amount>0; 10 provinces in total.",
+    "For each province, compute raw policy support intensity = that province's pharmaceutical-related policy count / national total of pharmaceutical-related policies.",
+    "For each province, compute raw financial buffer capacity = (total operating profit amount − total government reward and subsidy funds) / total assets, and raw innovation reserve = total cumulative Chinese invention patent grants / total R&D investment amount.",
+    "Perform min-max normalization to 0–100 for each of the three dimensions: policy support intensity, financial buffer capacity, and innovation reserve.",
+    "Sort by crisis response capability index in descending order; the top 3 provinces and their scores are: Guangdong Province (70.12), Shanghai (68.56), Beijing (69.23).",
+    "Compute average score = (70.12 + 68.56 + 69.23) / 3 = 69.40."
+  ],
+  "steps_num": 8,
+  "milestone": {
+    "National total of pharmaceutical-related policies": 21,
+    "Number of qualifying provinces": 10,
+    "Top 3 provinces and scores": [
+      {
+        "Province": "Guangdong Province",
+        "Score": 70.12
+      },
+      {
+        "Province": "Shanghai",
+        "Score": 68.56
+      },
+      {
+        "Province": "Beijing",
+        "Score": 69.23
+      }
+    ],
+    "Average score": 69.4
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/hard011.json b/assets/qa_gold/risk_assessment/hard011.json
new file mode 100644
index 0000000000000000000000000000000000000000..31b5823871840f5124d87cd6ea7ba7b2462712eb
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/hard011.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard011",
+  "question": "2022年，国家从消费品工业数字化转型、质量标准提升等多个维度出台了政策支持，部分省份也因势利导推出了地方消费电子产业相关政策方案。现聚焦受上述双重政策覆盖的省份（同时须满足：省内消费电子及电气业上市企业不少于5家，且全省总营业利润大于零）。在这批省份中，哪个省份的企业负债结构对利率最为敏感——即以全省总负债相对于总营业利润的倍率（总负债/总营业利润）来衡量，哪个省份的这一比值最高？在此基础上，若利率水平抬升2个百分点，以各企业总负债×2%估算其新增利息负担，全省新增利息成本总额将相当于该省消费电子及电气业总营业利润的百分之多少？",
+  "guidelines": "依次回答省份名称和额外利息成本占总营业利润的比例。比例以百分数表示，保留2位小数。如[\"广东省\", 23.36]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "江西省",
+    60.36
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"消费电子\"的政策记录，找到10条消费电子及电气业相关政策，其中4条为部委政策（国家层面），6条为地方政策。",
+    "经过对地方政策内容的分析，识别出台了地方消费电子产业发展政策的省份共5个：广东省（促进工业经济平稳增长措施，提出促进消费电子产品促销、推动8500家企业技术改造等）、江西省（打造全国新兴产业培育发展高地方案，提出电子信息产业规模突破12000亿元目标、新兴产业研发投入强度达2%）、四川省（承接制造业有序转移实施意见和成都制造业规划，提出承接电子信息配套产业、打造世界级电子信息产业集群）、重庆市（促进大中小企业融通发展方案）、海南省（激励企业上规模奖励资金细则）。",
+    "从company_profile.csv筛选行业=\"消费电子及电气业\"的企业，按省份统计，在上述省份中筛选企业总数>=5且总营业利润>0的省份：广东省150家、江西省5家、四川省8家符合条件。重庆市仅1家、海南省0家，不满足条件。",
+    "从company_operation_status.csv提取这3个省份消费电子及电气业企业的总负债和营业利润数据，计算各省份负债/营业利润倍数：江西省总负债166.30亿元/总营业利润5.51亿元=30.18倍、四川省总负债740.05亿元/总营业利润36.27亿元=20.40倍、广东省总负债16211.57亿元/总营业利润1387.98亿元=11.68倍。",
+    "江西省负债/营业利润倍数最高（30.18倍），利率敏感性风险最大。",
+    "计算江西省利率上升2个百分点的影响：额外利息成本=总负债166.30亿元×2%=3.33亿元，占总营业利润5.51亿元的比例=3.33/5.51×100%=60.36%。"
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "消费电子相关政策总数（条）": 10,
+    "国家层面政策数（条）": 4,
+    "地方政策涉及省份数（个）": 5,
+    "符合条件省份数（企业>=5且利润>0）": 3,
+    "江西省消费电子企业数（家）": 5,
+    "江西省总负债（亿元）": 166.3,
+    "江西省总营业利润（亿元）": 5.51,
+    "江西省负债/营业利润倍数": 30.18,
+    "江西省额外利息成本（亿元）": 3.33,
+    "江西省额外利息成本占总营业利润比例（%）": 60.36
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/hard012.json b/assets/qa_gold/risk_assessment/hard012.json
new file mode 100644
index 0000000000000000000000000000000000000000..58a7ab5967c133cf53c49fb349a658201231dd2e
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/hard012.json
@@ -0,0 +1,27 @@
+{
+  "id": "hard012",
+  "question": "In 2022, in the chemical raw materials and chemical products manufacturing industry, comprehensively assess each province's ability to withstand external shocks. What is the comprehensive score of the lowest-scoring province? (Assessment indicators: Profit Diversification 30%, Financial Safety Cushion 35%, Policy Buffer 35%. Profit Diversification = 1 − (largest enterprise operating profit / total operating profit) normalized score; Financial Safety Cushion = (total operating profit / total liabilities) normalized score; Policy Buffer = (total operating profit / (total government subsidies + 1)) normalized score; each dimension min-max normalized to 0–100; only provinces with total enterprises>=15 are included)",
+  "guidelines": "Answer format: numeric value (2 decimal places). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": 15.78,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter records with industry=\"Chemical Raw Materials and Chemical Products Manufacturing\" and total enterprises>=15, extract province, maximum operating profit amount, total operating profit amount, total liabilities, and total government reward and subsidy funds fields; 6 provinces found.",
+    "Filter provinces where all required fields are non-null, total operating profit amount>0, and total liabilities>0; 6 provinces in total.",
+    "For each province, compute raw profit diversification = 1 − (maximum operating profit amount / total operating profit amount).",
+    "For each province, compute raw financial safety cushion = total operating profit amount / total liabilities.",
+    "For each province, compute raw policy buffer = total operating profit amount / (total government reward and subsidy funds + 1).",
+    "Perform min-max normalization to 0–100 for profit diversification, financial safety cushion, and policy buffer.",
+    "For each province, compute regional shock resilience index = profit diversification score × 0.3 + financial safety cushion score × 0.35 + policy buffer score × 0.35; sort by index in ascending order; the lowest-scoring province is Shanghai, with a score of 15.78."
+  ],
+  "steps_num": 7,
+  "milestone": {
+    "Number of provinces with total enterprises>=15": 6,
+    "Number of provinces with complete required fields": 6,
+    "Lowest-scoring province": "Shanghai",
+    "Lowest score": 15.78
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/hard013.json b/assets/qa_gold/risk_assessment/hard013.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6f234b08ca49884ec82cca09163b739a2d35fd6
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/hard013.json
@@ -0,0 +1,31 @@
+{
+  "id": "hard013",
+  "question": "In 2022 regional data for the general purpose equipment manufacturing industry, for the province with the highest total government reward and subsidy funds: comprehensively rate its industrial competitiveness along two strategic dimensions. \"Industrial Upgrading Capacity\" combines the province's average rank across three sub-indicators (R&D investment growth rate, year-on-year change in Chinese patent applications, R&D personnel share); \"Industrial Base\" combines the province's average rank across three sub-indicators (total enterprises, total operating revenue, subsidy efficiency = total operating revenue / total government subsidies). The overall comprehensive performance rank is the average of the two dimension ranks. Answer only: Can this province's comprehensive industrial performance rank among the top 5 nationally?",
+  "guidelines": "Answer format: \"Yes\" or \"No\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "From regional_industry_status.csv, filter industry=\"General Purpose Equipment Manufacturing\", extract province and total government reward and subsidy funds fields; after filtering nulls, 16 provinces.",
+    "Sort by total subsidies in descending order; the province with the highest total government subsidies is Shanghai (2466523519.29 yuan).",
+    "From the same table, extract per-province mean year-on-year R&D investment change, mean R&D personnel share, total enterprises, total operating revenue amount, and total annual Chinese patent applications fields.",
+    "Compute per-province subsidy efficiency = total operating revenue amount / total government reward and subsidy funds (filter out zero denominator); 14 valid provinces.",
+    "Rank all provinces in descending order for each of the six indicators (R&D growth rate, annual patent applications, R&D personnel share, enterprise scale, revenue scale, subsidy efficiency); compute each province's rank on the six indicators.",
+    "Compute industrial upgrading capacity rank = (R&D growth rank + patent application rank + talent share rank) / 3; Shanghai's industrial upgrading capacity rank = 8.67.",
+    "Compute industrial base rank = (enterprise scale rank + revenue scale rank + subsidy efficiency rank) / 3; Shanghai's industrial base rank = 4.33.",
+    "Comprehensive performance rank = (industrial upgrading rank + industrial base rank) / 2; Shanghai's comprehensive performance rank value = 6.50.",
+    "Sort all provinces by comprehensive performance rank in ascending order; Shanghai's comprehensive rank is 6th, > 5; output \"No\"."
+  ],
+  "steps_num": 9,
+  "milestone": {
+    "Province with highest government subsidies": "Shanghai",
+    "Shanghai total government subsidies (yuan)": 2466523519.29,
+    "Shanghai industrial upgrading capacity rank": 8.67,
+    "Shanghai industrial base rank": 4.33,
+    "Shanghai comprehensive performance rank value": 6.5,
+    "Shanghai comprehensive rank": 6
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/medium001.json b/assets/qa_gold/risk_assessment/medium001.json
new file mode 100644
index 0000000000000000000000000000000000000000..718d9f0c7cd55bb4336dfed03a004667c00635eb
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/medium001.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium001",
+  "question": "Based on 2022 data, consider the following policy effect scenario: For provinces that have promulgated local industrial policies whose titles contain the keywords \"semiconductor\" or \"integrated circuit\", policy support will accelerate their semiconductor industry enterprises' R&D expansion pace to 2× the current growth rate over the next 3 years; for provinces that have not yet issued such policies, R&D growth remains unchanged. Using the median year-on-year change in enterprise R&D investment by province as the baseline growth rate, projected with 3-year compound growth, which province will rank first nationwide in total semiconductor industry R&D investment by 2025? What is the projected amount?",
+  "guidelines": "Answer format: [Province name, value (2 decimal places, unit: yuan)]. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": [
+    "Shanghai",
+    97732260069.03
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "From policy_resource.csv, filter policies whose titles contain \"semiconductor\" or \"integrated circuit\", 4 records. Extract the involved province list (deduplicated, excluding national policies), yielding 4 provinces: Guangdong, Zhejiang, Shanghai, Anhui.",
+    "From regional_industry_status.csv, filter records with industry = \"semiconductor industry\" and exclude provinces where either R&D investment amount or year-on-year R&D change is missing, 16 provinces: Guangdong Province, Beijing Municipality, Jiangsu Province, Shanghai Municipality, Zhejiang Province, Shandong Province, Sichuan Province, Anhui Province, Hong Kong SAR, Hunan Province, Hebei Province, Liaoning Province, Jilin Province, Xinjiang Uygur Autonomous Region, Shanxi Province.",
+    "Group by province, calculate total R&D investment and median year-on-year R&D change for each province, 16 provinces.",
+    "For each province, determine if it is a policy province: policy provinces use adjusted growth rate = median growth rate × 2; non-policy provinces use adjusted growth rate = median growth rate. The top-ranked province, Shanghai, is a policy province with adjusted growth rate 64.38%.",
+    "Calculate each province's projected 2025 total R&D investment = total R&D investment × (1 + adjusted growth rate/100)^3. Shanghai's projected 2025 R&D investment = 22003461800.09 × (1+64.38/100)^3 = 97732260069.03.",
+    "The top-ranked province is Shanghai, with projected 2025 total R&D investment of 97732260069.03 yuan."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Provinces with policy count": 4,
+    "Provinces with valid data count": 16,
+    "Shanghai 2022 total R&D investment (yuan)": 22003461800.09,
+    "Shanghai 2022 median YoY R&D change (%)": 32.19,
+    "Shanghai adjusted growth rate (%)": 64.38,
+    "Shanghai projected 2025 R&D investment (yuan)": 97732260069.03
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/medium002.json b/assets/qa_gold/risk_assessment/medium002.json
new file mode 100644
index 0000000000000000000000000000000000000000..65aa775a1cfb8ebfc870c6419b48033cc56b00a3
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/medium002.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium002",
+  "question": "In 2022, focusing on pharmaceutical manufacturing: among provinces included in the statistics (requiring total related enterprises in the province ≥ 10), if a province has R&D investment concentration CR3 greater than 60% and cumulative granted Chinese invention patent concentration CR3 also greater than 60%, classify that province as a high-risk \"R&D–patent dual head concentration\" province. How many provinces satisfy both dual-concentration conditions?",
+  "guidelines": "Answer format: an integer, e.g. \"3\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 3,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "From company_profile.csv, group enterprise records by province and count enterprises per province; filter to provinces with total enterprises ≥ 10, obtaining 11 provinces: Beijing Municipality, Shanghai Municipality, Jiangsu Province, Guangdong Province, Zhejiang Province, Shandong Province, Sichuan Province, Hubei Province, Hunan Province, Jilin Province, Hong Kong SAR.",
+    "From company_profile.csv, filter records in the above provinces with industry = \"Pharmaceutical Manufacturing\", extract enterprise name, bmCode, and province fields, 340 enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by enterprise name and bmCode, extract R&D investment amount and cumulative granted Chinese invention patents.",
+    "For each qualifying province, filter enterprises with non-null R&D investment amount, sort by R&D investment amount descending, take the top 3 enterprises, compute R&D concentration CR3 = (sum of top 3 enterprises' R&D / sum of all enterprises' R&D in the province) × 100%, and select provinces with R&D CR3 > 60%.",
+    "For each qualifying province, filter enterprises with non-null cumulative granted Chinese invention patents, sort by cumulative patents descending, take the top 3 enterprises, compute patent output concentration CR3 = (sum of top 3 enterprises' patents / sum of all enterprises' patents in the province) × 100%, and select provinces with patent CR3 > 60%.",
+    "Take the intersection of provinces with R&D CR3 > 60% and provinces with patent CR3 > 60%; count provinces satisfying both dual-concentration conditions: 3."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Provinces with total enterprises ≥ 10": 11,
+    "Provinces with R&D CR3 > 60%": [
+      "Hong Kong SAR",
+      "Jilin Province",
+      "Sichuan Province",
+      "Hubei Province"
+    ],
+    "Provinces with patent output CR3 > 60%": [
+      "Hong Kong SAR",
+      "Shanghai Municipality",
+      "Hubei Province",
+      "Sichuan Province"
+    ],
+    "Count of dual high-concentration risk provinces": 3
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/medium003.json b/assets/qa_gold/risk_assessment/medium003.json
new file mode 100644
index 0000000000000000000000000000000000000000..3982cbcd2b228b0220421b87a3f417144e6d5e2d
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/medium003.json
@@ -0,0 +1,27 @@
+{
+  "id": "medium003",
+  "question": "In 2022, assume all national R&D tax incentives are cancelled. Among manufacturing industries (only industries with total enterprises ≥ 20 and complete R&D investment and operating revenue data, positive revenue, excluding \"Financial Services\", \"Real Estate\", and \"Conglomerates\"), assume a corporate income tax rate of 25%; R&D tax benefits include 100% additional deduction. What is the sum of the declines (in percentage points) for the three industries with the largest average net profit margin decline, where net profit margin impact = R&D × 100% × 25% / operating revenue × 100%?",
+  "guidelines": "Answer format: a numeric value (2 decimal places). For example, 15.67 means the sum of the three industries' declines is 15.67 percentage points. If relevant data cannot be found, answer \"No relevant data found\".",
+  "answer": 601.08,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "From national_industry_status.csv, extract total enterprises for all industries; filter to manufacturing industries with total enterprises ≥ 20; exclude non-manufacturing industries such as \"Financial Services\", \"Real Estate\", and \"Conglomerates\", obtaining 41 qualifying industries. Identify 30 valid provinces: Guangdong Province, Zhejiang Province, Jiangsu Province, Beijing Municipality, Shanghai Municipality, Shandong Province, Hong Kong SAR, Fujian Province, Sichuan Province, Anhui Province, Hubei Province, Hunan Province, Henan Province, Liaoning Province, Hebei Province, Shaanxi Province, Tianjin Municipality, Chongqing Municipality, Xinjiang Uygur Autonomous Region, Jilin Province, Shanxi Province, Guangxi Zhuang Autonomous Region, Yunnan Province, Heilongjiang Province, Gansu Province, Guizhou Province, Inner Mongolia Autonomous Region, Hainan Province, Tibet Autonomous Region.",
+    "From company_profile.csv, filter all enterprises in the qualifying industries, extract enterprise name, bmCode, and industry fields; related enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by enterprise name and bmCode; keep records where R&D investment amount and operating revenue amount are both non-null and operating revenue amount > 0, 5416 enterprises.",
+    "For each enterprise, compute net profit margin impact = (R&D investment amount × 100% × 25%) / operating revenue amount × 100%.",
+    "For each industry, compute average enterprise net profit margin decline = Σ(enterprise net profit margin impact) / number of enterprises in that industry; average decline computed for 41 industries.",
+    "Sort by average net profit margin decline descending; take the top 3 industries (Pharmaceutical Manufacturing, Specialized Equipment Manufacturing, Information Transmission, Software and IT Services); sum their declines = 601.08 percentage points."
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "Qualifying industries count": 41,
+    "Qualifying provinces count": 30,
+    "Enterprises with complete data": 5416,
+    "Industries with computed average decline": 41,
+    "Sum of top-3 industry declines (percentage points)": 601.08
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/medium004.json b/assets/qa_gold/risk_assessment/medium004.json
new file mode 100644
index 0000000000000000000000000000000000000000..ceef7602b70dacfcc9e1c01a60c03e45db54f6f4
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/medium004.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium004",
+  "question": "2022年，《建材行业碳达峰实施方案》由多部委联合印发，各省也相继出台了碳达峰或节能减排地方行动方案，建材类企业由此面临双层合规压力。在同时处于上述两级政策约束之下的省份中，进一步筛选非金属矿物制品业样本：省内上市企业总数至少5家，且整体营业利润为正。对于符合条件的省份，若将碳排放合规成本设定为各企业运营成本（运营成本=营业收入-营业利润）的5%，并全部计入利润扣减项，那么，哪个省份的非金属矿物制品业总营业利润下降幅度最大？该省合规成本冲击后，总营业利润究竟会下降多少个百分点？",
+  "guidelines": "依次回答省份名称和营业利润下降百分比。百分比保留2位小数。如[\"湖南省\", 28.63]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "安徽省",
+    45.74
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"非金属矿物\"的政策记录，找到39条非金属矿物制品业相关政策。",
+    "在39条政策中，筛选policyClassification为\"地方政策\"且政策名称包含\"碳达峰\"或\"节能减排\"的地方政策，找到8条，涉及7个省份：四川省、宁夏回族自治区、安徽省、江西省、湖南省、贵州省、辽宁省。",
+    "从company_profile.csv筛选行业=\"非金属矿物制品业\"的企业，按省份统计，在上述7个省份中筛选有营业收入的企业总数>=5且总营业利润>0的省份，得到3个符合条件的省份：四川省（5家）、安徽省（5家）、湖南省（5家）。",
+    "从company_operation_status.csv提取这3个省份非金属矿物制品业企业的营业收入和营业利润数据，计算各省的运营成本（=营业收入-营业利润）和碳排放合规成本（=运营成本×5%），再计算营业利润下降百分比（=碳排放合规成本/总营业利润×100%）：四川省总营业利润11.78亿元，总运营成本19.30亿元，碳排放合规成本0.96亿元，下降8.19%；安徽省总营业利润263.50亿元，总运营成本2410.28亿元，碳排放合规成本120.51亿元，下降45.74%；湖南省总营业利润23.53亿元，总运营成本149.90亿元，碳排放合规成本7.49亿元，下降31.85%。",
+    "安徽省营业利润下降幅度45.74%最大，原因是安徽省非金属矿物制品业企业整体利润率较低（约9.86%），运营成本规模远大于利润，因此碳成本冲击的相对影响最严重。"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "非金属矿物制品业相关政策总数（条）": 39,
+    "国家碳达峰政策数（条）": 1,
+    "地方碳达峰/节能减排政策数（条）": 8,
+    "涉及省份数（个）": 7,
+    "符合条件省份数（企业>=5且总利润>0）": 3,
+    "安徽省企业数（家）": 5,
+    "安徽省总营业利润（亿元）": 263.5,
+    "安徽省总运营成本（亿元）": 2410.28,
+    "安徽省碳排放合规成本（亿元）": 120.51,
+    "安徽省营业利润下降百分比（%）": 45.74
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/medium005.json b/assets/qa_gold/risk_assessment/medium005.json
new file mode 100644
index 0000000000000000000000000000000000000000..cf09ba31857a1e72e3a30e5c68f0b1348ff0c72b
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/medium005.json
@@ -0,0 +1,32 @@
+{
+  "id": "medium005",
+  "question": "2022年，相关部委出台了促进钢铁工业高质量发展的指导意见，明确提出金属冶炼行业研发投入强度力争达到1.5%的政策目标。现聚焦以下范围：同时受到国家层面金属冶炼产业发展政策和地方金属冶炼和压延加工业相关政策覆盖的省份，且该省上市企业总数（以营业收入、营业利润和研发投入数据均不为空、营业收入大于零为准）不低于6家，同时全省总营业利润为正值。在此范围内，模拟地方政府出台强制合规要求——凡研发投入强度（研发投入÷营业收入）未达1.5%门槛的企业，须将研发投入强制补足至1.5%，补足部分直接计入成本并从营业利润中扣除。请问：在符合条件的省份中，哪个省份的企业需要补足的研发投入缺口总量最大？执行该合规要求后，该省金属冶炼和压延加工业的总营业利润预计下降多少个百分点？",
+  "guidelines": "依次回答省份名称和营业利润下降比例。下降比例以百分数表示，保留2位小数。如[\"山东省\", 3.75]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "江西省",
+    5.91
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"金属冶炼\"的政策记录，找到41条金属冶炼和压延加工业相关政策，其中地方政策32条，涉及17个省份。",
+    "从company_profile.csv筛选行业=\"金属冶炼和压延加工业\"的企业，按省份统计，在上述17个省份中筛选营业收入金额、营业利润金额和研发投入金额均不为空且营业收入金额>0的企业，保留企业数>=6且总营业利润>0的省份，得到5个符合条件的省份：云南省（7家）、安徽省（7家）、山东省（9家）、江西省（8家）、河南省（6家）。",
+    "从company_operation_status.csv获取这5个省份金属冶炼企业的营业收入金额和研发投入金额，计算每家企业的研发投入占比（研发投入金额/营业收入金额×100%），筛选研发投入占比<1.5%的企业，计算每家不达标企业的额外研发投入=营业收入金额×1.5%-实际研发投入金额，汇总得到各省份额外研发投入总额：江西省37.69亿元、山东省9.93亿元、云南省8.94亿元、河南省6.31亿元、安徽省0.39亿元。",
+    "江西省需要额外增加的研发投入总额最多（37.69亿元），其中3家不达标企业分别需增加约17.61亿元、17.38亿元和2.70亿元。",
+    "计算江西省营业利润下降比例=额外研发投入总额/总营业利润×100%=3768902884.89元/63806804479.63元×100%=5.91%。"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "地方金属冶炼相关政策数（条）": 32,
+    "涉及省份数（个）": 17,
+    "符合条件省份数（个）": 5,
+    "江西省有效企业数（家）": 8,
+    "江西省研发不达标企业数（家）": 3,
+    "江西省额外研发投入总额（亿元）": 37.69,
+    "江西省总营业利润（亿元）": 638.07,
+    "江西省营业利润下降比例（%）": 5.91
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/medium006.json b/assets/qa_gold/risk_assessment/medium006.json
new file mode 100644
index 0000000000000000000000000000000000000000..4411cc44e28f9de43fc92d2f5f73e9ffb339bd57
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/medium006.json
@@ -0,0 +1,35 @@
+{
+  "id": "medium006",
+  "question": "2022年，国家将橡胶制品业列入新污染物治理试点，同时也要求塑料制品行业加快推进绿色低碳转型。在同时具备国家层面新污染物治理政策约束和地方橡胶和塑料制品政策支持、且省内上市企业总数不低于5家、总营业利润为正的省份中，平均营业利润率（以总营业利润除以总营业收入的比值衡量）最低的是哪个省份？鉴于该省企业本身盈利空间有限，一旦面临新污染物合规落地——以各企业运营成本（=营业收入-营业利润）的3%估算额外环保支出，且这部分费用全额从营业利润中扣减——请问该省橡胶和塑料制品业的总营业利润将因此下降多大比例？",
+  "guidelines": "依次回答省份名称和总营业利润下降的百分比。百分比保留2位小数。如[\"山东省\", 56.12]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "上海市",
+    83.39
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段找到22条橡胶和塑料制品业相关政策，其中国家层面政策5条、地方政策17条。",
+    "分析17条地方政策覆盖的省份，共涉及12个省份。",
+    "从company_profile.csv筛选行业=\"橡胶和塑料制品业\"的企业共107家，按省份统计，在有地方政策覆盖的12个省份中筛选企业数>=5且总营业利润>0的省份，得到3个符合条件的省份：上海市（10家）、安徽省（8家）、山东省（5家）。",
+    "从company_operation_status.csv提取这3个省份橡胶和塑料制品业企业的营业收入金额和营业利润金额，计算各省平均营业利润率（总营业利润/总营业收入×100%）：上海市=7.49亿/215.69亿=3.47%、山东省=20.00亿/275.40亿=7.26%、安徽省=35.94亿/448.88亿=8.01%。上海市营业利润率最低。",
+    "计算上海市的新污染物合规成本冲击：各企业额外环保成本=各企业运营成本×3%，其中运营成本=营业收入-营业利润。上海市10家企业的额外环保成本总计=6.25亿元，总营业利润下降百分比=6.25/7.49×100%=83.39%。"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "橡胶和塑料制品业相关政策总数（条）": 22,
+    "国家层面政策数（条）": 5,
+    "地方政策覆盖省份数（个）": 12,
+    "橡胶和塑料制品业企业总数（家）": 107,
+    "符合条件省份数（个）": 3,
+    "上海市企业数（家）": 10,
+    "上海市总营业收入（亿元）": 215.69,
+    "上海市总营业利润（亿元）": 7.49,
+    "上海市营业利润率（%）": 3.47,
+    "上海市额外环保成本总计（亿元）": 6.25,
+    "上海市总营业利润下降百分比（%）": 83.39
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/medium007.json b/assets/qa_gold/risk_assessment/medium007.json
new file mode 100644
index 0000000000000000000000000000000000000000..af450aa1a664cdad2dbb154f683b580e20b79416
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/medium007.json
@@ -0,0 +1,38 @@
+{
+  "id": "medium007",
+  "question": "2022年，若要评估各省半导体行业的市场结构风险，需重点关注市场集中程度——集中度越高，单一龙头企业的经营波动对全省产业的冲击越大。请在满足以下两个条件的省份范围内作答：一是该省已出台半导体行业政策；二是省内半导体业上市企业总数不少于10家（仅统计营业收入不为空且大于零的企业）。基于赫芬达尔-赫希曼指数（HHI=各企业营业收入占省内总营业收入百分比的平方和），哪个省份市场集中度最高？在此基础上，进一步模拟：若该省营业收入体量最大的企业遭遇外部供应链冲击，营业收入收缩30%而省内其余企业保持不变，新的HHI指数会变为多少？",
+  "guidelines": "依次回答省份名称、原HHI指数和冲击后HHI指数，HHI指数保留2位小数。如[\"上海市\", 1285.07, 1223.31]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "浙江省",
+    1596.76,
+    1302.75
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"半导体\"的政策记录，找到44条半导体相关政策，其中全国性政策（province为\"全国\"）9条，地方政策35条，涉及15个省份。",
+    "从company_profile.csv筛选行业=\"半导体业\"的企业，按省份统计企业总数。在上述3个有集成电路专项政策的省份中，筛选企业总数不少于10家的省份：广东省54家、上海市27家、浙江省13家，均满足条件。",
+    "从company_operation_status.csv获取这3个省份半导体企业的营业收入金额，筛选营业收入金额不为空且大于0的企业。广东省54家、上海市27家、浙江省13家均全部满足条件。",
+    "计算各省份的HHI指数。HHI=Σ(企业营业收入/省内总营业收入×100)²。广东省：总营收2764.08亿元，HHI=680.27；上海市：总营收2474.39亿元，HHI=1285.07；浙江省：总营收255.75亿元，HHI=1596.76。浙江省HHI指数最高，表明市场集中度最高。",
+    "浙江省HHI最高，其营业收入排名第一的企业为华微创澜微电子公司，营收82.82亿元，市场份额占比32.38%。假设该企业因外部供应链冲击营收下降30%，新营收=82.82×0.70=57.98亿元。省内新总营收=255.75-82.82+57.98=230.90亿元。",
+    "重新计算冲击后浙江省各企业的市场份额并计算新HHI。龙头企业新份额=57.98/230.90×100=25.11%，其余12家企业份额相应调整。冲击后HHI=Σ(新份额)²=1302.75。HHI从1596.76下降至1302.75，下降293.01，虽然集中度有所降低，但仍处于中高集中度区间（HHI>1500为高集中度，1000-1500为中集中度），且省内总营收损失24.85亿元（占比9.72%），反映出浙江省半导体产业对龙头企业的高度依赖。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "半导体相关政策总数（条）": 44,
+    "地方政策数（条）": 35,
+    "有集成电路专项政策的省份数（个）": 3,
+    "广东省HHI": 680.27,
+    "上海市HHI": 1285.07,
+    "浙江省HHI（原）": 1596.76,
+    "浙江省龙头企业营收（亿元）": 82.82,
+    "浙江省龙头企业市场份额（%）": 32.38,
+    "冲击后龙头企业营收（亿元）": 57.98,
+    "冲击后浙江省总营收（亿元）": 230.9,
+    "冲击后浙江省HHI": 1302.75,
+    "浙江省总营收损失占比（%）": 9.72
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/medium008.json b/assets/qa_gold/risk_assessment/medium008.json
new file mode 100644
index 0000000000000000000000000000000000000000..8cb620008b9097a59860ec2e7007a1b53274230d
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/medium008.json
@@ -0,0 +1,42 @@
+{
+  "id": "medium008",
+  "question": "2022年，随着新能源汽车补贴政策进入退坡阶段，各省汽车制造业对政府补贴的依存程度成为衡量产业抗风险能力的关键变量。在已出台地方汽车产业专项扶持政策、且本省汽车制造业上市企业数量不低于10家的省份中（仅纳入政府奖励资金补贴、营业利润、营业收入三项数据均完整的企业，政府补贴依赖度定义为省内补贴总额与营业利润总额之比），哪个省份的补贴依赖度最高？如果对该省所有汽车制造企业同步实施50%补贴削减，且营业利润随之等额下降，则该省汽车制造业的整体营业利润将下降多少？",
+  "guidelines": "依次回答省份名称、补贴依赖度和营业利润下降比例。补贴依赖度和下降比例均以百分数表示，保留2位小数。如[\"广东省\", 45.20, 22.60]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "上海市",
+    66.53,
+    33.27
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"汽车\"的政策记录，找到69条汽车制造业相关政策，其中全国性政策16条，地方政策53条。",
+    "从policy_resource.csv中读取53条地方政策全文，分析哪些省份出台了汽车产业生产端专项扶持政策（含具体的产业补贴、补助、奖励或专项资金措施）。经分析，共有9个省份出台了此类政策：广东省5条（含广州市支持汽车及核心零部件产业稳链补链强链若干措施、广东省加快建设燃料电池汽车示范城市群行动计划等），上海市2条（含上海市制造业创新中心建设工程实施方案、充换电基础设施建设实施意见），山东省1条（青岛市加快新能源汽车产业高质量发展若干政策措施），江苏省1条（南京市新能源汽车换电模式应用试点实施方案），以及湖北省、湖南省、重庆市、辽宁省、吉林省各1条。",
+    "从company_profile.csv筛选行业=\"汽车制造业\"的企业共230家，按省份统计企业数。在上述9个有汽车产业扶持政策的省份中，筛选企业总数不少于10家的省份，得到4个符合条件的省份：上海市（22家）、山东省（18家）、广东省（27家）、江苏省（37家）。",
+    "从company_operation_status.csv获取这4个省份汽车制造业企业的政府奖励资金补贴、营业利润金额和营业收入金额数据，筛选三项数据均完整的企业。上海市20家、山东省18家、广东省26家、江苏省37家数据完整。",
+    "计算各省份的政府补贴依赖度=省内企业政府补贴总额/省内企业营业利润总额×100%。上海市：补贴总额50.91亿元/营业利润总额76.53亿元=66.53%；广东省：59.42亿元/453.02亿元=13.12%；江苏省：6.55亿元/60.89亿元=10.76%；山东省：15.85亿元/178.59亿元=8.87%。上海市补贴依赖度最高，达66.53%。",
+    "模拟上海市政府补贴削减50%的冲击：补贴削减额=50.91亿元×50%=25.46亿元，营业利润等额减少25.46亿元。原营业利润76.53亿元，新营业利润=76.53-25.46=51.07亿元，营业利润下降比例=25.46/76.53×100%=33.27%。上海市汽车制造业高补贴依赖度（66.53%）叠加低营业利润率（0.74%），使其在补贴退坡情景下面临的利润冲击最为显著。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "汽车相关政策总数（条）": 69,
+    "全国性政策数（条）": 16,
+    "地方政策数（条）": 53,
+    "有汽车产业扶持政策的省份数（个）": 9,
+    "符合条件省份数（企业>=10且有扶持政策）": 4,
+    "上海市数据完整企业数（家）": 20,
+    "上海市政府补贴总额（亿元）": 50.91,
+    "上海市营业利润总额（亿元）": 76.53,
+    "上海市营业收入总额（亿元）": 10279.21,
+    "上海市补贴依赖度（%）": 66.53,
+    "广东省补贴依赖度（%）": 13.12,
+    "江苏省补贴依赖度（%）": 10.76,
+    "山东省补贴依赖度（%）": 8.87,
+    "上海市补贴削减额（亿元）": 25.46,
+    "上海市新营业利润（亿元）": 51.07,
+    "上海市营业利润下降比例（%）": 33.27
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/medium009.json b/assets/qa_gold/risk_assessment/medium009.json
new file mode 100644
index 0000000000000000000000000000000000000000..bd4477f3f6e9c85f6e1b289dbfb76a9f5c903023
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/medium009.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium009",
+  "question": "2022年，纺织鞋服业受出口需求放缓与国内消费压力的双重影响，结构性风险进一步暴露。在同时受到国家层面纺织产业发展政策和地方纺织行业相关政策双重覆盖、且省内拥有正营业收入的上市企业总数不低于10家的省份范围内，哪个省份的纺织鞋服业市场结构最为集中——即按各企业营业收入占省内总营业收入的百分比计算市场份额后，所有市场份额百分比的平方加总（HHI）最大？进一步地，若该省第一大企业突遭外部需求骤降冲击，营业收入萎缩40%，而省内其余企业营收保持稳定，请重新计算此时的HHI指数。",
+  "guidelines": "依次回答省份名称和冲击后的HHI指数。HHI指数保留2位小数。如[\"广东省\", 1538.67]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "福建省",
+    2069.92
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"纺织\"的政策记录，找到21条纺织相关政策。地方层面共14条，涉及11个省份：上海市、广东省、福建省、山东省、四川省、辽宁省、广西壮族自治区、湖南省、陕西省、河北省、新疆维吾尔自治区。",
+    "从company_profile.csv筛选行业=\"纺织鞋服业\"的企业，关联company_operation_status.csv获取营业收入数据，仅保留营业收入大于0的企业，按省份统计企业数。在上述11个有地方政策覆盖的省份中，筛选企业数>=10的省份，得到4个符合条件的省份：上海市（13家）、山东省（11家）、广东省（33家）、福建省（17家）。",
+    "计算各省份的营业收入HHI指数（HHI=各企业营业收入占全省总营业收入百分比的平方和）：上海市HHI=2295.32、山东省HHI=2368.17、广东省HHI=1223.33、福建省HHI=2917.81。福建省HHI指数最高。",
+    "福建省纺织鞋服业总营业收入为1075.97亿元，营业收入排名第一的企业营业收入为536.51亿元，市场份额为49.86%。",
+    "模拟外部需求冲击：该企业营业收入下降40%后变为321.91亿元，福建省总营业收入变为861.36亿元。重新计算各企业市场份额并求HHI指数，冲击后HHI=2069.92。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "纺织相关政策总数（条）": 21,
+    "国家层面纺织政策数（条）": 7,
+    "地方层面纺织政策数（条）": 14,
+    "有地方政策覆盖的省份数（个）": 11,
+    "符合条件省份数（企业>=10且有地方政策）": 4,
+    "福建省纺织企业数（家）": 17,
+    "福建省总营业收入（亿元）": 1075.97,
+    "福建省原始HHI": 2917.81,
+    "福建省排名第一企业营业收入（亿元）": 536.51,
+    "冲击后排名第一企业营业收入（亿元）": 321.91,
+    "冲击后福建省总营业收入（亿元）": 861.36,
+    "冲击后福建省HHI": 2069.92
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/medium010.json b/assets/qa_gold/risk_assessment/medium010.json
new file mode 100644
index 0000000000000000000000000000000000000000..75d81ba42bec9a415fb6cb1ceb9d6d842e3f9fd4
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/medium010.json
@@ -0,0 +1,35 @@
+{
+  "id": "medium010",
+  "question": "2022年，食品饮料业上市企业在受到国家轻工业高质量发展政策指引的同时，部分省份也出台了针对性的地方食品产业政策。在同时满足两项条件的省份中（条件一：省内有国家轻工业政策与地方食品政策的双重覆盖；条件二：省内有营业收入记录的食品饮料业上市企业不少于6家），哪个省份企业的平均资产负债率水平最高，意味着其对外部融资依赖最深？在此前提下，若货币政策收紧、基准利率上升2个百分点，以各企业总负债额乘以2%估算新增利息负担，所有企业的额外利息成本加总后，这一总额相当于该省食品饮料业总营业利润的多少？",
+  "guidelines": "依次回答省份名称和额外利息成本占总营业利润的比例。比例以百分数表示，保留2位小数。如[\"湖北省\", 35.20]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "湖南省",
+    50.47
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"食品\"的政策记录，找到16条食品饮料业相关政策，其中国家层面（部委政策）3条，地方政策12条（不含标注为全国的1条）。",
+    "经过对地方政策内容的分析，提取出台了涉及食品饮料业的地方产业发展政策的省份共9个：甘肃省、河南省、云南省、四川省、湖南省、海南省、宁夏回族自治区、河北省、贵州省。",
+    "从company_profile.csv筛选行业为食品饮料业的企业，按省份统计，在上述9个有地方食品政策的省份中筛选有营业收入的企业数>=6的省份，得到4个符合条件的省份：河南省（8家）、四川省（9家）、湖南省（14家）、河北省（6家）。",
+    "从company_operation_status.csv提取这4个省份食品饮料业企业的资产负债率，计算各省份企业平均资产负债率：湖南省45.72%、四川省42.37%、河南省32.82%、河北省30.40%。湖南省平均资产负债率最高。",
+    "从company_operation_status.csv提取湖南省14家食品饮料业企业的总负债数据，计算总负债合计为415.40亿元，总营业利润合计为16.46亿元。",
+    "计算利率上升2个百分点带来的额外利息成本：额外利息成本=总负债×2%=415.40×2%=8.31亿元。额外利息成本占总营业利润的比例=8.31/16.46×100%=50.47%。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "食品饮料业相关政策总数（条）": 16,
+    "国家层面政策数（条）": 3,
+    "地方食品政策涉及省份数（个）": 9,
+    "符合条件省份数（企业>=6）": 4,
+    "湖南省食品饮料业企业数（家）": 14,
+    "湖南省平均资产负债率（%）": 45.72,
+    "湖南省食品饮料业总负债（亿元）": 415.4,
+    "湖南省食品饮料业总营业利润（亿元）": 16.46,
+    "额外利息成本（亿元）": 8.31,
+    "额外利息占总营业利润比例（%）": 50.47
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_gold/risk_assessment/medium011.json b/assets/qa_gold/risk_assessment/medium011.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f1f57d76cf30ffe632ac90526f83a9d2293b097
--- /dev/null
+++ b/assets/qa_gold/risk_assessment/medium011.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium011",
+  "question": "2022年，国家发布一系列采矿业相关产业政策，同时各省也出台了采矿业相关的产业发展政策。在双重政策覆盖之下，且有正营业收入的上市企业总数不低于6家、全省采矿业总营业利润大于零的省份中，哪个省份的采矿业市场份额分布最为集中（市场集中度以营业收入HHI衡量）？此外，鉴于采矿业对国际大宗商品价格高度敏感，若该省营业收入体量最大的采矿企业因国际市场剧烈波动遭遇营收下滑40%（其余企业营收保持不变），重新计算各企业市场份额后，全省采矿业的HHI指数将变动至多少？",
+  "guidelines": "依次回答省份名称和冲击后的HHI指数。HHI指数保留2位小数。如[\"内蒙古自治区\", 3200.50]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "answer": [
+    "新疆维吾尔自治区",
+    "4444.61"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  },
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"采矿\"的政策记录，找到27条采矿业相关政策。其中部委政策6条、地方政策21条。",
+    "分析21条地方采矿业相关政策，涉及17个省份。",
+    "从company_profile.csv筛选行业=\"采矿业\"的企业共143家，关联company_operation_status.csv获取运营数据。筛选营业收入>0且营业利润数据完整的企业，按省份统计。在上述17个有地方采矿业政策的省份中，筛选企业总数>=6家且总营业利润>0的省份，得到4个符合条件的省份：内蒙古自治区（8家）、山西省（8家）、新疆维吾尔自治区（6家）、河南省（8家）。",
+    "计算各省份采矿业营业收入HHI指数（HHI=各企业营业收入占省份总营业收入比例的百分比平方和）：内蒙古自治区HHI=3451.82，山西省HHI=2308.12，新疆维吾尔自治区HHI=4666.82，河南省HHI=3517.82。新疆维吾尔自治区HHI最高，为4666.82。",
+    "新疆维吾尔自治区6家采矿业企业营业收入分别为835.90亿元、594.09亿元、44.08亿元、20.15亿元、6.69亿元、1.97亿元，总营业收入1502.87亿元。排名第一企业占比55.62%。",
+    "模拟冲击：排名第一企业营业收入下降40%，从835.90亿元降至501.54亿元，其他企业不变。新总营业收入=1168.51亿元。重新计算各企业市场份额并求HHI：新HHI=(501.54/1168.51×100)²+(594.09/1168.51×100)²+(44.08/1168.51×100)²+(20.15/1168.51×100)²+(6.69/1168.51×100)²+(1.97/1168.51×100)²=4444.61。"
+  ],
+  "steps_num": 6,
+  "milestone": {
+    "采矿业相关政策总数（条）": 27,
+    "国家层面碳达峰绿色转型政策数（条）": 3,
+    "地方采矿业政策涉及省份数（个）": 17,
+    "符合条件省份数（>=6家企业且利润>0）": 4,
+    "新疆维吾尔自治区采矿业企业数（家）": 6,
+    "新疆维吾尔自治区总营业收入（亿元）": 1502.87,
+    "新疆维吾尔自治区原始HHI": 4666.82,
+    "排名第一企业营业收入（亿元）": 835.9,
+    "冲击后排名第一企业营业收入（亿元）": 501.54,
+    "冲击后总营业收入（亿元）": 1168.51,
+    "冲击后HHI": 4444.61
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/comprehensive_decision/easy001_result.json b/assets/qa_raw/comprehensive_decision/easy001_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..40498875271a783422e08ab3377e1e5d4f664707
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/easy001_result.json
@@ -0,0 +1,26 @@
+{
+  "id": "easy001",
+  "question": "In which province is the enterprise with the highest R&D investment ratio nationwide in 2022 located?",
+  "guidelines": "The answer must be \"yes\" or \"no\". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all records with year=2022 from company_operation_status.csv, and extract the enterprise name and R&D investment ratio fields.",
+    "Filter enterprise records with non-null R&D investment ratio, sort by R&D investment ratio in descending order, and identify the enterprise with the highest R&D investment ratio as \"Yaoshi Shenkang Medical Equipment Company\" with an R&D investment ratio of 934642.80%.",
+    "Look up the record for enterprise name \"Yaoshi Shenkang Medical Equipment Company\" in company_profile.csv, extract the province field, and obtain the province where the enterprise is located as \"Jiangsu Province\"."
+  ],
+  "steps_num": 3,
+  "evidence": [
+    "R&D investment ratio data for enterprises in 2022 was found in company_operation_status.csv.",
+    "Province information for each enterprise was found in company_profile.csv."
+  ],
+  "milestone": {
+    "Enterprise with highest R&D investment ratio": "Yaoshi Shenkang Medical Equipment Company",
+    "R&D investment ratio (%)": 934642.8,
+    "Province of location": "Jiangsu Province"
+  },
+  "answer": "Jiangsu Province",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/easy002_result.json b/assets/qa_raw/comprehensive_decision/easy002_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..6a351bfa538fce66c77f742e12c9e365d9de836d
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/easy002_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "easy002",
+  "question": "In 2022, nationwide, how is the chemical raw materials and chemical products manufacturing industry ranked by asset scale? Please list the top five provinces.",
+  "guidelines": "Answer format: [Province A, Province B, ...]. Output only province names, do not add any other explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all records with industry=\"Chemical Raw Materials and Chemical Products Manufacturing\" from regional_industry_status.csv, and extract the province and total assets fields.",
+    "Sort all provincial data by total assets in descending order to determine the asset scale ranking of each province. Extract the top five provinces and their total assets: Shandong Province (544109686731.85), Zhejiang Province (401644798867.80), Jiangsu Province (250743006622.33), Shanghai (230472528900.23), Guangdong Province (176926240169.03)."
+  ],
+  "steps_num": 2,
+  "milestone": {
+    "Total assets in Shandong Province (yuan)": 544109686731.85,
+    "Total assets in Zhejiang Province (yuan)": 401644798867.8,
+    "Total assets in Jiangsu Province (yuan)": 250743006622.33,
+    "Total assets in Shanghai (yuan)": 230472528900.23,
+    "Total assets in Guangdong Province (yuan)": 176926240169.03,
+    "Top five provinces": [
+      "Shandong Province",
+      "Zhejiang Province",
+      "Jiangsu Province",
+      "Shanghai",
+      "Guangdong Province"
+    ]
+  },
+  "answer": [
+    "Shandong Province",
+    "Zhejiang Province",
+    "Jiangsu Province",
+    "Shanghai",
+    "Guangdong Province"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/easy003_result.json b/assets/qa_raw/comprehensive_decision/easy003_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..48b2f17b636ac99b179670e11e8868b0448234fa
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/easy003_result.json
@@ -0,0 +1,23 @@
+{
+  "id": "easy003",
+  "question": "In 2022, in the Qilu region (Shandong Province), how many policies support Zhongbai Jinmao Chain Company in its industry?",
+  "guidelines": "The answer must be an exact number. Output only the number, do not add units, commas, or any textual explanation. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Search for records of \"Zhongbai Jinmao Chain Company\" in company_profile.csv, extract the industry field, and determine that its industry is \"Wholesale and Retail Trade\".",
+    "Filter from policy_release_status.csv for all policy records with province=\"Shandong Province\", and policies where the industry field contains \"Wholesale and Retail Trade\". Found 2 policies: 1 from \"Local Policy - Shandong Provincial People's Government General Office Policy Count\" and 1 from \"Local Policy - Shandong Provincial Development and Reform Commission Policy Count\".",
+    "Count the number of policies meeting the criteria, which is 2."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Zhongbai Jinmao Chain Company": "Wholesale and Retail Trade",
+    "Shandong Province Local Policy - Shandong Provincial People's Government General Office Policy Count": 1,
+    "Shandong Province Local Policy - Shandong Provincial Development and Reform Commission Policy Count": 1,
+    "Number of policies in Shandong Province related to Wholesale and Retail Trade": 2
+  },
+  "answer": 2,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/easy004_result.json b/assets/qa_raw/comprehensive_decision/easy004_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..67a28987aba1536ed146a137101b641aa9af4888
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/easy004_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "easy004",
+  "question": "In 2022, nationwide, what is the ranking of provinces by asset size in the Information Transmission, Software and Information Technology Services industry? Please list the top five provinces.",
+  "guidelines": "Answer format: [Province A, Province B, ...]. Output only the province ranking, do not add any other explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from regional_industry_status.csv all records with industry=\"Information Transmission, Software and Information Technology Services\", and extract the province and total assets fields.",
+    "Sort all provincial data in descending order by total assets to determine the asset size ranking of each province. Extract the top five provinces and their total assets: Beijing (10297490896006.5), Zhejiang Province (4115693929492.25), Guangdong Province (2262247330030.01), Shanghai (1282711003966.55), Jiangsu Province (177568006242.47)."
+  ],
+  "steps_num": 2,
+  "milestone": {
+    "Total assets of Beijing (CNY)": 10297490896006.5,
+    "Total assets of Zhejiang Province (CNY)": 4115693929492.25,
+    "Total assets of Guangdong Province (CNY)": 2262247330030.01,
+    "Total assets of Shanghai (CNY)": 1282711003966.55,
+    "Total assets of Jiangsu Province (CNY)": 177568006242.47,
+    "Top five provinces by ranking": [
+      "Beijing",
+      "Zhejiang Province",
+      "Guangdong Province",
+      "Shanghai",
+      "Jiangsu Province"
+    ]
+  },
+  "answer": [
+    "Beijing",
+    "Zhejiang Province",
+    "Guangdong Province",
+    "Shanghai",
+    "Jiangsu Province"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/easy005_result.json b/assets/qa_raw/comprehensive_decision/easy005_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..afcb9a31d09fa7a23cc2fc96b051c0130a2545e0
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/easy005_result.json
@@ -0,0 +1,37 @@
+{
+  "id": "easy005",
+  "question": "In 2022, nationwide, what is the ranking of provinces by profitability in the Real Estate industry? Please list the top five provinces.",
+  "guidelines": "Answer format: [Province A, Province B, ...]. Output only the province ranking, do not add any other explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from regional_industry_status.csv all records with industry=\"Real Estate\", extract province and total net profit amount fields, finding data for 34 provinces/regions in the Real Estate industry.",
+    "Filter province records where total net profit amount is not null, totaling 34 valid records. Sort all provincial data in descending order by total net profit amount to determine the profitability ranking of each province.",
+    "Extract the top five provinces and their total net profit amounts: Hong Kong SAR (86497465420.52 CNY), Guangdong Province (70559018502.22 CNY), Zhejiang Province (12297928184.06 CNY), Beijing (8975618268.55 CNY), Jilin Province (675521026.00 CNY)."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Total net profit amount of Hong Kong SAR (CNY)": 86497465420.52,
+    "Total net profit amount of Guangdong Province (CNY)": 70559018502.22,
+    "Total net profit amount of Zhejiang Province (CNY)": 12297928184.06,
+    "Total net profit amount of Beijing (CNY)": 8975618268.55,
+    "Total net profit amount of Jilin Province (CNY)": 675521026.0,
+    "Top five provinces by ranking": [
+      "Hong Kong SAR",
+      "Guangdong Province",
+      "Zhejiang Province",
+      "Beijing",
+      "Jilin Province"
+    ]
+  },
+  "answer": [
+    "Hong Kong SAR",
+    "Guangdong Province",
+    "Zhejiang Province",
+    "Beijing",
+    "Jilin Province"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/easy006_result.json b/assets/qa_raw/comprehensive_decision/easy006_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..271c08e777eee82a563ba6881c6b1f6e3f37f9a3
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/easy006_result.json
@@ -0,0 +1,21 @@
+{
+  "id": "easy006",
+  "question": "In 2022, what is Sichuan Province's national ranking by average R&D investment in the Information Transmission, Software and Information Technology Services industry?",
+  "guidelines": "The answer must be an exact number representing the ranking. Output only the number, do not add units, commas, or any textual explanation. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from regional_industry_status.csv all records with industry=\"Information Transmission, Software and Information Technology Services\", finding data for 34 provinces/regions.",
+    "Extract the \"mean R&D investment amount\" field for each province to obtain mean R&D investment data for all provinces. Sort all provinces by mean R&D investment amount in descending order to determine the R&D investment ranking of each province.",
+    "Locate Sichuan Province's position in the sorted list. Sichuan Province's mean R&D investment amount is 147274301.44 CNY, and its ranking is 10th."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Mean R&D investment amount of Sichuan Province (CNY)": 147274301.44,
+    "Sichuan Province ranking": 10
+  },
+  "answer": 10,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard001_result.json b/assets/qa_raw/comprehensive_decision/hard001_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f463123d133501ddd1a078ce71c993867c761140
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard001_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "hard001",
+  "question": "In 2022, a strategic consulting firm was commissioned by a provincial government to quantitatively rank the comprehensive attractiveness of pharmaceutical manufacturing across provinces, in order to identify priority target regions for attracting leading enterprises. The company designed a four-dimensional weighted scoring system: four original indicators—enterprise agglomeration level (weight 30%), R&D expenditure as a share of revenue (weight 30%), regional policy coverage intensity (weight 20%), and R&D human resource penetration rate (weight 20%)—were normalized (min-max) and then weighted to produce a composite score. Among these, agglomeration level is measured by the proportion of enterprises in each province to the national total in pharmaceutical manufacturing; policy intensity is measured by the ratio of relevant policy items in each province to the total number of relevant policies nationwide; and human resource penetration rate is the total number of R&D personnel in each province divided by total employees. What is the specific composite score value of the province with the highest weighted composite score after normalization across provinces?",
+  "guidelines": "The answer should be a numerical value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter records with industry=\"Pharmaceutical Manufacturing\" from regional_industry_status.csv, extract province, total enterprises, total R&D expenditure, total operating revenue, total R&D personnel, and total employees; 34 provincial records were found.",
+    "Filter records with industry=\"Pharmaceutical Manufacturing\" from national_industry_status.csv to obtain the national total of 449 pharmaceutical manufacturing enterprises.",
+    "Filter policy records from policy_release_status.csv where the industry field contains \"Pharmaceutical Manufacturing\"; 80 relevant policies were found across 22 provinces. Group by province to count policy numbers per province.",
+    "Calculate four original indicators for each province: industry agglomeration = total enterprises/449, R&D intensity = total R&D expenditure/total operating revenue, policy support = province policy count/80, talent density = total R&D personnel/total employees. Filter to 16 valid provinces with all four indicators non-null.",
+    "Apply min-max normalization to each of the four indicators across the 16 valid provinces: normalized value = (original value - min)/(max - min).",
+    "Calculate composite score = normalized industry agglomeration × 0.3 + normalized R&D intensity × 0.3 + normalized policy support × 0.2 + normalized talent density × 0.2.",
+    "Sort by composite score in descending order. Shanghai has the highest composite score, with industry agglomeration 0.1203, R&D intensity 0.2548, policy support 0.1375, talent density 0.1620, and composite score = 0.9160."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "34 provincial records for pharmaceutical manufacturing were found in regional_industry_status.csv.",
+    "National total of 449 pharmaceutical manufacturing enterprises was obtained from national_industry_status.csv.",
+    "80 pharmaceutical manufacturing-related policies were found in policy_release_status.csv, distributed across 22 provinces."
+  ],
+  "milestone": {
+    "National total pharmaceutical manufacturing enterprises": 449.0,
+    "National total pharmaceutical manufacturing-related policies": 80,
+    "Number of valid provinces": 16,
+    "Shanghai industry agglomeration": 0.1203,
+    "Shanghai R&D intensity": 0.2548,
+    "Shanghai policy support": 0.1375,
+    "Shanghai talent density": 0.162,
+    "Shanghai composite score": 0.916
+  },
+  "answer": 0.92,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard002_result.json b/assets/qa_raw/comprehensive_decision/hard002_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2187e74a75f08df4624aad2fcd279eb8d7ac3e18
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard002_result.json
@@ -0,0 +1,33 @@
+{
+  "id": "hard002",
+  "question": "In 2022, conduct a quantitative assessment of the investment value of the semiconductor industry across provinces. The evaluation framework requires incorporating three dimensions: first, industry scale (40% weight), measured by the inter-provincial rank percentile of total operating revenue in each province; second, profitability quality (30% weight), reflected by the inter-provincial rank percentile of operating profit margin (total operating profit divided by total operating revenue) in each province; third, technology output intensity (30% weight), measured by the inter-provincial rank percentile of the ratio of total patent applications to R&D expenditure (converted to 100 million yuan) in each province. The rank percentile for each indicator is calculated by sorting values from low to high, using the formula (rank - 1) / (total number of provinces - 1). Note that only provinces with complete data for all three indicators are included in the calculation. Under this weighted scoring system, what is the final score of the province ranked first in comprehensive investment value?",
+  "guidelines": "The answer should be a numerical value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter records with industry=\"Semiconductor Industry\" from regional_industry_status.csv, extract province, total operating revenue, total operating profit, and total R&D expenditure; 34 records were found.",
+    "Filter enterprises with industry=\"Semiconductor Industry\" from company_profile.csv, extract company name, bmCode, and province; 172 enterprises were found.",
+    "Join with company_operation_status.csv via bmCode to extract annual domestic invention patent applications. Filter 149 valid records with non-null values, group by province to sum annual domestic invention patent applications; 22 provinces have patent data.",
+    "Inner join regional data with enterprise-level patent summary by province, filter records with total operating revenue > 0 and total R&D expenditure > 0; 13 valid provinces. Calculate three original indicators: industry scale = total operating revenue, profitability = total operating profit / total operating revenue, innovation output = total annual domestic invention patent applications / total R&D expenditure (in 100 million yuan).",
+    "Rank each indicator by value from low to high (rank method = min), calculate rank percentile = (rank - 1) / (13 - 1).",
+    "Calculate investment value composite score = industry scale rank percentile × 0.4 + profitability rank percentile × 0.3 + innovation output rank percentile × 0.3.",
+    "Sort by composite score in descending order. Zhejiang Province has the highest composite score, with industry scale rank percentile 0.6667, profitability rank percentile 0.5000, innovation output rank percentile 0.8333, and composite score = 0.6667."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "34 provincial records for semiconductor industry were found in regional_industry_status.csv.",
+    "172 semiconductor industry enterprises were found in company_profile.csv.",
+    "Enterprise-level patent data was obtained from company_operation_status.csv, with 149 valid records."
+  ],
+  "milestone": {
+    "Number of valid provinces": 13,
+    "Zhejiang Province industry scale rank percentile": 0.6667,
+    "Zhejiang Province profitability rank percentile": 0.5,
+    "Zhejiang Province innovation output rank percentile": 0.8333,
+    "Zhejiang Province composite score": 0.6667
+  },
+  "answer": 0.67,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard003_result.json b/assets/qa_raw/comprehensive_decision/hard003_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c2203f00007bc03286b49adc422c8ad5a9786b6a
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard003_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "hard003",
+  "question": "In 2022, an automotive manufacturing enterprise commissioned a third-party institution to score and rate the industrial supporting capacity of each province before selecting a site for a new plant. The scoring rules are as follows: first, rank provinces in descending order by the number of government policies related to automotive manufacturing, and take the top five provinces by policy count as the candidate pool; then, within the candidate provinces, calculate the industrial supporting composite index, which is a weighted combination of three components—upstream and downstream supply chain density (weight 0.4), local labor reserve (weight 0.3), and government subsidy intensity per enterprise (weight 0.3). Supply chain density is defined as the ratio of total automotive manufacturing enterprises in the province to the national total in the industry; labor reserve is defined as the ratio of total industry employees in the province to the national total in the industry; subsidy intensity is defined as total government rewards and subsidies for automotive manufacturing in the province divided by the number of enterprises in the province (subsidy intensity must be normalized across all provinces before being used in the formula). Among the top five provinces by policy ranking, what is the composite index value of the province with the highest industrial supporting composite index?",
+  "guidelines": "Answer format: numerical value (4 decimal places). Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter policy records from policy_release_status.csv where industry field contains \"Automotive Manufacturing\"; 69 records found. Group by province to count policy numbers per province (excluding \"National\" level), sort by policy count descending; top 5 provinces by policy support are ['Guangdong Province', 'Shanghai', 'Hunan Province', 'Sichuan Province', 'Chongqing'].",
+    "Filter records with industry=\"Automotive Manufacturing\" from regional_industry_status.csv, extract province, total enterprises, total employees, and total government rewards and subsidies; 34 records found.",
+    "Filter records with industry=\"Automotive Manufacturing\" from national_industry_status.csv to obtain national total of 230 automotive manufacturing enterprises and 3,254,510 total employees.",
+    "Calculate three original indicators for each province: upstream-downstream enterprise density = total enterprises/230, talent reserve = total employees/3,254,510, subsidy intensity = total government rewards and subsidies/total enterprises.",
+    "Apply min-max normalization to subsidy intensity: normalized value = (subsidy intensity - 5468453.88)/(476294284.71 - 5468453.88).",
+    "Calculate industrial supporting composite index = upstream-downstream enterprise density × 0.4 + talent reserve × 0.3 + normalized subsidy intensity × 0.3.",
+    "Filter among top 5 policy provinces ['Guangdong Province', 'Shanghai', 'Hunan Province', 'Sichuan Province', 'Chongqing'] (Chongqing excluded due to missing data for composite index calculation). Guangdong Province has the highest composite index: upstream-downstream enterprise density 0.1174, talent reserve 0.4499, normalized subsidy intensity 0.4558, composite index = 0.3187."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "69 automotive manufacturing-related policies were found in policy_release_status.csv.",
+    "34 automotive manufacturing provincial records were found in regional_industry_status.csv.",
+    "National total of 230 automotive manufacturing enterprises and 3,254,510 employees were obtained from national_industry_status.csv."
+  ],
+  "milestone": {
+    "National total automotive manufacturing enterprises": 230.0,
+    "National total automotive manufacturing employees": 3254510.0,
+    "Top 5 provinces by policy": [
+      "Guangdong Province",
+      "Shanghai",
+      "Hunan Province",
+      "Sichuan Province",
+      "Chongqing"
+    ],
+    "Guangdong Province upstream-downstream enterprise density": 0.1174,
+    "Guangdong Province talent reserve": 0.4499,
+    "Guangdong Province normalized subsidy intensity": 0.4558,
+    "Guangdong Province composite index": 0.3187
+  },
+  "answer": 0.3187,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard004_result.json b/assets/qa_raw/comprehensive_decision/hard004_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..99dee1355e1bc81a259dd479ebfef55acd802859
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard004_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "hard004",
+  "question": "In 2022, a provincial development and reform commission, when reviewing the effectiveness of fiscal subsidies for the chemical raw materials and chemical products manufacturing industry, needed to identify enterprises with misallocated subsidy resources. Specifically, analysts must first define the scope: only examine enterprises located in provinces that have policy entries for \"Chemical Raw Materials and Chemical Products Manufacturing\" in the policy release status data; then use the industry-wide median of government subsidies and the median operating profit margin (profit margin = operating profit ÷ operating revenue × 100%) as dual thresholds to identify \"capital misallocation\" enterprises—those that simultaneously have \"subsidy amount above the industry median\" but \"profit margin below the industry median\". Among the valid enterprises in the policy-covered provinces, what is the proportion of capital misallocation enterprises as a percentage of total valid enterprises in those provinces (express the result as a percentage with 2 decimal places, without the % symbol)?",
+  "guidelines": "The answer should be a percentage value with 2 decimal places. Output only the number, without the % symbol or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter policy records from policy_release_status.csv where industry field contains \"Chemical Raw Materials and Chemical Products Manufacturing\"; 60 records found. Extract province field, remove nulls and exclude \"National\" level; 23 policy-covered provinces: ['Shanghai', 'Yunnan Province', 'Inner Mongolia Autonomous Region', 'Sichuan Province', 'Ningxia Hui Autonomous Region', 'Anhui Province', 'Shandong Province', 'Shanxi Province', 'Guangdong Province', 'Guangxi Zhuang Autonomous Region', 'Xinjiang Uygur Autonomous Region', 'Jiangxi Province', 'Hebei Province', 'Henan Province', 'Hainan Province', 'Hubei Province', 'Hunan Province', 'Gansu Province', 'Fujian Province', 'Guizhou Province', 'Liaoning Province', 'Shaanxi Province', 'Heilongjiang Province'].",
+    "Filter enterprise records with industry=\"Chemical Raw Materials and Chemical Products Manufacturing\" from company_profile.csv, extract company name, bmCode, and province; 364 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract government rewards and subsidies, operating profit, and operating revenue; 364 records after merge.",
+    "Filter valid samples with non-null values for government rewards and subsidies, operating profit, and operating revenue; 362 enterprises.",
+    "Calculate industry-wide median benchmarks for valid enterprises: operating profit margin = operating profit/operating revenue × 100%; median government subsidy is 10,019,029.08 yuan, median operating profit margin is 10.00%.",
+    "Filter valid enterprises whose province is in the policy-covered province list; 233 enterprises.",
+    "Among the 233 valid enterprises in policy-covered provinces, filter \"high subsidy, low output\" enterprises with government subsidy > 10,019,029.08 and operating profit margin < 10.00%; 54 enterprises, proportion = 54/233 × 100% = 23.18%."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "60 policies for chemical raw materials and chemical products manufacturing were found in policy_release_status.csv, covering 23 provinces (excluding \"National\" level).",
+    "364 chemical raw materials and chemical products manufacturing enterprises were found in company_profile.csv.",
+    "2022 government subsidy and operating data for 364 enterprises were obtained from company_operation_status.csv; 362 valid samples."
+  ],
+  "milestone": {
+    "Number of policy-covered provinces": 23,
+    "Total chemical enterprises": 364,
+    "Number of valid samples": 362,
+    "Median government subsidy (yuan)": 10019029.08,
+    "Median operating profit margin (%)": 10.0,
+    "Valid enterprises in policy-covered provinces": 233,
+    "High subsidy low output enterprises": 54,
+    "Proportion (%)": 23.18
+  },
+  "answer": 23.18,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard005_result.json b/assets/qa_raw/comprehensive_decision/hard005_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca32d13988ed43719b9f9e4b8c41a6b2e7d3503d
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard005_result.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard005",
+  "question": "In 2022, for the information transmission, software and information technology services industry, an industry research institute sought to obtain a policy-adjusted comprehensive innovation efficiency indicator by superimposing the incentive effect of local policy support on top of raw innovation efficiency. The calculation logic is as follows: first, exclude from enterprise microdata any samples with missing R&D expenditure or annual domestic invention patent grants; for the remaining valid enterprises, aggregate by province and calculate the ratio of total invention patent grants to total R&D expenditure (converted to 100 million yuan) for each province as the province's raw innovation efficiency benchmark; then use the proportion of policy items in that province out of all information technology policies as the policy support coefficient, and multiply the raw efficiency benchmark by (1 plus the policy support coefficient) to obtain the final policy-adjusted innovation efficiency. Among all provinces with data, what is the specific value of the province with the highest adjusted efficiency?",
+  "guidelines": "The answer should be a numerical value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter policy records from policy_release_status.csv where industry field contains \"Information Transmission, Software and Information Technology Services\"; 206 records found. Group by province to count policy numbers per province.",
+    "Filter enterprise records with industry=\"Information Transmission, Software and Information Technology Services\" from company_profile.csv, extract company name, bmCode, and province; 644 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract R&D expenditure and annual domestic invention patent grants; 644 records after merge.",
+    "Filter valid enterprises with non-null values for both R&D expenditure and annual domestic invention patent grants; 432 enterprises.",
+    "Group by province to sum R&D expenditure and annual domestic invention patent grants; 28 provinces have valid data.",
+    "Convert total R&D expenditure per province to 100 million yuan, calculate raw innovation efficiency = total patent grants / total R&D expenditure (100 million yuan). Merge with policy data, calculate policy support coefficient = province policy count / 206.",
+    "Calculate policy-adjusted innovation efficiency = raw innovation efficiency × (1 + policy support coefficient), sort by adjusted efficiency descending. Hong Kong Special Administrative Region has the highest: raw efficiency 63.7360, policy support coefficient 0.0000, adjusted efficiency = 63.7360."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "206 information technology-related policies were found in policy_release_status.csv.",
+    "644 information technology services enterprises were found in company_profile.csv.",
+    "2022 R&D and patent data for 644 enterprises were obtained from company_operation_status.csv; 432 valid enterprises."
+  ],
+  "milestone": {
+    "Total information technology-related policies": 206,
+    "Total information technology services enterprises": 644,
+    "Number of valid enterprises": 432,
+    "Number of valid provinces": 28,
+    "Hong Kong SAR raw innovation efficiency": 63.736,
+    "Hong Kong SAR policy support coefficient": 0.0,
+    "Hong Kong SAR adjusted innovation efficiency": 63.74
+  },
+  "answer": 63.74,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard006_result.json b/assets/qa_raw/comprehensive_decision/hard006_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac3c4b6f4ad53933d0263e485b41b4dd3926b0cf
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard006_result.json
@@ -0,0 +1,37 @@
+{
+  "id": "hard006",
+  "question": "In 2022, to measure the impact of different ownership backgrounds on the operating performance of specialized equipment manufacturing enterprises, an analysis team compared each enterprise's return on equity (ROE) level with the industry-wide return level in its province to calculate \"excess ROE\" as a relative performance indicator. Specifically: enterprise ROE is calculated as net profit divided by net assets (total assets minus total liabilities) multiplied by 100%; provincial industry benchmark ROE is extracted from provincial industry summary tables, calculated as total industry net profit divided by total industry net assets (total assets minus total liabilities) multiplied by 100%; each enterprise's excess ROE is the difference between its own ROE and its province's benchmark ROE. After grouping by ownership type, which ownership category has the highest mean excess ROE among enterprises? What is that mean value in percentage points?",
+  "guidelines": "The answer should be a percentage value with 2 decimal places. Output only the number, without the % symbol or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter enterprise records with industry=\"Specialized Equipment Manufacturing\" from company_profile.csv, extract company name, bmCode, ownership, and province; 447 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract net profit, total assets, total liabilities, and operating revenue; 447 records after merge.",
+    "Filter records with industry=\"Specialized Equipment Manufacturing\" from regional_industry_status.csv, calculate provincial industry average ROE = total net profit / (total assets - total liabilities) × 100%; 15 valid provinces.",
+    "Filter valid enterprises with total assets > total liabilities and non-null operating revenue; 444 enterprises.",
+    "Calculate net assets = total assets - total liabilities for each enterprise, then ROE = net profit / net assets × 100%.",
+    "Inner join enterprise data with provincial industry average ROE by province; 389 enterprises matched. Calculate excess ROE = enterprise ROE - provincial industry average ROE for each enterprise.",
+    "Group by ownership to calculate mean excess ROE for each ownership type; 6 ownership types. Collective enterprises (2 enterprises) have the highest mean excess ROE = 5.17%."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "447 specialized equipment manufacturing enterprises were found in company_profile.csv.",
+    "2022 financial data for 447 enterprises were obtained from company_operation_status.csv; 444 valid enterprises.",
+    "Industry summary data for 15 provinces in specialized equipment manufacturing were obtained from regional_industry_status.csv."
+  ],
+  "milestone": {
+    "Total specialized equipment manufacturing enterprises": 447,
+    "Number of valid enterprises": 444,
+    "Enterprises matched with provincial data": 389,
+    "Number of ownership types": 6,
+    "Collective enterprise count": 2,
+    "Collective enterprise mean excess ROE (%)": 5.17
+  },
+  "answer": [
+    "Collective Enterprise",
+    5.17
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard007_result.json b/assets/qa_raw/comprehensive_decision/hard007_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..88231261a8a20a7332ebcbf41f39aaf651199e73
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard007_result.json
@@ -0,0 +1,44 @@
+{
+  "id": "hard007",
+  "question": "In 2022, when a research institution was reviewing the implementation effectiveness of provincial industrial policies in the food and beverage industry, it found that although some provinces had issued many support policies, the profitability of enterprises within their jurisdictions was not ideal. To identify such \"policy-heavy, low-return\" provinces, the institution planned to analyze separately those provinces with a relatively large number of policies (including national-level policies, totaling 3 or more): sum the operating profit amounts of all food and beverage industry enterprises in these provinces and divide by the sum of operating revenue amounts to obtain the comprehensive operating profit margin for each province, then identify the province with the lowest profit margin. What is the profit margin value (as a percentage, rounded to two decimal places) for the province with the lowest comprehensive operating profit margin?",
+  "guidelines": "The answer should be a percentage value with 2 decimal places. Output only the number, without the % symbol or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter policy records with industry field containing \"Food and Beverage Industry\" from policy_release_status.csv, 16 records in total. Group by province field to count policy numbers per province; 4 national-level policies need to be added to each province. After adding national-level policy count to each province's count, filter provinces with policy count >= 3, totaling 9 provinces: ['Yunnan', 'Sichuan', 'Ningxia', 'Hebei', 'Henan', 'Hainan', 'Hunan', 'Gansu', 'Guizhou'].",
+    "Filter all enterprise records with industry=\"Food and Beverage Industry\" from company_profile.csv, extract company name, bmCode, and province fields; 247 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract operating profit amount and operating revenue amount fields.",
+    "Filter valid enterprises with non-null operating revenue amount; 247 enterprises in total.",
+    "Group by province field, aggregate the sum of operating profit amounts and sum of operating revenue amounts for each province.",
+    "Calculate comprehensive operating profit margin for each province = sum of operating profit amounts / sum of operating revenue amounts × 100%.",
+    "Among the 9 provinces with >=3 policies, sort by operating profit margin in ascending order; the province with the lowest operating profit margin is Hainan, with total operating profit of 78,834,917.61 yuan, total operating revenue of 13,274,274,000.99 yuan, and operating profit margin = 0.59%."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "16 food and beverage industry related policies were found in policy_release_status.csv.",
+    "247 food and beverage industry enterprises were found in company_profile.csv.",
+    "2022 operating profit and operating revenue data for 247 enterprises were obtained from company_operation_status.csv; 247 valid enterprises."
+  ],
+  "milestone": {
+    "Total food and beverage industry policies": 16,
+    "Provinces with >=3 policies": [
+      "Yunnan",
+      "Sichuan",
+      "Ningxia",
+      "Hebei",
+      "Henan",
+      "Hainan",
+      "Hunan",
+      "Gansu",
+      "Guizhou"
+    ],
+    "Number of valid enterprises": 247,
+    "Hainan total operating profit": 78834917.61,
+    "Hainan total operating revenue": 13274274000.99,
+    "Hainan operating profit margin (%)": 0.59
+  },
+  "answer": 0.59,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard008_result.json b/assets/qa_raw/comprehensive_decision/hard008_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..8fcc3562a4c99ca33bc4923f1237d53ddcf3f0f7
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard008_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "hard008",
+  "question": "In 2022, a private equity institution sought to identify high-quality provinces in the electricity, heat, gas and water production and supply industry that combine growth potential, market undervaluation, and innovation resilience. The screening logic has three layers: The first layer requires that the median year-over-year change in operating revenue of enterprises within the province be positive (>0%), to exclude regions where revenue is already shrinking; The second layer, based on the first layer results, further requires that the province's market valuation level be relatively low, i.e., the P/S ratio of all enterprises in the province must be lower than the median P/S ratio across all provinces nationwide (national median is calculated from the provincial P/S ratio series); The third layer adds an innovation requirement, i.e., the mean R&D investment ratio of enterprises in the province must be higher than the mean of all enterprises in the industry with R&D investment ratio records. How many provinces satisfy all three conditions simultaneously? ",
+  "guidelines": "The answer should be an integer. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all provincial records with industry=\"Electricity, Heat, Gas and Water Production and Supply\" from regional_industry_status.csv, extract province, median year-over-year change in operating revenue, total company market cap, total operating revenue amount, and number of enterprises; 34 records in total.",
+    "Filter enterprises with industry=\"Electricity, Heat, Gas and Water Production and Supply\" from company_profile.csv, extract company name, bmCode, and province fields; 189 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract R&D investment ratio field. 122 enterprises have non-null R&D investment ratio; national average R&D investment ratio is 1.1470%. Group by province to calculate average R&D investment ratio per province.",
+    "Filter 16 valid provinces with non-null and >0 total operating revenue amount; calculate P/S ratio per province = total company market cap / total operating revenue amount (converted to 100 million yuan); national median P/S ratio is 1.024358.",
+    "Filter high-growth provinces with median year-over-year change in operating revenue > 0%; 15 provinces in total.",
+    "Among high-growth provinces, filter low-valuation provinces with P/S ratio < national median 1.024358; 7 provinces in total.",
+    "Among high-growth, low-valuation provinces, further filter provinces where average enterprise R&D investment ratio > 1.1470%; 4 provinces ultimately satisfy all three conditions: ['Guangdong', 'Shanghai', 'Henan', 'Hebei']."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "34 provincial records in the electricity industry were found in regional_industry_status.csv.",
+    "189 electricity industry enterprises were found in company_profile.csv.",
+    "R&D investment ratio data for 189 enterprises were obtained from company_operation_status.csv; 122 valid enterprises."
+  ],
+  "milestone": {
+    "Number of valid provinces": 16,
+    "National median P/S ratio": 1.024358,
+    "National average enterprise R&D investment ratio (%)": 1.147,
+    "Number of high-growth provinces": 15,
+    "Number of high-growth, low-valuation provinces": 7,
+    "Number of provinces meeting all conditions": 4
+  },
+  "answer": 4,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard009_result.json b/assets/qa_raw/comprehensive_decision/hard009_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d5a47f9cf49c52376fd84e23c84cf0e517fd6350
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard009_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "hard009",
+  "question": "In 2022, an investment manager at a merger and acquisition fund was seeking \"high R&D, low valuation\" M&A targets in the textile, footwear and apparel industry, but the scope was limited to provinces covered by textile, footwear and apparel industry-related policies. The prerequisite for screening valid enterprises is: net profit amount strictly greater than zero, and both R&D investment ratio and company market cap fields have data records. On this basis, first use all valid enterprises in the industry as the benchmark population to calculate the median R&D investment ratio and the median P/E ratio respectively; then from the subset of valid enterprises located in policy-covered provinces, filter enterprises whose R&D investment ratio is higher than the industry median and whose P/E ratio is lower than the industry median. How many enterprises satisfy the above dual screening conditions? (P/E ratio = company market cap (100 million yuan) ÷ net profit amount (100 million yuan))",
+  "guidelines": "The answer should be an integer. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter policy records with industry field containing \"Textile, Footwear and Apparel\" from policy_release_status.csv, 21 records in total. Extract province field, remove nulls and exclude \"National\" level, to obtain 11 policy-covered provinces: ['Shanghai', 'Sichuan', 'Shandong', 'Guangdong', 'Guangxi', 'Xinjiang', 'Hebei', 'Hunan', 'Fujian', 'Liaoning', 'Shaanxi'].",
+    "Filter all enterprise records with industry=\"Textile, Footwear and Apparel\" from company_profile.csv, extract company name, bmCode, and province fields; 177 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract R&D investment ratio, net profit amount, and company market cap fields; 177 records after merge.",
+    "Filter valid enterprises with net profit amount > 0 and both R&D investment ratio and company market cap non-null; 81 enterprises in total.",
+    "Confirm data units: company market cap is in 100 million yuan, net profit amount is in yuan. Unify units for P/E calculation: P/E = company market cap (100 million yuan) ÷ (net profit amount (yuan) ÷ 100000000), i.e., P/E = company market cap (100 million yuan) ÷ net profit amount (100 million yuan).",
+    "Calculate industry-wide median benchmarks for valid enterprises: median R&D investment ratio is 2.8, median P/E = company market cap (100 million yuan) / net profit amount (100 million yuan) is 23.13.",
+    "Among valid enterprises, filter those whose province is in the policy-covered province list; 29 enterprises in total.",
+    "Among the 29 valid enterprises in policy-covered provinces, filter enterprises with R&D investment ratio > 2.8 and P/E < 23.13; 9 enterprises in total."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "21 textile, footwear and apparel industry related policies were found in policy_release_status.csv, covering 11 provinces (excluding \"National\" level).",
+    "177 textile, footwear and apparel industry enterprises were found in company_profile.csv.",
+    "2022 R&D and market cap data for 177 enterprises were obtained from company_operation_status.csv; 81 valid enterprises."
+  ],
+  "milestone": {
+    "Number of policy-covered provinces": 11,
+    "Total textile, footwear and apparel enterprises": 177,
+    "Number of valid enterprises": 81,
+    "Median R&D investment ratio": 2.8,
+    "Median P/E": 23.13,
+    "Valid enterprises in policy-covered provinces": 29,
+    "Number of enterprises meeting conditions": 9
+  },
+  "answer": 9,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard010_result.json b/assets/qa_raw/comprehensive_decision/hard010_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..779d3d031c56d5c3e4bfb09a9c7ccbaa6d3ce604
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard010_result.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard010",
+  "question": "In 2022, to quantify the comprehensive competitive strength of the construction industry across regions, an industry association constructed a provincial competitiveness index system. The index is composed of four weighted sub-dimensions: market size share of national total (weight 30%), asset operation efficiency i.e. operating profit to total assets ratio (weight 30%), technology accumulation level i.e. cumulative invention patent grants to number of enterprises in jurisdiction ratio (weight 20%), and talent structure i.e. R&D personnel as share of total employees (weight 20%). The four raw indicators are each min-max normalized across all valid provinces, then weighted and summed to obtain the final index. Only provinces with data records for all four indicators are included in the calculation. Finally, please calculate the index difference between the first-ranked province and the last-ranked province in the competitiveness index ranking (rounded to two decimal places).",
+  "guidelines": "The answer should be a numeric value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all provincial records with industry=\"Construction\" from regional_industry_status.csv, extract province, total operating revenue amount, total operating profit amount, total assets, total cumulative Chinese invention patent grants, number of enterprises, total R&D personnel count, and total employee count fields; 34 records in total.",
+    "Filter records with industry=\"Construction\" from national_industry_status.csv to obtain national construction industry total operating revenue amount of 12,683,425,500,139.00 yuan and total number of enterprises 148.",
+    "Filter enterprises with industry=\"Construction\" from company_profile.csv, 148 enterprises in total, for cross-validation of provincial-level data.",
+    "Calculate four raw indicators per province: scale index = total operating revenue amount / 12,683,425,500,139.00, efficiency index = total operating profit amount / total assets, innovation index = total cumulative Chinese invention patent grants / number of enterprises, talent index = total R&D personnel count / total employee count. Filter 16 valid provinces with all four indicators non-null.",
+    "Apply min-max normalization to each of the four indicators: normalized value = (raw value - min) / (max - min).",
+    "Calculate competitiveness index = normalized scale index × 0.3 + normalized efficiency index × 0.3 + normalized innovation index × 0.2 + normalized talent index × 0.2.",
+    "Sort by competitiveness index in descending order; the difference between first-ranked Beijing (0.7287) and last-ranked Liaoning (0.0142) = 0.7145."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "34 provincial records in the construction industry were found in regional_industry_status.csv.",
+    "National construction industry total operating revenue amount of 12,683,425,500,139.00 yuan and total number of enterprises 148 were obtained from national_industry_status.csv.",
+    "148 construction industry enterprises were found in company_profile.csv."
+  ],
+  "milestone": {
+    "Number of valid provinces": 16,
+    "National total construction enterprises": 148,
+    "First-ranked province": "Beijing",
+    "First-ranked score": 0.7287,
+    "Last-ranked province": "Liaoning",
+    "Last-ranked score": 0.0142,
+    "Difference": 0.7145
+  },
+  "answer": 0.71,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard011_result.json b/assets/qa_raw/comprehensive_decision/hard011_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e68623745868d2ff71aa416a4b51939111b5014
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard011_result.json
@@ -0,0 +1,55 @@
+{
+  "id": "hard011",
+  "question": "In 2022, a think tank was commissioned to study the impact of policy intervention on R&D behavior in the communication transmission equipment industry. The research design divides all enterprises with R&D investment ratio data records into two groups: one group from provinces that have appeared in policy release information with \"Communication Transmission Equipment\" related policy entries (\"National\" level entries do not count as provinces and are not included in either group); the other group from provinces that have never appeared in the above policy entries. After grouping, calculate the arithmetic mean of R&D investment ratio for each group respectively, then compute the difference between them (policy-covered provinces mean minus non-policy-covered provinces mean). This difference reflects the association between policy coverage and R&D intensity of communication transmission equipment enterprises within the jurisdiction. What is this difference in percentage points?",
+  "guidelines": "The answer should be a numeric value with 2 decimal places. A positive number indicates policy-covered provinces are higher; a negative number indicates non-policy-covered provinces are higher. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter policy records with industry field containing \"Communication Transmission Equipment\" from policy_release_status.csv, 70 records in total. Extract unique province values and exclude \"National\" to obtain 17 policy-covered provinces: ['Anhui', 'Shandong', 'Guangdong', 'Sichuan', 'Hubei', 'Fujian', 'Jiangxi', 'Chongqing', 'Hunan', 'Yunnan', 'Guizhou', 'Henan', 'Shaanxi', 'Hainan', 'Beijing', 'Shanghai', 'Xinjiang'].",
+    "Filter enterprise records with industry=\"Communication Transmission Equipment\" from company_profile.csv, extract company name, bmCode, and province fields; 120 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract R&D investment ratio field; 120 records after merge.",
+    "Filter valid enterprises with non-null R&D investment ratio; 117 enterprises in total.",
+    "Based on the policy-covered province list from step 1, divide valid enterprises into two groups: 86 enterprises in policy-covered provinces, 31 enterprises in non-policy-covered provinces.",
+    "Calculate mean R&D investment ratio for each group: policy-covered provinces average = 14.35%, non-policy-covered provinces average = 9.45%.",
+    "Calculate difference = policy-covered average R&D ratio - non-policy-covered average R&D ratio = 14.35 - 9.45 = 4.90 percentage points."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "70 communication transmission equipment industry related policies were found in policy_release_status.csv, covering 17 provinces.",
+    "120 communication transmission equipment industry enterprises were found in company_profile.csv.",
+    "R&D investment ratio data for 120 enterprises were obtained from company_operation_status.csv; 117 valid enterprises."
+  ],
+  "milestone": {
+    "Number of communication transmission equipment policies": 70,
+    "Policy-covered provinces": [
+      "Anhui",
+      "Shandong",
+      "Guangdong",
+      "Sichuan",
+      "Hubei",
+      "Fujian",
+      "Jiangxi",
+      "Chongqing",
+      "Hunan",
+      "Yunnan",
+      "Guizhou",
+      "Henan",
+      "Shaanxi",
+      "Hainan",
+      "Beijing",
+      "Shanghai",
+      "Xinjiang"
+    ],
+    "Number of communication transmission equipment enterprises": 120,
+    "Number of valid enterprises": 117,
+    "Enterprises in policy-covered provinces": 86,
+    "Enterprises in non-policy-covered provinces": 31,
+    "Policy-covered average R&D ratio (%)": 14.35,
+    "Non-policy-covered average R&D ratio (%)": 9.45,
+    "Difference (percentage points)": 4.9
+  },
+  "answer": 4.9,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard012_result.json b/assets/qa_raw/comprehensive_decision/hard012_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6380ab42c4420e2110d7c13a0cbf351f64838c6
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard012_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "hard012",
+  "question": "In 2022, an antitrust research team analyzed the provincial market structure of the metal smelting and rolling processing industry. To ensure statistical reliability, only provinces with operating revenue amount records and at least 5 enterprises in the industry within the jurisdiction were included. Among qualifying provinces, the Herfindahl-Hirschman Index (HHI) was used to measure market concentration in each province: calculate each enterprise's operating revenue as a share of total operating revenue of all valid enterprises in the province, sum the squares of these shares and multiply by 100% to obtain the province's HHI value. Higher HHI indicates more concentrated markets and greater monopoly risk. After identifying the province with the highest HHI, extract the province's total operating profit amount and total operating revenue amount from provincial industry summary data, and calculate the corresponding operating profit margin. What is the operating profit margin of the province with the highest HHI? ",
+  "guidelines": "The answer should be a percentage value with 2 decimal places. Output only the number, without the % symbol or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter enterprise records with industry=\"Metal Smelting and Rolling Processing\" from company_profile.csv, extract company name, bmCode, and province fields; 145 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract operating revenue amount field; 145 records after merge.",
+    "Filter 111 enterprises with non-null operating revenue amount; group by province to count enterprises per province; retain 13 provinces with enterprise count >= 5: ['Shanghai', 'Yunnan', 'Beijing', 'Sichuan', 'Anhui', 'Shandong', 'Guangdong', 'Jiangsu', 'Jiangxi', 'Henan', 'Zhejiang', 'Liaoning', 'Hong Kong'].",
+    "Within each valid province, calculate each enterprise's market share = enterprise operating revenue amount / sum of operating revenue amounts of all enterprises in the province.",
+    "Calculate Herfindahl-Hirschman Index (HHI) for each province = sum of squares of enterprise market shares × 100%; sort by HHI in descending order.",
+    "Among qualifying provinces, using the same valid enterprise sample as for HHI calculation (non-null operating revenue and province enterprise count ≥ 5), aggregate total operating profit amount and total operating revenue amount per province; calculate operating profit margin per province = sum of operating profit amounts / sum of operating revenue amounts × 100%.",
+    "The province with the highest HHI is Shanghai, HHI = 88.47. Total operating profit of valid enterprises in this province is 16,186,839,594.21 yuan, total operating revenue is 391,233,407,230.44 yuan; operating profit margin = 4.14%."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "145 metal smelting and rolling processing enterprises were found in company_profile.csv.",
+    "Operating revenue amount data for 145 enterprises were obtained from company_operation_status.csv; 111 valid enterprises with non-null operating revenue.",
+    "Total operating profit amount and total operating revenue amount data for metal smelting and rolling processing by province were obtained from regional_industry_status.csv."
+  ],
+  "milestone": {
+    "Number of metal smelting and rolling processing enterprises": 145,
+    "Enterprises with non-null operating revenue": 111,
+    "Number of valid provinces (enterprises >= 5)": 13,
+    "Province with highest HHI": "Shanghai",
+    "HHI value": 88.47,
+    "Shanghai operating profit margin (%)": 4.14
+  },
+  "answer": 4.14,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard013_result.json b/assets/qa_raw/comprehensive_decision/hard013_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c9f0afc58c341a31063da8e98a41f1a7f161aa3e
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard013_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "hard013",
+  "question": "In 2022, a provincial industry and information department sought to evaluate the government subsidy utilization efficiency of enterprises of different sizes in the rubber and plastic products industry. After dividing enterprises into three groups by total assets—large (top 1/3 rounded up), medium (middle 1/3 rounded up), and small (bottom 1/3)—which enterprise size group has the highest subsidy utilization efficiency? What is its efficiency value? ",
+  "guidelines": "The answer should be a numeric value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all enterprise records with industry=\"Rubber and Plastic Products\" from company_profile.csv, extract company name, bmCode, and province fields; 107 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract total assets, operating revenue amount, and government reward funds and subsidies fields; 107 records after merge.",
+    "Filter records with industry=\"Rubber and Plastic Products\" from national_industry_status.csv to obtain national benchmark data: total enterprises 107, total operating revenue 313,571,405,678.69 yuan, total government subsidies 1,966,394,638.00 yuan, total assets 476,387,663,666.96 yuan.",
+    "Filter valid enterprises with all three fields (total assets, operating revenue amount, government reward funds and subsidies) non-null and government reward funds and subsidies > 0; 107 enterprises in total.",
+    "Sort by total assets in descending order; divide 107 enterprises into three groups: large enterprise group (top 36, highest 1/3 by total assets), medium enterprise group (middle 36), small enterprise group (bottom 35, lowest 1/3 by total assets).",
+    "Calculate subsidy utilization efficiency per group = sum of operating revenue amounts of enterprises in the group / sum of government reward funds and subsidies of enterprises in the group. Small group efficiency = 95.58, medium group efficiency = 161.13, large group efficiency = 166.33.",
+    "The group with the highest subsidy utilization efficiency is the large enterprise group, efficiency value = 166.33."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "107 rubber and plastic products enterprises were found in company_profile.csv.",
+    "2022 total assets, operating revenue, and subsidy data for 107 enterprises were obtained from company_operation_status.csv; 107 valid enterprises.",
+    "National aggregate benchmark data for rubber and plastic products industry were obtained from national_industry_status.csv; total enterprises 107."
+  ],
+  "milestone": {
+    "National total rubber and plastic enterprises": 107,
+    "Number of valid enterprises": 107,
+    "Small group enterprise count": 35,
+    "Medium group enterprise count": 36,
+    "Large group enterprise count": 36,
+    "Small group efficiency": 95.58,
+    "Medium group efficiency": 161.13,
+    "Large group efficiency": 166.33,
+    "Highest efficiency group": "Large"
+  },
+  "answer": [
+    "Large",
+    166.33
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard014_result.json b/assets/qa_raw/comprehensive_decision/hard014_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..cf0372c4defc551ab30bf7b2b493d53b89e6008e
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard014_result.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard014",
+  "question": "In 2022, a technology innovation fund evaluated the \"R&D-patent conversion\" full-chain efficiency of the consumer electronics and electrical industry across provinces, seeking to identify the province with optimal conversion efficiency (only provinces with valid enterprise count >= 3 are included). What is the R&D-patent conversion efficiency value of that province? (R&D-patent conversion efficiency = sum of annual Chinese invention patent grants / sum of annual Chinese invention patent applications × R&D output density; R&D output density = sum of annual Chinese invention patent applications / sum of R&D investment amount (100 million yuan))",
+  "guidelines": "The answer should be a numeric value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all enterprise records with industry=\"Consumer Electronics and Electrical\" from company_profile.csv, extract company name, bmCode, and province fields; 358 enterprises found.",
+    "Join with company_operation_status.csv via bmCode to extract annual Chinese invention patent applications, annual Chinese invention patent grants, and R&D investment amount fields; 358 records after merge.",
+    "Filter records with industry=\"Consumer Electronics and Electrical\" from national_industry_status.csv to obtain national benchmark data: total enterprises 358, total R&D investment amount 245,156,000,000.00 yuan, total annual invention patent applications 63,940, total annual invention patent grants 43,780.",
+    "Filter 266 valid enterprises with all three fields (annual Chinese invention patent applications, annual Chinese invention patent grants, R&D investment amount) non-null; retain 13 provinces with valid enterprise count >= 3 after grouping by province.",
+    "Aggregate by province: sum of annual Chinese invention patent applications, sum of annual Chinese invention patent grants, and sum of R&D investment amount.",
+    "Calculate R&D output density per province = sum of annual Chinese invention patent applications / sum of R&D investment amount (100 million yuan); conversion efficiency = (sum of annual Chinese invention patent grants / sum of annual Chinese invention patent applications) × R&D output density.",
+    "Sort by conversion efficiency in descending order; the province with the highest conversion efficiency is Shandong, with 24,694 patent applications, 15,143 patent grants, R&D output density 77.1105, conversion efficiency = 47.29."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "358 consumer electronics and electrical enterprises were found in company_profile.csv.",
+    "2022 patent and R&D data for 358 enterprises were obtained from company_operation_status.csv; 266 valid enterprises.",
+    "National aggregate benchmark data for consumer electronics and electrical industry were obtained from national_industry_status.csv; total enterprises 358."
+  ],
+  "milestone": {
+    "National total consumer electronics and electrical enterprises": 358,
+    "Number of valid enterprises": 266,
+    "Number of valid provinces": 13,
+    "Shandong patent applications": 24694.0,
+    "Shandong patent grants": 15143.0,
+    "Shandong R&D output density": 77.1105,
+    "Shandong conversion efficiency": 47.29
+  },
+  "answer": 47.29,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard015_result.json b/assets/qa_raw/comprehensive_decision/hard015_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..edfe045a2908716138fb637be3e70387839b4f2f
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard015_result.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard015",
+  "question": "In 2022, a provincial government evaluated the comprehensive financial health of real estate enterprises to decide which provinces (where the province has an effective enterprise count >= 3) should face strengthened risk supervision for real estate firms. What is the health score of the province with the lowest financial health? (Financial health = Profitability score × 0.4 + Solvency score × 0.3 + Growth capability score × 0.3; Profitability is measured by the average net profit margin of enterprises in that province, where net profit margin = net profit amount / operating revenue amount; Solvency is measured as 1 − the arithmetic mean of enterprises' asset-liability ratio in that province / 100; Growth capability is measured as the median of enterprises' year-over-year change in operating revenue in that province / 100; each indicator is min-max normalized across all valid provinces before being substituted into the formula.)",
+  "guidelines": "The answer should be a numerical value with 2 decimal places. Output only the number without units or text explanation. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter records with industry=\"房地产业\" from company_profile.csv; extract enterprise name, bmCode, and province — 295 enterprises.",
+    "Join with company_operation_status.csv on enterprise name and bmCode for year=2022; extract net profit amount, operating revenue amount, asset-liability ratio, and year-over-year change in operating revenue — 295 rows after merge.",
+    "Keep records with operating revenue amount > 0 and non-null net profit amount and asset-liability ratio — 295 rows; do not additionally exclude rows by asset-liability ratio; all participate in within-province arithmetic means and subsequent normalization.",
+    "Count enterprises by province; define valid provinces as those with effective enterprise count >= 3; in this data there are 17 valid provinces, and growth capability (median YoY operating revenue change) can be computed for all of them.",
+    "On those 17 valid provinces only, compute three raw indicators per province: Profitability = arithmetic mean over enterprises of (net profit amount / operating revenue amount); Solvency = 1 − arithmetic mean of asset-liability ratio / 100; Growth capability = median of year-over-year change in operating revenue / 100.",
+    "Apply min-max normalization to profitability, solvency, and growth capability separately across all 17 valid provinces (i.e. \"all valid provinces\" means this set only, excluding provinces with only 1–2 enterprises).",
+    "Compute financial health = normalized profitability × 0.4 + normalized solvency × 0.3 + normalized growth capability × 0.3.",
+    "Sort the 17 valid provinces by financial health ascending; the lowest is Beijing: raw profitability ≈ −1.0774, raw solvency ≈ −21.9888, raw growth capability ≈ −0.1341, health score ≈ 0.0713, rounded to two decimals as 0.07."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "From company_profile.csv, there are 295 real estate industry enterprises.",
+    "From company_operation_status.csv, obtain 2022 operating data for those enterprises merged with profile — 295 rows; rows with revenue > 0 and non-null net profit and asset-liability ratio number 295, all used for per-province indicator calculation."
+  ],
+  "milestone": {
+    "real_estate_enterprises_profile_merged_2022": 295,
+    "enterprise_records_province_aggregation_no_al_ratio_exclusion": 295,
+    "valid_provinces_count_enterprises_ge_3": 17,
+    "Beijing_raw_profitability": -1.0774,
+    "Beijing_raw_solvency": -21.9888,
+    "Beijing_raw_growth_capability": -0.1341,
+    "Beijing_financial_health": 0.0713
+  },
+  "answer": 0.07,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard016_result.json b/assets/qa_raw/comprehensive_decision/hard016_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4309367af2f0fb26e5013e430ca2a162c939875f
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard016_result.json
@@ -0,0 +1,38 @@
+{
+  "id": "hard016",
+  "question": "In 2022, an institutional investor plans to build an equity portfolio among Haishan Chang Industrial Equipment Company, Zhongbai Jinmao Chain Company, and Sansan Dateng Heavy Industry Company. The total portfolio weight must equal 1, the portfolio-weighted asset-liability ratio must equal exactly 45%, and the portfolio-weighted year-on-year operating revenue change must equal exactly 0%. Based on these three companies' 2022 operating data, find their portfolio weights and compute the portfolio-weighted ROE. Note: each company's asset-liability ratio is computed as total liabilities ÷ total assets × 100%.",
+  "guidelines": "Answer format: weight of Haishan Chang Industrial Equipment Company, weight of Zhongbai Jinmao Chain Company, weight of Sansan Dateng Heavy Industry Company, weighted ROE. The first three weights to four decimal places; weighted ROE to three decimal places. Output numbers and commas only, with no explanatory text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "steps": [
+    "From company_operation_status.csv, filter by bmCode for Haishan Chang Industrial Equipment Company (100071), Zhongbai Jinmao Chain Company (100120), and Sansan Dateng Heavy Industry Company (100260) to obtain their 2022 records; 3 valid rows found. Extract total assets, total liabilities, year-on-year change in operating revenue, and net profit amount.",
+    "Compute each firm's asset-liability ratio from total assets and total liabilities using total liabilities / total assets × 100%. Results: Haishan Chang Industrial Equipment Company 86.3880%, Zhongbai Jinmao Chain Company 57.6074%, Sansan Dateng Heavy Industry Company 13.6112%.",
+    "Derive shareholder equity from total assets minus total liabilities; then compute ROE = net profit amount / shareholder equity × 100%. ROE for the three firms is 15.2002%, 11.5040%, and 7.1713%, respectively.",
+    "Let the three firms' weights be w1, w2, and w3. Set up the simultaneous equations w1+w2+w3=1, 86.3880w1+57.6074w2+13.6112w3=45, and −19.28w1+12.21w2−9.41w3=0.",
+    "Solve the system of three linear equations to obtain w1=0.13180488, w2=0.49541694, w3=0.37277818.",
+    "Compute portfolio-weighted ROE=15.2002%×0.13180488+11.5040%×0.49541694+7.1713%×0.37277818; the result is 10.376%."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "From company_operation_status.csv, 2022 total assets, total liabilities, net profit amount, and year-on-year operating revenue data were found for all three companies."
+  ],
+  "milestone": {
+    "Firm count": 3,
+    "Haishan Chang Industrial Equipment Company asset-liability ratio": 86.388,
+    "Zhongbai Jinmao Chain Company asset-liability ratio": 57.6074,
+    "Sansan Dateng Heavy Industry Company asset-liability ratio": 13.6112,
+    "Haishan Chang Industrial Equipment Company weight": 0.1318,
+    "Zhongbai Jinmao Chain Company weight": 0.4954,
+    "Sansan Dateng Heavy Industry Company weight": 0.3728,
+    "Portfolio-weighted ROE": 10.376
+  },
+  "answer": [
+    0.1318,
+    0.4954,
+    0.3728,
+    10.376
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard017_result.json b/assets/qa_raw/comprehensive_decision/hard017_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..07990c2c60b225509f39a1e58b68201143c83bb4
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard017_result.json
@@ -0,0 +1,32 @@
+{
+  "id": "hard017",
+  "question": "In 2022, in the chemical raw materials and chemical products manufacturing industry covered by the Implementation Plan for \"Three Products\" in Raw Materials Industry, Hualu Runyuan Technology Co., Ltd. plans to restore profitability through product upgrade and price increases. After implementing the \"Three Products\" reforms, the company can obtain two types of certain profit improvements: one from process and quality improvement, equal to 1.5% of that year's operating revenue; the other from special support and subsidies. Assuming sales volume is unchanged, price increases have no effect on costs, and the goal is to bring net profit exactly to zero, find the minimum price increase rate required based on the company's 2022 operating data and policy information.",
+  "guidelines": "Answer format: minimum price increase rate. Four decimal places. Output the number only, no percent sign or text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "steps": [
+    "From policy_resource.jsonl, confirm that the Implementation Plan for \"Three Products\" in Raw Materials Industry specifies special support and subsidy of 5,000,000 yuan.",
+    "From company_profile.csv, filter record with bmCode=100639; confirm Hualu Runyuan Technology belongs to chemical raw materials and chemical products manufacturing.",
+    "From company_operation_status.csv, filter 2022 record with bmCode=100639; extract operating revenue and net profit: operating revenue 664,310,105.79 yuan, net profit −109,823,137.00 yuan.",
+    "Process and quality improvement profit gain = 664,310,105.79 × 1.5% = 9,964,651.59 yuan; add special support and subsidy 5,000,000.00 yuan.",
+    "Remaining profit gap = 109,823,137.00 − 9,964,651.59 − 5,000,000.00 = 94,858,485.41 yuan.",
+    "Under the assumption that sales volume is unchanged and all price-increase revenue flows to profit, minimum price increase rate = 94,858,485.41 / 664,310,105.79 = 14.2792%."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "From policy_resource.jsonl, the \"Three Products\" policy covering chemical raw materials and chemical products manufacturing specifies special support and subsidy of 5,000,000 yuan.",
+    "From company_profile.csv and company_operation_status.csv, Hualu Runyuan Technology's 2022 industry classification, operating revenue, and net profit were found."
+  ],
+  "milestone": {
+    "Operating revenue (yuan)": 664310105.79,
+    "Net profit (yuan)": -109823137,
+    "Process improvement profit gain (yuan)": 9964651.59,
+    "Subsidy profit gain (yuan)": 5000000,
+    "Remaining profit gap (yuan)": 94858485.41,
+    "Minimum price increase rate (%)": 14.2792
+  },
+  "answer": 14.2792,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard018_result.json b/assets/qa_raw/comprehensive_decision/hard018_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..faae389eea52670a1db158ef0236716393a8e8b8
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard018_result.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard018",
+  "question": "In 2022, does Lianji Chuangji Machine Tool Company meet the basic conditions under the Several Policies on Supporting the Construction of a Strong Province of Skilled Workers? If yes, treat it as a sample firm eligible for skills training subsidies and apply the following: all R&D personnel receive individual skill improvement subsidies at the \"senior technician\" rate; all non-R&D employees at the \"technician\" rate; assume all new subsidies are used to offset that year's R&D investment. Calculate by how many basis points the adjusted R&D investment ratio is lower than the disclosed 2022 R&D investment ratio.",
+  "guidelines": "Answer format: the number of basis points of decline. Two decimal places. Output the number only, no units or text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "steps": [
+    "From company_profile.csv, confirm Lianji Chuangji Machine Tool Company is located in Anhui Province, satisfying the geographic condition for Anhui's local skills policy.",
+    "From policy_resource.jsonl, locate the Notice of Anhui Provincial People's Government on Several Policies Supporting the Construction of a Strong Province of Skilled Workers; extract individual skill improvement subsidy rates: technician 3,500 yuan per person, senior technician 5,000 yuan per person.",
+    "From company_operation_status.csv, filter records with bmCode=100791 and year=2022; extract total employees, R&D personnel count, R&D investment amount, operating revenue, and disclosed R&D investment ratio. Results: 1,276 employees, 290 R&D personnel, R&D investment 58,357,098.35 yuan, operating revenue 1,082,827,667.36 yuan, disclosed R&D ratio 5.37%.",
+    "Non-R&D employees = 1,276 − 290 = 986.",
+    "Total new subsidies = 290 × 5,000 + 986 × 3,500 = 4,901,000.00 yuan.",
+    "Adjusted R&D investment = 58,357,098.35 − 4,901,000.00 = 53,456,098.35 yuan.",
+    "Adjusted R&D ratio = 53,456,098.35 / 1,082,827,667.36 × 100% = 4.9367%. Compared with disclosed 5.37%, decline = 5.37% − 4.9367% = 0.4333%, or 43.33 basis points."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "From company_profile.csv, Lianji Chuangji Machine Tool Company was confirmed to be located in Anhui Province.",
+    "From policy_resource.jsonl, Anhui's strong-province-of-skilled-workers policy was found, with technician subsidy 3,500 yuan/person and senior technician 5,000 yuan/person.",
+    "From company_operation_status.csv, 2022 employee structure, R&D investment, operating revenue, and disclosed R&D ratio were obtained for the company."
+  ],
+  "milestone": {
+    "Total employees": 1276,
+    "R&D personnel count": 290,
+    "Non-R&D employee count": 986,
+    "Total new subsidies (yuan)": 4901000,
+    "Adjusted R&D investment (yuan)": 53456098.35,
+    "Adjusted R&D ratio (%)": 4.9367,
+    "Decline (basis points)": 43.33
+  },
+  "answer": "43.33",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/hard019_result.json b/assets/qa_raw/comprehensive_decision/hard019_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a4cf2862282114bf9c6269258de6ea8d5d3d238d
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/hard019_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "hard019",
+  "question": "2022年，以山安泽医疗科技公司具有“微生态活菌业务”和“高温合金业务”双主业特征。某基金经理希望用以山安泽医疗科技公司与三三达腾重工公司构建一个两股票组合，来替代连机创机机床公司的增长暴露，并进一步检验该替代组合在剔除补贴后的盈利质量与研发强度溢价。若组合要求加权营业收入同比增减幅恰好等于连机创机机床公司2022年的对应指标，请基于本地数据计算：以山安泽医疗科技公司的组合权重、剔除政府奖励资金和补贴后的组合加权净利率，以及该组合研发投入占比相对连机创机机床公司高出的基点数。",
+  "guidelines": "答案格式为：以山安泽医疗科技公司权重,剔除补贴后的组合加权净利率,研发投入占比高出的基点数。前两项按百分比口径保留2位小数，最后一项保留2位小数。仅输出数字和逗号，不要添加单位或文字说明。如无法找到相关数据，请回答“未查询到相关数据”。",
+  "steps": [
+    "从company_core.jsonl中定位以山安泽医疗科技公司，确认其核心竞争力描述同时覆盖微生态活菌业务与高温合金业务，满足题目中的双主业前提。",
+    "从company_operation_status.csv中筛选bmCode分别为764661、100260、100791且year=2022的记录，提取营业收入同比增减幅、净利润金额、营业收入金额、政府奖励资金、补贴和研发投入占比字段。其中，以山安泽医疗科技公司的研发投入占比字段值为18.11%，三三达腾重工公司的研发投入占比字段值为8.90%，连机创机机床公司的研发投入占比字段值为5.37%。",
+    "得到以山安泽医疗科技公司营业收入同比增减幅20.92%，三三达腾重工公司营业收入同比增减幅-9.41%，连机创机机床公司营业收入同比增减幅2.29%。设以山安泽医疗科技公司权重为w，则20.92%×w+(-9.41%)×(1-w)=2.29%，解得w=38.5757%，三三达腾重工公司权重为61.4243%。",
+    "分别计算剔除补贴后的净利率。以山安泽医疗科技公司剔除补贴后的净利润=105565919.71-27049179.17=78516740.54元，剔除补贴后的净利率=78516740.54/793847988.60×100%=9.8907%。三三达腾重工公司剔除补贴后的净利润=155125599.81-11937935.97=143187663.84元，剔除补贴后的净利率=143187663.84/799392301.84×100%=17.9121%。",
+    "按组合权重计算剔除补贴后的组合加权净利率=38.5757%×9.8907%+61.4243%×17.9121%=14.8178%。",
+    "按组合权重计算组合研发投入占比=38.5757%×18.11%+61.4243%×8.90%=12.4528%。连机创机机床公司研发投入占比字段值为5.37%，因此组合研发投入占比高出12.4528%-5.37%=7.0828个百分点，即708.28个基点。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从company_operation_status.csv中获取了以山安泽医疗科技公司、三三达腾重工公司和连机创机机床公司2022年的增长、盈利、补贴和研发投入数据。"
+  ],
+  "milestone": {
+    "以山安泽医疗科技公司权重(%)": 38.58,
+    "三三达腾重工公司权重(%)": 61.42,
+    "以山安泽医疗科技公司剔除补贴后净利率(%)": 9.89,
+    "三三达腾重工公司剔除补贴后净利率(%)": 17.91,
+    "剔除补贴后的组合加权净利率(%)": 14.82,
+    "组合研发投入占比(%)": 12.45,
+    "相对连机创机机床公司高出的基点数": 708.28
+  },
+  "answer": [
+    38.58,
+    14.82,
+    708.28
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium001_result.json b/assets/qa_raw/comprehensive_decision/medium001_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..521115bc28087309e423b1bb56c67e07584d3b41
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium001_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium001",
+  "question": "For 2022 pharmaceutical manufacturing industry data by province, if R&D funding intensity is measured as each province's total R&D expenditure as a percentage of its total operating revenue, among all provinces with complete data records, what is the specific value of this ratio for the province with the highest level? Which company has the highest R&D funding intensity in that province?",
+  "guidelines": "The first answer is a numeric value (2 decimal places), unit is %; the second answer is the full company name, which must exactly match the \"Company Name\" field in company_profile.csv. If either question cannot be answered, respond with \"No relevant data found\".",
+  "steps": [
+    "From regional_industry_status.csv, filter industry=\"Pharmaceutical Manufacturing\", extract province, total R&D expenditure, total operating revenue; exclude records with missing total R&D expenditure or total operating revenue, or with total operating revenue ≤ 0, yielding 16 valid province records.",
+    "For each province, compute R&D funding intensity = total R&D expenditure ÷ total operating revenue × 100%, sort by intensity descending. Shanghai has the highest: total R&D expenditure 40,798,081,760.73 yuan, total operating revenue 160,133,198,188.25 yuan, intensity = 40,798,081,760.73 ÷ 160,133,198,188.25 × 100% = 25.48%.",
+    "From company_profile.csv, filter province=\"Shanghai\" and industry=\"Pharmaceutical Manufacturing\", obtain company names and bmCode list for pharmaceutical manufacturing firms in that city.",
+    "From company_operation_status.csv, filter year=2022 and bmCode in the above set, with non-null R&D expenditure and operating revenue, and operating revenue > 0.",
+    "For each company, compute R&D funding intensity = R&D expenditure ÷ operating revenue × 100% using the same formula. The highest is \"Kangsheng Anjian Biopharmaceutical Company\" (bmCode=505404): R&D expenditure 809,733,452.00 yuan, operating revenue 12,792,315.00 yuan, intensity = 809,733,452.00 ÷ 12,792,315.00 × 100% = 6329.84%."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "In regional_industry_status.csv, \"Pharmaceutical Manufacturing\" has 16 provinces after complete-data filtering; Shanghai ranks first in provincial R&D funding intensity.",
+    "Joining company_profile.csv with company_operation_status.csv (2022), R&D funding intensity can be computed for Shanghai pharmaceutical manufacturing firms; Kangsheng Anjian Biopharmaceutical Company has the highest ratio."
+  ],
+  "milestone": {
+    "Province with highest R&D funding intensity": "Shanghai",
+    "Shanghai total R&D expenditure (yuan)": 40798081760.73,
+    "Shanghai total operating revenue (yuan)": 160133198188.25,
+    "Provincial R&D funding intensity (%)": 25.48,
+    "Company with highest R&D funding intensity in that province": "Kangsheng Anjian Biopharmaceutical Company",
+    "Company R&D expenditure (yuan)": 809733452.0,
+    "Company operating revenue (yuan)": 12792315.0,
+    "Company R&D funding intensity (%)": 6329.84
+  },
+  "answer": [
+    25.48,
+    "Kangsheng Anjian Biopharmaceutical Company"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium002_result.json b/assets/qa_raw/comprehensive_decision/medium002_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b84ac54c5f9cb966c3c7c5a8b3c747c30f86476d
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium002_result.json
@@ -0,0 +1,35 @@
+{
+  "id": "medium002",
+  "question": "In 2022, a semiconductor company plans to expand production and wishes to locate in the province with the highest enterprise concentration to gain industrial synergy effects. What is the proportion of that province's semiconductor industry enterprise count to the national total? What proportion does that province's total operating profit in the semiconductor industry account for of the national semiconductor industry's total operating profit?",
+  "guidelines": "Two answers required, both numeric values (2 decimal places), unit is %. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From regional_industry_status.csv, filter records with industry=\"Semiconductor Industry\", extract province, total enterprise count, total operating profit; 34 province records obtained.",
+    "From national_industry_status.csv, filter records with industry=\"Semiconductor Industry\", extract total enterprise count and total operating profit: national total 172 enterprises, national total operating profit 411,298,557,285.26 yuan.",
+    "Sort by total enterprise count in regional_industry_status.csv descending; Guangdong Province has the highest with 54 enterprises; corresponding total operating profit 25,562,691,329.46 yuan.",
+    "Compute enterprise concentration = Guangdong enterprise count / national enterprise count × 100% = 54 / 172 × 100% = 31.40%.",
+    "Compute that province's semiconductor industry operating profit share = Guangdong total operating profit / national total operating profit × 100% = 25,562,691,329.46 / 411,298,557,285.26 × 100% = 6.22%."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "regional_industry_status.csv has 34 semiconductor industry province records; Guangdong Province has the highest enterprise count (54), with total operating profit 25,562,691,329.46 yuan.",
+    "national_industry_status.csv shows national semiconductor industry total of 172 enterprises, national total operating profit 411,298,557,285.26 yuan."
+  ],
+  "milestone": {
+    "Province with highest enterprise concentration": "Guangdong Province",
+    "Guangdong semiconductor industry enterprise count": 54,
+    "National semiconductor industry enterprise count": 172,
+    "Enterprise concentration (%)": 31.4,
+    "Guangdong semiconductor industry total operating profit (yuan)": 25562691329.46,
+    "National semiconductor industry total operating profit (yuan)": 411298557285.26,
+    "Guangdong operating profit share (%)": 6.22
+  },
+  "answer": [
+    31.4,
+    6.22
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium003_result.json b/assets/qa_raw/comprehensive_decision/medium003_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a1be38b72e9814c5317dff32ca8cb9d03e9ba8d
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium003_result.json
@@ -0,0 +1,35 @@
+{
+  "id": "medium003",
+  "question": "In 2022, measuring per capita output efficiency of automobile manufacturing by province using revenue per capita (total operating revenue ÷ total employee count), among all provinces nationwide, what is the specific value in yuan per person for the province with the highest indicator? Compared to the national average, by what percentage (1 decimal place) is that province's per capita revenue higher?",
+  "guidelines": "Two answers required: first is a numeric value (2 decimal places), unit yuan/person; second is a percentage (1 decimal place), unit %. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From regional_industry_status.csv, filter all province records with industry=\"Automobile Manufacturing\", extract province, total operating revenue, total employee count; 34 records obtained.",
+    "Filter province records with non-null total operating revenue and total employee count, and total employee count > 0; 14 valid provinces obtained.",
+    "Compute per capita revenue for each valid province = total operating revenue / total employee count, sort by this indicator descending; Beijing has the highest: 341,596,317,171.50 / 87,614 = 3,898,878.23 yuan/person.",
+    "From national_industry_status.csv, filter national records with industry=\"Automobile Manufacturing\", extract total operating revenue and total employee count: national total operating revenue 4,614,449,954,119.47 yuan, national total employee count 3,254,510, national average per capita revenue = 1,417,863.20 yuan/person.",
+    "Compute Beijing's excess over national average = (3,898,878.23 - 1,417,863.20) / 1,417,863.20 × 100% = 174.98%, rounded to 1 decimal place per requirement: 175.0%."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "regional_industry_status.csv has 34 province records for automobile manufacturing; 14 provinces have valid operating revenue and employee count; Beijing has the highest per capita revenue.",
+    "national_industry_status.csv shows automobile manufacturing national total operating revenue 4,614,449,954,119.47 yuan, national total employee count 3,254,510; national average per capita revenue = 1,417,863.20 yuan/person."
+  ],
+  "milestone": {
+    "Beijing total operating revenue (yuan)": 341596317171.5,
+    "Beijing total employee count": 87614,
+    "Beijing per capita revenue (yuan/person)": 3898878.23,
+    "National total operating revenue (yuan)": 4614449954119.47,
+    "National total employee count": 3254510,
+    "National average per capita revenue (yuan/person)": 1417863.2,
+    "Beijing excess over national average (%)": 175.0
+  },
+  "answer": [
+    3898878.23,
+    175.0
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium004_result.json b/assets/qa_raw/comprehensive_decision/medium004_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..470161fca9f47f37903f87c5bf9c2c630188adf7
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium004_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium004",
+  "question": "In 2022, rank provinces by profitability in chemical raw materials and chemical products manufacturing. Provincial operating profit margin is computed as total operating profit divided by total operating revenue. Using this as the ranking criterion, what is Guangdong Province's rank? Apply the same ranking to all relevant enterprises within Guangdong Province—which enterprise ranks first?",
+  "guidelines": "Two answers required: first is Guangdong Province's rank in the provincial ranking (integer, e.g. \"6\" means 6th place); second is the full name of the top-ranked enterprise in Guangdong, which must match the \"Company Name\" field in company_profile.csv. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From regional_industry_status.csv, filter province records with industry=\"Chemical Raw Materials and Chemical Products Manufacturing\", extract province, total operating profit, total operating revenue; 34 records obtained.",
+    "Filter records with non-null total operating profit and total operating revenue, and total operating revenue ≠ 0; 16 valid provinces. Compute provincial operating profit margin = total operating profit / total operating revenue × 100%, sort by this indicator descending.",
+    "In the provincial ranking, Guangdong Province has total operating profit 11,690,448,651.68 yuan, total operating revenue 101,800,752,670.91 yuan, operating profit margin = 11.4837%, rank 6th.",
+    "From company_profile.csv, filter enterprises with province=\"Guangdong Province\" and industry=\"Chemical Raw Materials and Chemical Products Manufacturing\", obtain company names and bmCode; from company_operation_status.csv, filter year=2022 and bmCode in the above set, extract operating profit and operating revenue.",
+    "For Guangdong enterprises, compute operating profit margin = operating profit / operating revenue × 100% (requiring operating revenue ≠ 0), sort by operating profit margin descending. First place: \"Hengyi Changhua Technology Co., Ltd.\" (bmCode=533611), operating profit margin = 2,190,338,633.59 / 3,466,111,075.75 × 100% = 63.19%."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "regional_industry_status.csv has 34 province records for chemical raw materials and chemical products manufacturing; 16 provinces have valid operating profit and operating revenue data; Guangdong Province ranks 6th by operating profit margin.",
+    "Joining company_profile.csv with company_operation_status.csv yields 41 valid 2022 enterprise records for Guangdong in this industry; Hengyi Changhua Technology Co., Ltd. ranks first by enterprise operating profit margin."
+  ],
+  "milestone": {
+    "Guangdong total operating profit (yuan)": 11690448651.68,
+    "Guangdong total operating revenue (yuan)": 101800752670.91,
+    "Guangdong operating profit margin (%)": 11.4837,
+    "Guangdong provincial rank": 6,
+    "Top-ranked enterprise in Guangdong": "Hengyi Changhua Technology Co., Ltd.",
+    "Top enterprise operating profit (yuan)": 2190338633.59,
+    "Top enterprise operating revenue (yuan)": 3466111075.75,
+    "Top enterprise operating profit margin (%)": 63.19
+  },
+  "answer": [
+    6,
+    "Hengyi Changhua Technology Co., Ltd."
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium005_result.json b/assets/qa_raw/comprehensive_decision/medium005_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..1662123603555bba365ab96fe5242abee9747f66
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium005_result.json
@@ -0,0 +1,27 @@
+{
+  "id": "medium005",
+  "question": "In 2022, from all private enterprises in the food and beverage industry, aggregate government rewards and subsidy amounts by province of registration to find the province with the highest provincial subsidy total. How many hundred million yuan in government subsidies did private enterprises in that province receive in total?",
+  "guidelines": "The answer should be a numerical value (2 decimal places), unit is hundred million yuan. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all enterprise records with industry=\"Food and Beverage\" and ownership=\"Private Enterprise\" from company_profile.csv, extract enterprise name, bmCode and province fields, finding 161 food and beverage private enterprises.",
+    "Join with company_operation_status.csv via bmCode to obtain 2022 data for these enterprises, extract government reward funds and subsidy fields.",
+    "Filter records where government reward funds and subsidy are non-null, resulting in 157 enterprises with subsidy data.",
+    "Group by province and calculate total government subsidies received by private enterprises in each province.",
+    "Sort provinces by total government subsidies in descending order. The province with the highest total subsidies is Inner Mongolia, with total subsidies of 1,259,874,619.23 yuan; convert from yuan to hundred million yuan (divide by 100,000,000), yielding 12.60 hundred million yuan."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found 161 food and beverage private enterprises from company_profile.csv.",
+    "Found 2022 government subsidy data for 157 food and beverage private enterprises from company_operation_status.csv."
+  ],
+  "milestone": {
+    "Inner Mongolia food and beverage private enterprise total subsidies (yuan)": 1259874619.23,
+    "Inner Mongolia food and beverage private enterprise total subsidies (hundred million yuan)": 12.6
+  },
+  "answer": 12.6,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium006_result.json b/assets/qa_raw/comprehensive_decision/medium006_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c34f6e049a3456e1ac62f312992c13cd4c19bb00
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium006_result.json
@@ -0,0 +1,26 @@
+{
+  "id": "medium006",
+  "question": "In 2022, a local government planned to introduce support policies for the specialized equipment manufacturing industry and needed to understand the province's R&D personnel investment level in this industry. Is the average proportion of R&D personnel to total employees in specialized equipment manufacturing enterprises in Zhejiang Province higher than the national average for this industry? ",
+  "guidelines": "The answer should be \"Yes\" or \"No\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter records with industry=\"Specialized Equipment Manufacturing\" and province=\"Zhejiang\" from regional_industry_status.csv, extract average R&D personnel proportion field; Zhejiang's average R&D personnel proportion is 18.25%.",
+    "Filter records with industry=\"Specialized Equipment Manufacturing\" from national_industry_status.csv, extract average R&D personnel proportion field; national average R&D personnel proportion is 20.10%.",
+    "Compare Zhejiang's average R&D personnel proportion (18.25%) with the national average (20.10%).",
+    "Zhejiang's value is not greater than the national value, therefore the answer is \"No\"."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "Found average R&D personnel proportion data for Zhejiang's specialized equipment manufacturing in regional_industry_status.csv.",
+    "Found national average R&D personnel proportion data for specialized equipment manufacturing in national_industry_status.csv."
+  ],
+  "milestone": {
+    "Zhejiang average R&D personnel proportion (%)": 18.25,
+    "National average R&D personnel proportion (%)": 20.1
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium007_result.json b/assets/qa_raw/comprehensive_decision/medium007_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ecb162397e98e8650b11b25ecc65bd304a482a16
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium007_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium007",
+  "question": "In 2022, analyze government subsidy leverage in the information transmission, software and information technology services industry. Define government subsidy leverage effect as the ratio of each province's total operating profit to total government subsidies. What is the specific ratio for the province with the highest government subsidy leverage effect? Which enterprise in that province has the highest leverage effect?",
+  "guidelines": "Two answers required: first is a numeric value (2 decimal places, unitless ratio); second is the full company name. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From company_profile.csv, filter industry=\"Information Transmission, Software and Information Technology Services\", extract company name, bmCode, and province.",
+    "Inner-join with company_operation_status.csv on year=2022 by company name and bmCode to obtain the 2022 enterprise sample for this industry (644 records, covering 31 provincial-level regions).",
+    "Group by province: sum operating profit to get each province's total operating profit; sum government rewards and subsidies to get each province's total government subsidies.",
+    "Keep provinces with total government subsidies > 0 and non-null total operating profit; compute government subsidy leverage effect = total operating profit / total government subsidies — 31 valid provinces; sort descending: first is Jiangxi Province — total operating profit 93,463,073.13 yuan, total government rewards and subsidies 1,391,005.09 yuan, provincial leverage = 67.19 (2 decimal places).",
+    "Sort enterprises by government subsidy leverage effect descending; top enterprise: Dongche Kexin Systems Company (bmCode=591984); operating profit 20,080,241.24 yuan, government rewards and subsidies 1,391,005.09 yuan, enterprise-level leverage ≈ 14.44; Zhongke Ruanchuang Software Company is excluded from enterprise-level ranking due to missing subsidy data."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found 644 enterprise records for information transmission, software and information technology services in company_profile.csv.",
+    "Found 644 year-2022 records in company_operation_status.csv matching those enterprises in company_profile.csv."
+  ],
+  "milestone": {
+    "Provincial aggregation note": "Aggregate by province from the enterprise table; do not use Jiangxi totals missing from the regional table",
+    "Jiangxi total operating profit (yuan)": 93463073.13,
+    "Jiangxi total government rewards and subsidies (yuan)": 1391005.09,
+    "Jiangxi government subsidy leverage effect": 67.19,
+    "Enterprise with highest government subsidy leverage effect": "Dongche Kexin Systems Company",
+    "Enterprise operating profit (yuan)": 20080241.24,
+    "Enterprise government rewards and subsidies (yuan)": 1391005.09,
+    "Enterprise government subsidy leverage effect": 14.44
+  },
+  "answer": [
+    67.19,
+    "Dongche Kexin Systems Company"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium008_result.json b/assets/qa_raw/comprehensive_decision/medium008_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..6664d57dab914484c456a71aa382930576f8beb3
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium008_result.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium008",
+  "question": "In 2022, to study the capital turnover of central state-owned enterprises in the electricity, heat, gas and water production and supply industry, calculate the asset turnover ratio for each enterprise by dividing its annual operating revenue by its total assets. Find the arithmetic mean of the asset turnover ratios for these enterprises.",
+  "guidelines": "The answer should be a numerical value (rounded to 4 decimal places). If the relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter records from company_profile.csv where industry=\"Electricity, heat, gas and water production and supply\" and ownership=\"Central state-owned enterprise\", extract company name and bmCode fields, finding 44 enterprises in total.",
+    "Join with company_operation_status.csv by bmCode to retrieve 2022 data for these enterprises, extract operating revenue amount and total assets fields.",
+    "Filter records where both operating revenue amount and total assets are not empty and total assets is greater than 0, resulting in 44 valid enterprises.",
+    "Calculate asset turnover ratio for each enterprise = operating revenue amount / total assets.",
+    "Calculate the average asset turnover ratio for all qualifying central state-owned enterprises = sum of asset turnover ratios (14.369275) / number of enterprises (44) = 0.3266."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found 44 central state-owned enterprises in the electricity, heat, gas and water production and supply industry from company_profile.csv.",
+    "Found 2022 operating revenue and total assets data for 44 enterprises from company_operation_status.csv."
+  ],
+  "milestone": {
+    "Sum of Asset Turnover Ratios": 14.369275,
+    "Number of Valid Enterprises": 44,
+    "Average Asset Turnover Ratio": 0.3266
+  },
+  "answer": 0.3266,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium009_result.json b/assets/qa_raw/comprehensive_decision/medium009_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e38704e44fb597644a8a69f17564f6cf8da79e3f
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium009_result.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium009",
+  "question": "In 2022, a mining industry enterprise plans to go public for financing and hopes to list on the exchange with the highest average market capitalization among enterprises in this industry. Among the four exchanges—Shenzhen Stock Exchange, Hong Kong Stock Exchange, Shanghai Stock Exchange, and Beijing Stock Exchange—which exchange has the highest average market capitalization of listed mining enterprises?",
+  "guidelines": "The answer should be the exchange name (e.g., \"Shenzhen Stock Exchange\", \"Hong Kong Stock Exchange\", \"Shanghai Stock Exchange\", \"Beijing Stock Exchange\"). If the relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter records from company_profile.csv where industry=\"Mining\", extract company name, bmCode and exchange fields, finding 143 mining enterprises in total.",
+    "Join with company_operation_status.csv by bmCode to retrieve 2022 data for these enterprises, extract company market capitalization field.",
+    "Filter records where company market capitalization is not empty, resulting in 142 enterprises with market capitalization data.",
+    "Group by exchange field and calculate the average market capitalization of mining enterprises for each exchange.",
+    "Sort by average market capitalization in descending order. Shanghai Stock Exchange has the highest average market capitalization, with 49 mining enterprises, total market capitalization of 4,771.8 billion CNY, and average market capitalization of 973.84 billion CNY."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found 143 mining enterprises from company_profile.csv.",
+    "Found 2022 market capitalization data for 142 mining enterprises from company_operation_status.csv."
+  ],
+  "milestone": {
+    "Shanghai Stock Exchange Total Market Capitalization of Mining Enterprises (billion CNY)": 47718.0,
+    "Number of Mining Enterprises on Shanghai Stock Exchange": 49,
+    "Shanghai Stock Exchange Average Market Capitalization of Mining Enterprises (billion CNY)": 973.84
+  },
+  "answer": "Shanghai Stock Exchange",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium010_result.json b/assets/qa_raw/comprehensive_decision/medium010_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c20da2a6262c8a292848e4b569928ed3c499867
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium010_result.json
@@ -0,0 +1,32 @@
+{
+  "id": "medium010",
+  "question": "In 2022, in the provincial data for the construction industry, each province has an indicator reflecting the average asset-liability ratio (financial leverage level) of enterprises in that province's industry (considering only enterprises with valid total assets and total liabilities). Among the provinces covered by valid data, which province has the lowest value for this mean indicator, and what is that value?",
+  "guidelines": "The answer should be a numerical value (rounded to 2 decimal places), in units of %. If the relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter records from company_profile.csv where industry=\"Construction\", extract company name, bmCode and province fields, finding 148 enterprises in total.",
+    "Join with company_operation_status.csv by bmCode, extract total liabilities and total assets fields, resulting in 148 records after merging.",
+    "Filter enterprise records where both total liabilities and total assets are not empty and total assets is greater than 0, resulting in 148 valid enterprises covering 22 provinces.",
+    "Calculate asset-liability ratio for each enterprise = total liabilities / total assets × 100%, then group by province and calculate the mean asset-liability ratio for each province.",
+    "Sort all provinces by mean asset-liability ratio in ascending order. Shanxi Province has the lowest mean asset-liability ratio of 27.17%."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found 148 construction industry enterprises from company_profile.csv.",
+    "Retrieved total liabilities and total assets data for 148 enterprises from company_operation_status.csv; 22 provinces have valid data records."
+  ],
+  "milestone": {
+    "Total Number of Construction Enterprises": 148,
+    "Number of Valid Enterprises (total liabilities and total assets not empty)": 148,
+    "Number of Valid Provinces": 22,
+    "Shanxi Province Mean Asset-Liability Ratio (%)": 27.17
+  },
+  "answer": [
+    "Shanxi Province",
+    27.17
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium011_result.json b/assets/qa_raw/comprehensive_decision/medium011_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c07cf08bda9c05b58809fd3a5f585f434331155b
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium011_result.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium011",
+  "question": "In 2022, among all enterprises in the rubber and plastic products industry with R&D investment records, what is the R&D concentration CR5? ",
+  "guidelines": "The answer should be a numerical value (rounded to 2 decimal places), in units of %. If the relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter records from company_profile.csv where industry=\"Rubber and plastic products\", extract company name and bmCode fields, finding 107 enterprises in total.",
+    "Join with company_operation_status.csv by bmCode to retrieve 2022 data for these enterprises, extract R&D investment amount field.",
+    "Filter records where R&D investment amount is not empty, resulting in 106 enterprises with R&D investment data. Sort by R&D investment amount in descending order, extract the top 5 enterprises and their R&D investment amounts.",
+    "Calculate the sum of R&D investment amounts of the top 5 enterprises as 4,221,126,553.77 CNY. Calculate the total R&D investment amount of all valid enterprises in the industry as 12,179,847,530.98 CNY.",
+    "Calculate R&D concentration CR5 = (4,221,126,553.77 / 12,179,847,530.98) × 100% = 34.66%."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found 107 rubber and plastic products enterprises from company_profile.csv.",
+    "Found 2022 R&D investment amount data for 106 enterprises from company_operation_status.csv."
+  ],
+  "milestone": {
+    "Sum of R&D Investment Amount of Top 5 Enterprises (CNY)": 4221126553.77,
+    "Total R&D Investment Amount of the Industry (CNY)": 12179847530.98,
+    "R&D Concentration CR5 (%)": 34.66
+  },
+  "answer": 34.66,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium012_result.json b/assets/qa_raw/comprehensive_decision/medium012_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..df0419b1396d8731e11d0abfd072fd0f535e49fd
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium012_result.json
@@ -0,0 +1,33 @@
+{
+  "id": "medium012",
+  "question": "In 2022, among all enterprises in Guangdong Province belonging to the wholesale and retail trade industry, using each enterprise's net profit margin as the comparison standard, what is the indicator value for the enterprise with the highest net profit margin? What is that enterprise's rank among all enterprises in this industry nationwide? ",
+  "guidelines": "Two answers required: first is a numeric value (2 decimal places, unit %); second is the rank number (integer, e.g. \"7\" means 7th place). If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From company_profile.csv, filter all enterprise records with industry=\"Wholesale and Retail Trade\" and province=\"Guangdong Province\", extract company name and bmCode; 43 Guangdong wholesale and retail enterprises found.",
+    "From company_operation_status.csv, join 2022 data by bmCode for these enterprises, extract net profit and operating revenue; 2022 data found for 43 enterprises.",
+    "Filter enterprise records with non-null net profit and operating revenue, and operating revenue ≠ 0; 43 valid records.",
+    "Compute each enterprise's net profit margin = net profit / operating revenue × 100%. Sort all enterprises by net profit margin descending; the enterprise with the highest net profit margin is \"Yonghui Changda Wholesale Company\", net profit 292,221,119.71 yuan, operating revenue 935,248,730.59 yuan, net profit margin = 31.25%.",
+    "From company_profile.csv, filter all enterprises nationwide with industry=\"Wholesale and Retail Trade\"; from company_operation_status.csv, take year=2022 with operating revenue > 0 and non-null net profit. Sort nationwide valid enterprises by net profit margin descending; Yonghui Changda Wholesale Company ranks 7th."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found 43 Guangdong wholesale and retail enterprises from company_profile.csv.",
+    "Found 2022 net profit and operating revenue data for 43 Guangdong wholesale and retail enterprises from company_operation_status.csv; Yonghui Changda Wholesale Company has the highest net profit margin (31.25%).",
+    "Nationwide (same criteria: 2022 and operating revenue > 0) there are 273 valid enterprises; Yonghui Changda Wholesale Company ranks 7th in net profit margin in this industry nationwide."
+  ],
+  "milestone": {
+    "Yonghui Changda Wholesale Company net profit (yuan)": 292221119.71,
+    "Yonghui Changda Wholesale Company operating revenue (yuan)": 935248730.59,
+    "Yonghui Changda Wholesale Company net profit margin (%)": 31.25,
+    "Yonghui Changda Wholesale Company nationwide rank": 7
+  },
+  "answer": [
+    31.25,
+    7
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium013_result.json b/assets/qa_raw/comprehensive_decision/medium013_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..55313e24c0d5e32dcb6badd2403aaedb61a17605
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium013_result.json
@@ -0,0 +1,33 @@
+{
+  "id": "medium013",
+  "question": "In 2022, a scientific research and technical services enterprise wishes to identify the province with the fastest net profit growth in the industry to guide market expansion. What is the indicator value for the province with the highest median year-on-year net profit growth rate in the national scientific research and technical services industry? What is the nationwide rank of the enterprise with the highest such indicator in that province among all enterprises in this industry?",
+  "guidelines": "Two answers required: first is a numeric value (2 decimal places, unit %), i.e. the indicator value for the province with the highest \"median year-on-year net profit growth rate\" in this industry nationwide; second is a rank number (integer, e.g. \"23\" means 23rd place), i.e. the nationwide rank of the enterprise with the highest \"year-on-year net profit growth rate\" in that province among all enterprises in this industry. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From regional_industry_status.csv, filter all records with industry=\"Scientific Research and Technical Services\", extract province and median year-on-year net profit growth rate; 34 province records found.",
+    "Filter province records with non-null median year-on-year net profit growth rate; 16 valid provinces.",
+    "Sort by median year-on-year net profit growth rate descending; first place: Anhui Province, median year-on-year net profit growth rate 13.81%. From company_profile.csv, filter enterprises with province=\"Anhui Province\" and industry=\"Scientific Research and Technical Services\", obtain enterprise bmCode set.",
+    "From company_operation_status.csv, filter year=2022 and bmCode in that set, extract \"year-on-year net profit growth rate\", sort by this indicator descending. Highest enterprise in that province: Zhongqi Shengyuan Technology Research Institute, year-on-year net profit growth rate 32.58%.",
+    "From company_profile.csv, filter nationwide enterprises with industry=\"Scientific Research and Technical Services\"; from company_operation_status.csv, filter year=2022 with valid year-on-year net profit growth rate. Sort by year-on-year net profit growth rate descending; Zhongqi Shengyuan Technology Research Institute ranks 23rd nationwide."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "regional_industry_status.csv: scientific research and technical services has 34 province records; 16 provinces have valid median year-on-year net profit growth rate after filtering; Anhui Province has the highest at 13.81.",
+    "Company level: among 2022 valid enterprises in Anhui Province (industry=scientific research and technical services), Zhongqi Shengyuan Technology Research Institute has the highest year-on-year net profit growth rate (32.58%); this enterprise ranks 23rd nationwide in year-on-year net profit growth rate among all enterprises in this industry."
+  ],
+  "milestone": {
+    "Valid province count for scientific research and technical services (non-null median YoY net profit growth)": 16,
+    "Anhui Province median year-on-year net profit growth rate (%)": 13.81,
+    "Enterprise with highest indicator in Anhui Province": "Zhongqi Shengyuan Technology Research Institute",
+    "Enterprise year-on-year net profit growth rate (%)": 32.58,
+    "Enterprise nationwide rank": 23
+  },
+  "answer": [
+    13.81,
+    23
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium014_result.json b/assets/qa_raw/comprehensive_decision/medium014_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..efa8b41e57e1f8bafa2cbdc8023e6715e22a2aa7
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium014_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium014",
+  "question": "In 2022, for the metal smelting and rolling processing industry, among provinces with valid records for both total government subsidies and total industry employee count, per capita subsidy is computed as each province's total government rewards and subsidies divided by that province's industry employee count. What is the per capita subsidy in yuan for the province with the highest per capita subsidy? What is the nationwide rank of the enterprise with the highest such indicator in that province among all enterprises in this industry?",
+  "guidelines": "Two answers required: first is a numeric value (2 decimal places, unit yuan/person), i.e. the highest provincial per capita subsidy; second is a rank number (integer), indicating the nationwide rank of the enterprise with the highest indicator in that province among all enterprises in this industry. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From regional_industry_status.csv, filter all province records with industry=\"Metal Smelting and Rolling Processing\", extract province, total government rewards and subsidies, total employee count; 34 province records obtained.",
+    "Filter records with non-null total government rewards and subsidies and total employee count, total employee count > 0 and total government rewards and subsidies > 0; 16 valid provinces.",
+    "Compute each province's per capita subsidy (yuan/person) = total government rewards and subsidies / total employee count, sort by per capita subsidy descending. Province with highest per capita subsidy: Shanghai; 893,080,778.37 / 50,830 = 17,569.95 yuan/person.",
+    "Within Shanghai: from company_profile.csv, filter enterprises with industry=\"Metal Smelting and Rolling Processing\" and province=\"Shanghai\", obtain bmCode set; from company_operation_status.csv, filter year=2022 and bmCode in that set, with government rewards and subsidies > 0 and total employee count > 0. Compute each enterprise's per capita subsidy = government rewards and subsidies / total employee count, sort descending; first place: Xin Ge Jinze Materials Company.",
+    "Nationwide, for all enterprises in this industry (year=2022, government rewards and subsidies > 0, total employee count > 0), compute per capita subsidy using the same formula and sort descending; Xin Ge Jinze Materials Company ranks 12th nationwide."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "regional_industry_status.csv: metal smelting and rolling processing has 34 province records; 16 provinces have valid and positive total government rewards and subsidies and total employee count after filtering.",
+    "Shanghai enterprises: among enterprises in this industry with 2022 data, government rewards and subsidies > 0 and total employee count > 0, Xin Ge Jinze Materials Company has the highest per capita subsidy; it ranks 12th nationwide under the same criteria."
+  ],
+  "milestone": {
+    "Shanghai total government rewards and subsidies (yuan)": 893080778.37,
+    "Shanghai total employee count": 50830.0,
+    "Shanghai per capita subsidy (yuan/person)": 17569.95,
+    "Enterprise with highest indicator in Shanghai": "Xin Ge Jinze Materials Company",
+    "Enterprise per capita subsidy (yuan/person)": 31941.25,
+    "Enterprise nationwide rank": 12
+  },
+  "answer": [
+    17569.95,
+    12
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium015_result.json b/assets/qa_raw/comprehensive_decision/medium015_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d5a3830138e8f17779f4bef03ce8903cfcd9854
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium015_result.json
@@ -0,0 +1,65 @@
+{
+  "id": "medium015",
+  "question": "List the 2022 indicators for which Shandong Province's financial industry enterprise averages are below the national financial industry medians.",
+  "guidelines": "The answer must list all qualifying indicator names, separated by semicolons. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From regional_industry_status.csv, filter records with industry=\"Financial Industry\" and province=\"Shandong Province\", extract total enterprise count; 12 Shandong financial industry enterprises found.",
+    "From national_industry_status.csv, filter records with industry=\"Financial Industry\" and district=\"National\", extract total enterprise count; 297 national financial industry enterprises found.",
+    "From regional_industry_status.csv, extract all indicator averages for Shandong financial industry enterprises, including average total assets, average net profit, average total employee count, average operating revenue, average asset-liability ratio, average R&D expenditure, etc.",
+    "From national_industry_status.csv, extract all indicator medians for national financial industry enterprises, including median total assets, median net profit, median total employee count, median operating revenue, median asset-liability ratio, median R&D expenditure, etc.",
+    "Compare each Shandong financial industry indicator average with the corresponding national financial industry median; exclude indicators such as enterprise count and totals that are not suitable for comparison. Filter indicators where Shandong average is below national median; 28 indicators found, mainly including year-on-year R&D personnel growth rate, year-on-year operating profit growth rate, year-on-year net profit growth rate, annual Chinese patent applications, cumulative Chinese patent applications, R&D personnel count, R&D personnel ratio, etc. Final result: Shandong Province financial industry enterprises have 28 indicators with averages below the national financial industry medians."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Shandong Province financial industry enterprise count": 12,
+    "National financial industry enterprise count": 297,
+    "Indicator count below national median": 28,
+    "Main indicator list": [
+      "Year-on-year R&D personnel growth rate",
+      "Year-on-year operating profit growth rate",
+      "Year-on-year net profit growth rate",
+      "Year-on-year employee growth rate",
+      "Capitalized R&D expenditure",
+      "Year-on-year capitalized R&D expenditure growth rate",
+      "Annual PCT patent applications",
+      "Annual PCT invention patent applications",
+      "Provincial/ministerial science and technology progress award",
+      "Participation in drafting national standards"
+    ]
+  },
+  "answer": [
+    "Year-on-year R&D personnel growth rate",
+    "Year-on-year operating profit growth rate",
+    "Year-on-year net profit growth rate",
+    "Year-on-year employee growth rate",
+    "Capitalized R&D expenditure",
+    "Year-on-year capitalized R&D expenditure growth rate",
+    "Annual PCT patent applications",
+    "Annual PCT invention patent applications",
+    "Provincial/ministerial science and technology progress award",
+    "Participation in drafting national standards",
+    "Participation in drafting industry standards",
+    "Annual Chinese patent applications",
+    "Annual Chinese invention patent applications",
+    "Annual Chinese patent grants",
+    "Cumulative Chinese invention patent applications",
+    "Annual Chinese invention patent grants",
+    "Cumulative PCT patent applications",
+    "Cumulative PCT invention patent applications",
+    "Cumulative Chinese patent applications",
+    "Cumulative Chinese invention patent grants",
+    "Cumulative patent citations",
+    "R&D personnel ratio",
+    "R&D personnel count",
+    "Year-on-year R&D expenditure growth rate",
+    "Cumulative Chinese invention patent lapses",
+    "Company market value",
+    "Asset-liability ratio",
+    "Total employee count"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium016_result.json b/assets/qa_raw/comprehensive_decision/medium016_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4b64de5a29e18d88b77499693d9aff3889ff7531
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium016_result.json
@@ -0,0 +1,22 @@
+{
+  "id": "medium016",
+  "question": "In 2022, between Sichuan Province's top enterprise by operating revenue and Shandong Province's top enterprise by net profit in the pharmaceutical manufacturing industry, which one has the highest tax payment?",
+  "guidelines": "The answer must be the company name. Output only the company name, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From company_profile.csv, filter all enterprise records with province=\"Sichuan Province\", extract company name and bmCode; 198 Sichuan enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by company name and bmCode, extract operating revenue; sort by operating revenue descending. Top enterprise by operating revenue: Dexi Jinjin Intelligent Electrical Company (bmCode: 451895), operating revenue 142,422,544,758.90 yuan.",
+    "From company_profile.csv, filter all enterprise records with province=\"Shandong Province\" and industry=\"Pharmaceutical Manufacturing\", extract company name and bmCode; 22 Shandong pharmaceutical manufacturing enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by company name and bmCode, extract net profit; sort by net profit descending. Top enterprise by net profit: Jianming Anyuan Medical Technology Company (bmCode: 199134), net profit 2,950,153,469.00 yuan.",
+    "Checked all fields in company_operation_status.csv, company_profile.csv, company_core.csv, etc.; no tax payment related fields found. Since the data files do not contain tax payment information, the tax payments of the two enterprises cannot be compared."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Tax payment field exists": false
+  },
+  "answer": "No relevant data found",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium017_result.json b/assets/qa_raw/comprehensive_decision/medium017_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4fca9974b719536bc7a8e1f197f988b9aaba7548
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium017_result.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium017",
+  "question": "In 2022, the industry where Zhongbai Jinmao Chain Company operates, is the enterprise with the highest year-on-year R&D expenditure growth rate also the one with the highest R&D expenditure?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From company_profile.csv, find the record for Zhongbai Jinmao Chain Company, extract industry field; industry is \"Wholesale and Retail Trade\". From company_profile.csv, filter all enterprise records with industry=\"Wholesale and Retail Trade\", extract company name and bmCode; 273 wholesale and retail enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by company name and bmCode, extract R&D expenditure and year-on-year R&D expenditure growth rate; 2022 data found for 273 enterprises.",
+    "Filter enterprise records with non-null year-on-year R&D expenditure growth rate (129 records); sort all enterprises by year-on-year R&D expenditure growth rate descending. Enterprise with highest year-on-year R&D expenditure growth rate: Yonghui Zesheng Chain Company, growth rate 2221.3%, R&D expenditure 516,766.56 yuan.",
+    "Filter enterprise records with non-null R&D expenditure; sort all enterprises by R&D expenditure descending. Enterprise with highest R&D expenditure: Bubusheng Jin Commerce Company, R&D expenditure 2,800,235,364.00 yuan, year-on-year R&D expenditure growth rate 11.8%.",
+    "Compare the enterprise with highest year-on-year R&D expenditure growth rate (Yonghui Zesheng Chain Company) and the enterprise with highest R&D expenditure (Bubusheng Jin Commerce Company); they are not the same enterprise, so the answer is \"No\"."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Enterprise with highest year-on-year R&D expenditure growth rate": "Yonghui Zesheng Chain Company",
+    "Highest growth rate enterprise YoY R&D expenditure growth rate (%)": 2221.3,
+    "Highest growth rate enterprise R&D expenditure (yuan)": 516766.56,
+    "Enterprise with highest R&D expenditure": "Bubusheng Jin Commerce Company",
+    "Highest R&D enterprise R&D expenditure (yuan)": 2800235364.0,
+    "Highest R&D enterprise YoY R&D expenditure growth rate (%)": 11.8,
+    "Is same enterprise": false
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium018_result.json b/assets/qa_raw/comprehensive_decision/medium018_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5756d1f525015aef740c7fef6c92e38b8f74b181
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium018_result.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium018",
+  "question": "In 2022 Nationwide, is the province with the highest R&D expenditure growth rate also the province with the lowest R&D expenditure?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From regional_industry_status.csv, read R&D expenditure data for all provinces, including province, total R&D expenditure, average year-on-year R&D expenditure growth rate, etc.",
+    "Aggregate total R&D expenditure and average year-on-year R&D expenditure growth rate by province; exclude provinces with total R&D expenditure of 0 (missing data).",
+    "Identify the province with the highest average year-on-year R&D expenditure growth rate: Hong Kong Special Administrative Region, R&D expenditure growth rate 212.35%, total R&D expenditure 14,248,827,635.96 yuan.",
+    "Identify the province with the lowest total R&D expenditure (excluding 0): Jilin Province, total R&D expenditure 8,574,294,268.49 yuan, R&D expenditure growth rate 16.95%.",
+    "Compare whether the province with the highest R&D expenditure growth rate (Hong Kong SAR) and the province with the lowest R&D expenditure (Jilin Province) are the same province.Conclusion: The province with the highest R&D expenditure growth rate is not the province with the lowest R&D expenditure; the answer is No."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province with highest R&D expenditure growth rate": "Hong Kong Special Administrative Region",
+    "Highest R&D expenditure growth rate (%)": 212.35,
+    "Highest growth province total R&D expenditure (yuan)": 14248827635.96,
+    "Province with lowest R&D expenditure": "Jilin Province",
+    "Lowest R&D expenditure (yuan)": 8574294268.49,
+    "Lowest R&D province R&D expenditure growth rate (%)": 16.95,
+    "Is same province": false
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium019_result.json b/assets/qa_raw/comprehensive_decision/medium019_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0366ffb23ad713870cd143e255a2e4e379bec603
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium019_result.json
@@ -0,0 +1,30 @@
+{
+  "id": "medium019",
+  "question": "In 2022, Sichuan Province, is Zhongbai Jinmao Chain Company's R&D expenditure higher than the R&D expenditure of the enterprise ranked 15th nationwide in its industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From company_profile.csv, find the record for company name=\"Zhongbai Jinmao Chain Company\", extract industry and province; industry is \"Wholesale and Retail Trade\", province is \"Sichuan Province\".",
+    "From company_operation_status.csv, filter 2022 data for company name=\"Zhongbai Jinmao Chain Company\", extract R&D expenditure; Zhongbai Jinmao Chain Company's R&D expenditure is 11,270,987.0 yuan.",
+    "From company_profile.csv, filter all enterprise records with industry=\"Wholesale and Retail Trade\", extract company name or bmCode; 273 wholesale and retail enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by company name or bmCode, extract R&D expenditure; 2022 R&D expenditure data found for 273 wholesale and retail enterprises.",
+    "Filter enterprise records with non-null R&D expenditure; sort all enterprises by R&D expenditure descending to determine R&D expenditure ranking. Enterprise ranked 15th: Lianhua Tongze Commerce Company, R&D expenditure 265,616,054.7 yuan, province Beijing. Compare Zhongbai Jinmao Chain Company's R&D expenditure (11,270,987.0 yuan) with the 15th-ranked nationwide wholesale and retail enterprise's R&D expenditure (265,616,054.7 yuan). Since 11,270,987.0 < 265,616,054.7, the answer is No."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found Zhongbai Jinmao Chain Company's record in company_profile.csv; confirmed industry \"Wholesale and Retail Trade\", province \"Sichuan Province\".",
+    "Found Zhongbai Jinmao Chain Company's 2022 R&D expenditure of 11,270,987.0 yuan in company_operation_status.csv.",
+    "Found 273 wholesale and retail enterprises in company_profile.csv, including company name or bmCode.",
+    "Found 2022 R&D expenditure data for 273 wholesale and retail enterprises in company_operation_status.csv."
+  ],
+  "milestone": {
+    "Zhongbai Jinmao Chain Company R&D expenditure (yuan)": 11270987.0,
+    "15th-ranked nationwide wholesale and retail enterprise R&D expenditure (yuan)": 265616054.7,
+    "Comparison result": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium020_result.json b/assets/qa_raw/comprehensive_decision/medium020_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ada4a4fcfa8a8c15414f67afa006b5077f2f9766
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium020_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium020",
+  "question": "In 2022, the chemical raw materials and chemical products manufacturing industry in Shandong Province, is the market capitalization of the leading enterprises by operating revenue among the top 3 in this industry nationwide?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From company_profile.csv, filter all enterprise records with province=\"Shandong Province\" and industry=\"Chemical raw materials and chemical products manufacturing\", extract company names, identify enterprises in the chemical raw materials and chemical products manufacturing industry in Shandong Province.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by bmCode or company name, extract operating revenue amount and company market capitalization, obtain operating revenue and market capitalization for each enterprise.",
+    "Filter enterprises with non-null operating revenue amount and company market capitalization, sort by operating revenue amount descending, determine the leading enterprise by operating revenue as \"Hengyi Changhua Fine Chemical Company\" with operating revenue of 165,565,462,711.66 yuan and company market capitalization of 2,754.0.",
+    "From company_profile.csv, filter all enterprise records with industry=\"Chemical raw materials and chemical products manufacturing\", from company_operation_status.csv obtain 2022 company market capitalization for these enterprises by bmCode or company name.",
+    "Filter enterprises with non-null company market capitalization, sort by company market capitalization descending, determine nationwide market capitalization ranking for this industry; identify top 3: No.1 \"Hengyi Changhua Fine Chemical Company\" (Shandong Province) 2,754.0, No.2 \"Rongsheng Jinsheng Chemical Company\" (Qinghai Province) 1,040.0, No.3 \"Hengyi Yuanjin Fine Chemical Company\" (Ningxia Hui Autonomous Region) 927.0. Determine whether the Shandong Province revenue-leading enterprise \"Hengyi Changhua Fine Chemical Company\" ranks among the nationwide top 3 by market capitalization: this enterprise is No.1 nationwide by market capitalization, thus it is among the top 3, conclusion: Yes."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found 41 enterprises in chemical raw materials and chemical products manufacturing in Shandong Province from company_profile.csv, along with their bmCode or names.",
+    "Found 2022 operating revenue amount and company market capitalization data for these enterprises from company_operation_status.csv.",
+    "Found nationwide chemical raw materials and chemical products manufacturing enterprises from company_profile.csv, along with their bmCode or names.",
+    "Found their 2022 company market capitalization data from company_operation_status.csv."
+  ],
+  "milestone": {
+    "Shandong Province operating revenue leading enterprise name": "Hengyi Changhua Fine Chemical Company",
+    "Shandong Province operating revenue leading enterprise operating revenue (yuan)": 165565462711.66,
+    "Shandong Province operating revenue leading enterprise company market capitalization": 2754.0,
+    "Chemical raw materials and chemical products manufacturing nationwide market cap No.1 enterprise": "Hengyi Changhua Fine Chemical Company",
+    "Chemical raw materials and chemical products manufacturing nationwide market cap No.3 enterprise": "Hengyi Yuanjin Fine Chemical Company",
+    "Chemical raw materials and chemical products manufacturing nationwide market cap No.3 enterprise market capitalization": 927.0,
+    "Comparison result": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium021_result.json b/assets/qa_raw/comprehensive_decision/medium021_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..178dbf20ea08f3f28e91f7f2b59986099104f06e
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium021_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium021",
+  "question": "In 2022, the chemical raw materials and chemical products manufacturing industry in Shandong Province, is the market capitalization of the enterprise with the highest operating revenue among the top 3 in this industry nationwide?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From company_profile.csv, filter all enterprise records with province=\"Shandong Province\" and industry=\"Chemical raw materials and chemical products manufacturing\", extract company names, identify enterprises in the chemical raw materials and chemical products manufacturing industry in Shandong Province.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by bmCode or company name, extract operating revenue amount and company market capitalization, obtain operating revenue and market capitalization for each enterprise.",
+    "Filter enterprises with non-null operating revenue amount and company market capitalization, sort by operating revenue amount descending, determine the leading enterprise by operating revenue as \"Hengyi Changhua Fine Chemical Company\" with operating revenue of 165,565,462,711.66 yuan and company market capitalization of 2,754.0.",
+    "From company_profile.csv, filter all enterprise records with industry=\"Chemical raw materials and chemical products manufacturing\", from company_operation_status.csv obtain 2022 company market capitalization for these enterprises by bmCode or company name.",
+    "Filter enterprises with non-null company market capitalization, sort by company market capitalization descending, determine nationwide market capitalization ranking for this industry; identify top 3: No.1 \"Hengyi Changhua Fine Chemical Company\" (Shandong Province) 2,754.0, No.2 \"Rongsheng Jinsheng Chemical Company\" (Qinghai Province) 1,040.0, No.3 \"Hengyi Yuanjin Fine Chemical Company\" (Ningxia Hui Autonomous Region) 927.0. Determine whether the Shandong Province revenue-leading enterprise \"Hengyi Changhua Fine Chemical Company\" ranks among the nationwide top 3 by market capitalization: this enterprise is No.1 nationwide by market capitalization, thus it is among the top 3, conclusion: Yes."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found 41 enterprises in chemical raw materials and chemical products manufacturing in Shandong Province from company_profile.csv, along with their bmCode or names.",
+    "Found 2022 operating revenue amount and company market capitalization data for these enterprises from company_operation_status.csv.",
+    "Found nationwide chemical raw materials and chemical products manufacturing enterprises from company_profile.csv, along with their bmCode or names.",
+    "Found their 2022 company market capitalization data from company_operation_status.csv."
+  ],
+  "milestone": {
+    "Shandong Province operating revenue leading enterprise operating revenue (yuan)": 165565462711.66,
+    "Shandong Province operating revenue leading enterprise company market capitalization": 2754.0,
+    "Chemical raw materials and chemical products manufacturing nationwide market cap No.3 enterprise market capitalization": 927.0,
+    "Comparison result": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium022_result.json b/assets/qa_raw/comprehensive_decision/medium022_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..32fc8e8de69f21f7e74b413c43938d1c11235fd1
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium022_result.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium022",
+  "question": "In 2022 is the region with the highest average R&D investment growth rate also the one with the most policy quantity?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From regional_industry_status.csv, filter all provincial records, extract province and year-on-year R&D investment growth rate mean fields, obtain R&D investment growth rate data for each province and industry.",
+    "Aggregate R&D investment growth rate data by province, calculate the average R&D investment growth rate for each province (average of year-on-year R&D investment growth rate mean across all industries).",
+    "Identify the region with the highest average R&D investment growth rate as \"Hong Kong Special Administrative Region\" with an average R&D investment growth rate of 212.35%.",
+    "From policy_release_status.csv, filter all provincial records, count policy quantity by province, obtain policy quantity for each province.",
+    "Identify the region with the most policies as \"Guangdong Province\" with 59 policies. Compare the region with the highest average R&D investment growth rate (Hong Kong Special Administrative Region) with the region with the most policies (Guangdong Province), determine they are not the same, conclusion: No."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found year-on-year R&D investment growth rate mean data for each province and industry from regional_industry_status.csv.",
+    "Found policy data for each province from policy_release_status.csv."
+  ],
+  "milestone": {
+    "Region with highest average R&D investment growth rate": "Hong Kong Special Administrative Region",
+    "Region with most policies": "Guangdong Province",
+    "Comparison result": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium023_result.json b/assets/qa_raw/comprehensive_decision/medium023_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..8ccb11edf29affb424ccafb02a941c9c59d6a265
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium023_result.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium023",
+  "question": "In 2022 is the average year-on-year net profit growth rate for the industry where Haishan Changgong Equipment Company operates higher than the average R&D investment growth rate?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From company_profile.csv, find the record with company name=\"Haishan Changgong Equipment Company\", extract the industry field, determine its industry as \"General equipment manufacturing\".",
+    "From national_industry_status.csv, filter records with industry=\"General equipment manufacturing\" and district=\"National\", extract net profit year-on-year growth rate mean and R&D investment year-on-year growth rate mean fields.",
+    "Obtain the net profit year-on-year growth rate mean for general equipment manufacturing as -79.47492958%, and the R&D investment year-on-year growth rate mean as 10.1399523809524%.",
+    "Compare the average year-on-year net profit growth rate (-79.47492958%) with the R&D investment growth rate (10.1399523809524%), determine that -79.47492958 is less than 10.1399523809524, conclusion: No."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "Found information for Haishan Changgong Equipment Company from company_profile.csv, confirmed its industry as \"General equipment manufacturing\".",
+    "Found net profit year-on-year growth rate mean and R&D investment year-on-year growth rate mean for national general equipment manufacturing data from national_industry_status.csv."
+  ],
+  "milestone": {
+    "Industry where Haishan Changgong Equipment Company operates": "General equipment manufacturing",
+    "General equipment manufacturing net profit year-on-year growth rate mean (%)": -79.47492958,
+    "General equipment manufacturing R&D investment year-on-year growth rate mean (%)": 10.1399523809524,
+    "Comparison result": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium024_result.json b/assets/qa_raw/comprehensive_decision/medium024_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..6ddb9ba613a0a6a886f50f746aa88578139d7090
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium024_result.json
@@ -0,0 +1,27 @@
+{
+  "id": "medium024",
+  "question": "In 2022, is the region with the highest total operating revenue nationwide also the one with the largest total enterprise market capitalization?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, with no additional text. If relevant data cannot be found, respond with \"No relevant data found\".",
+  "steps": [
+    "From regional_industry_status.csv, filter all provincial records (excluding \"National\"), extract province, total operating revenue amount, and total company market capitalization fields.",
+    "Aggregate total operating revenue amount (yuan) across all industries by province, obtain total operating revenue amount (yuan) for each province. Note: operating revenue amount in the file is in yuan, no conversion needed.",
+    "Identify the region with the highest total operating revenue amount (yuan) as \"Beijing\" with total operating revenue of 54,657,099,005,195.0 yuan.",
+    "Convert total company market capitalization from 10,000 yuan unit to yuan unit (multiply by 10,000), aggregate total company market capitalization (yuan) across all industries by province, obtain total enterprise market capitalization (yuan) for each province.",
+    "Identify the region with the largest total enterprise market capitalization (yuan) as \"Beijing\" with total enterprise market capitalization of 3,959,736,000.0 yuan. Compare the region with the highest total operating revenue amount (yuan) (Beijing) with the region with the largest total enterprise market capitalization (yuan) (Beijing), determine they are the same, conclusion: Yes."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found total operating revenue amount (yuan) and total company market capitalization (10,000 yuan) data for each province and industry from regional_industry_status.csv."
+  ],
+  "milestone": {
+    "Region with highest total operating revenue amount (yuan)": "Beijing",
+    "Region with largest total enterprise market capitalization (yuan)": "Beijing",
+    "Comparison result": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium025_result.json b/assets/qa_raw/comprehensive_decision/medium025_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f14b24ae5491e2d79b6187bcd6a4e114c147214
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium025_result.json
@@ -0,0 +1,29 @@
+{
+  "id": "medium025",
+  "question": "In 2022, is the industry with the highest R&D investment growth rate also the industry with the largest R&D investment?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Only output the answer, do not add any explanation. If the relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all records with district=\"National\" from national_industry_status.csv, extract the fields: industry, year-over-year change in R&D investment (average), and total R&D investment amount.",
+    "Filter industry records where both the year-over-year change in R&D investment (average) and total R&D investment amount are not empty (notna). Note: The total R&D investment amount in the file is in yuan (CNY), no conversion is needed.",
+    "Sort all industries by the year-over-year change in R&D investment (average) field in descending order. The industry with the highest R&D investment growth rate is \"Information transmission, software and information technology services\", with an R&D investment growth rate of 179.78%.",
+    "Sort all industries by the total R&D investment amount field in descending order. The industry with the largest total R&D investment amount is \"Information transmission, software and information technology services\", with a total R&D investment amount of 616535246522.23 yuan.",
+    "Compare the industry with the highest R&D investment growth rate (Information transmission, software and information technology services) with the industry with the largest total R&D investment amount (Information transmission, software and information technology services). Conclude that they are the same, so the answer is: Yes."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found year-over-year change in R&D investment (average) and total R&D investment amount (in yuan) for national data of each industry from national_industry_status.csv."
+  ],
+  "milestone": {
+    "Industry with the highest R&D investment growth rate": "Information transmission, software and information technology services",
+    "Maximum R&D investment growth rate (%)": 179.78,
+    "Industry with the largest total R&D investment amount": "Information transmission, software and information technology services",
+    "Maximum total R&D investment amount (yuan)": 616535246522.23,
+    "Comparison result": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium026_result.json b/assets/qa_raw/comprehensive_decision/medium026_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0cd835b793d9745e9d87a35f1017d8e1ea75e285
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium026_result.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium026",
+  "question": "In 2022, does the industry with the highest average asset-liability ratio in Guangdong Province belong to the industry with the highest average asset-liability ratio nationwide?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Only output the answer, do not add any explanation. If the relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all records with province=\"Guangdong Province\" from regional_industry_status.csv, extract the fields: industry and average asset-liability ratio.",
+    "Filter industry records where the average asset-liability ratio is not empty. Sort by the average asset-liability ratio field in descending order. The industry with the highest average asset-liability ratio in Guangdong Province is \"Textile, footwear and apparel industry\", with an average asset-liability ratio of 724.30%.",
+    "Filter all records with district=\"National\" from national_industry_status.csv, extract the fields: industry and average asset-liability ratio.",
+    "Filter industry records where the average asset-liability ratio is not empty. Sort by the average asset-liability ratio field in descending order. The industry with the highest average asset-liability ratio nationwide is \"Financial industry\", with an average asset-liability ratio of 518.42%.",
+    "Compare the industry with the highest average asset-liability ratio in Guangdong Province (Textile, footwear and apparel industry) with the industry with the highest average asset-liability ratio nationwide (Financial industry). Conclude that they are not the same, so the answer is: No."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found average asset-liability ratio data for industries in Guangdong Province from regional_industry_status.csv.",
+    "Found average asset-liability ratio data for national data of each industry from national_industry_status.csv."
+  ],
+  "milestone": {
+    "Industry with the highest average asset-liability ratio in Guangdong Province": "Textile, footwear and apparel industry",
+    "Industry with the highest average asset-liability ratio nationwide": "Financial industry",
+    "Comparison result": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium027_result.json b/assets/qa_raw/comprehensive_decision/medium027_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..993def3e5afa225e87c6c4d125bc7fca654b46ad
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium027_result.json
@@ -0,0 +1,28 @@
+{
+  "id": "medium027",
+  "question": "In 2022, Compare the enterprise with the highest R&D input-output ratio in Guangdong Province and the enterprise with the highest R&D input-output ratio in Sichuan Province (R&D input-output ratio = operating revenue amount / R&D investment amount). Which enterprise has higher operating revenue? The unit of operating revenue is yuan.",
+  "guidelines": "The answer must be \"yes\" or \"no\". Output only the answer, without any additional explanation. If the relevant data cannot be found, please answer \"No relevant data retrieved\".",
+  "steps": [
+    "Filter all enterprise records with province=\"Guangdong Province\" and province=\"Sichuan Province\" from company_profile.csv, and extract the enterprise name and province fields.",
+    "Filter all records with year=2022 from company_operation_status.csv, and extract the enterprise name, operating revenue amount, and R&D investment amount fields. Note: The operating revenue amount in the file is in yuan; no conversion is needed.",
+    "Filter enterprise records where both operating revenue amount and R&D investment amount are not null and R&D investment amount is greater than 0. Calculate R&D input-output ratio = operating revenue amount / R&D investment amount.",
+    "Identify the enterprise with the highest R&D input-output ratio in Guangdong Province as \"Wulichangyuan Wholesale Company\", with an R&D input-output ratio of 8100.17 and operating revenue of 73,443,148,744.17 yuan.",
+    "Identify the enterprise with the highest R&D input-output ratio in Sichuan Province as \"Guotouzeyuan New Energy Company\", with an R&D input-output ratio of 56708.37 and operating revenue of 2,377,118,263.90 yuan. Compare the enterprise with the highest R&D input-output ratio in Guangdong Province (Wulichangyuan Wholesale Company, operating revenue 73,443,148,744.17 yuan) with the enterprise with the highest R&D input-output ratio in Sichuan Province (Guotouzeyuan New Energy Company, operating revenue 2,377,118,263.90 yuan), and determine that the Guangdong Province enterprise has higher operating revenue."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found enterprise information for Guangdong Province and Sichuan Province from company_profile.csv.",
+    "Found operating revenue amount (in yuan) and R&D investment amount for enterprises in 2022 from company_operation_status.csv."
+  ],
+  "milestone": {
+    "Enterprise with highest R&D input-output ratio in Guangdong Province": "Wulichangyuan Wholesale Company",
+    "Enterprise with highest R&D input-output ratio in Sichuan Province": "Guotouzeyuan New Energy Company",
+    "Enterprise with higher operating revenue": "Enterprise with highest R&D input-output ratio in Guangdong Province (Wulichangyuan Wholesale Company)"
+  },
+  "answer": "The enterprise with the highest R&D input-output ratio in Guangdong Province (Wulichangyuan Wholesale Company) has higher operating revenue",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium028_result.json b/assets/qa_raw/comprehensive_decision/medium028_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4bc4fa0d739a05cb76d06e3799020ef8f2daa6e6
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium028_result.json
@@ -0,0 +1,63 @@
+{
+  "id": "medium028",
+  "question": "In 2022, for the enterprise ranked first in year-over-year growth rate of R&D investment in the chemical raw materials and chemical products manufacturing industry, which of its indicators perform below the industry average?",
+  "guidelines": "Output only the indicator name(s) as the answer, without any additional explanation. If the relevant data cannot be found, please answer \"No relevant data retrieved\"",
+  "steps": [
+    "Filter all enterprise records with industry=\"chemical raw materials and chemical products manufacturing\" from company_profile.csv, and extract the enterprise name and industry fields.",
+    "Filter all records with year=2022 from company_operation_status.csv, and extract the enterprise name, year-over-year change in R&D investment, and various indicator fields. Note: All amount-related indicators are in yuan; no conversion is needed.",
+    "Filter enterprise records where the year-over-year change in R&D investment is not null, sort by the year-over-year change in R&D investment in descending order, and identify the enterprise ranked first in R&D investment year-over-year growth rate as \"Huayichangze Technology Co., Ltd.\", with an R&D investment year-over-year growth rate of 2766.85%.",
+    "Calculate the average of each indicator for all enterprises in the chemical raw materials and chemical products manufacturing industry (excluding the year-over-year change in R&D investment itself).",
+    "Compare each indicator of this enterprise (Huayichangze Technology Co., Ltd.) with the industry average, and identify the indicators that are below the industry average. A total of 17 indicators were found: total assets, R&D investment ratio, capitalized R&D investment, year-over-year change in capitalized R&D investment, company market capitalization, cumulative China patent applications, cumulative China invention patent grants, cumulative China invention patent invalidations, annual China patent applications, annual China invention patent applications, annual China patent grants, annual China invention patent grants, R&D personnel ratio, total liabilities, cumulative citations of all patents, cumulative China invention patent applications, participation in drafting national standards."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Found enterprise information for the chemical raw materials and chemical products manufacturing industry from company_profile.csv.",
+    "Found year-over-year change in R&D investment and various indicator data for enterprises in 2022 from company_operation_status.csv (amount-related indicators are in yuan)."
+  ],
+  "milestone": {
+    "Enterprise ranked first in R&D investment year-over-year growth rate": "Huayichangze Technology Co., Ltd.",
+    "Indicators below industry average": [
+      "Total assets",
+      "R&D investment ratio",
+      "Capitalized R&D investment",
+      "Year-over-year change in capitalized R&D investment",
+      "Company market capitalization",
+      "Cumulative China patent applications",
+      "Cumulative China invention patent grants",
+      "Cumulative China invention patent invalidations",
+      "Annual China patent applications",
+      "Annual China invention patent applications",
+      "Annual China patent grants",
+      "Annual China invention patent grants",
+      "R&D personnel ratio",
+      "Total liabilities",
+      "Cumulative citations of all patents",
+      "Cumulative China invention patent applications",
+      "Participation in drafting national standards"
+    ]
+  },
+  "answer": [
+    "Total assets",
+    "R&D investment ratio",
+    "Capitalized R&D investment",
+    "Year-over-year change in capitalized R&D investment",
+    "Company market capitalization",
+    "Cumulative China patent applications",
+    "Cumulative China invention patent grants",
+    "Cumulative China invention patent invalidations",
+    "Annual China patent applications",
+    "Annual China invention patent applications",
+    "Annual China patent grants",
+    "Annual China invention patent grants",
+    "R&D personnel ratio",
+    "Total liabilities",
+    "Cumulative citations of all patents",
+    "Cumulative China invention patent applications",
+    "Participation in drafting national standards"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium029_result.json b/assets/qa_raw/comprehensive_decision/medium029_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c7980a51659ee8da6a6003f79c1fa7c0423e3612
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium029_result.json
@@ -0,0 +1,47 @@
+{
+  "id": "medium029",
+  "question": "Total number of all enterprises affected by the policy 'Notice of the General Office of Guangdong Provincial People's Government on Printing and Distributing Several Measures of Guangdong Province for Further Promoting Steady Growth of Industrial Economy' in 2022",
+  "guidelines": "The answer must be \"yes\" or \"no\". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Search policy_resource.csv for policy records whose policy name contains \"Guangdong Province's Promotion of Steady Industrial Economic Growth\" and \"Guangdong Province\" to identify the policy.",
+    "Extract the industries involved from the industry field of the policy record, obtaining 8 industries: Consumer Electronics and Electrical Equipment, Automobile Manufacturing, Textile and Footwear, Special Purpose Equipment Manufacturing, General Purpose Equipment Manufacturing, Communication Transmission Equipment, Semiconductor, and Other Manufacturing.",
+    "Filter all enterprise records with province=\"Guangdong Province\" from company_profile.csv, and extract the province and industry fields.",
+    "Count the number of enterprises in each involved industry in Guangdong Province: Consumer Electronics and Electrical Equipment 150, Automobile Manufacturing 27, Textile and Footwear 33, Special Purpose Equipment Manufacturing 80, General Purpose Equipment Manufacturing 20, Communication Transmission Equipment 38, Semiconductor 54, Other Manufacturing 14.",
+    "Calculate the total number of enterprises across all involved industries: 150+27+33+80+20+38+54+14=416."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "The policy and its 8 involved industries were found in policy_release_status.csv.",
+    "Enterprise data for each industry in Guangdong Province was found in company_profile.csv."
+  ],
+  "milestone": {
+    "Number of industries involved in the policy": 8,
+    "Involved industries": [
+      "Consumer Electronics and Electrical Equipment",
+      "Automobile Manufacturing",
+      "Textile and Footwear",
+      "Special Purpose Equipment Manufacturing",
+      "General Purpose Equipment Manufacturing",
+      "Communication Transmission Equipment",
+      "Semiconductor",
+      "Other Manufacturing"
+    ],
+    "Enterprise count per industry": {
+      "Consumer Electronics and Electrical Equipment": 150,
+      "Automobile Manufacturing": 27,
+      "Textile and Footwear": 33,
+      "Special Purpose Equipment Manufacturing": 80,
+      "General Purpose Equipment Manufacturing": 20,
+      "Communication Transmission Equipment": 38,
+      "Semiconductor": 54,
+      "Other Manufacturing": 14
+    },
+    "Total number of enterprises": 416
+  },
+  "answer": 416,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium030_result.json b/assets/qa_raw/comprehensive_decision/medium030_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b99e6a4f1c10d5e28d621d780df2b91d1575ff8e
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium030_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium030",
+  "question": "Did both the operating revenue and R&D investment of Haishan Changgong Equipment Company and Sansan Daten Heavy Industry Company show an upward trend in 2022?",
+  "guidelines": "The answer must be \"yes\" or \"no\". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all records with enterprise names \"Haishan Changgong Equipment Company\" and \"Sansan Daten Heavy Industry Company\" from company_operation_status.csv, and extract the enterprise name, year, operating revenue amount, R&D investment amount, year-over-year change in operating revenue, and year-over-year change in R&D investment fields. Note: amounts are in yuan and do not require conversion.",
+    "Check the data years for both companies and find that only 2022 data is available; therefore use year-over-year change in operating revenue and year-over-year change in R&D investment to determine trends.",
+    "Analyze Haishan Changgong Equipment Company: year-over-year change in operating revenue is -19.28% (decline), year-over-year change in R&D investment is -14.09% (decline); neither operating revenue nor R&D investment shows an upward trend.",
+    "Analyze Sansan Daten Heavy Industry Company: year-over-year change in operating revenue is -9.41% (decline), year-over-year change in R&D investment is 26.97% (increase); operating revenue does not show an upward trend, but R&D investment does.",
+    "Comprehensive conclusion: Neither operating revenue nor R&D investment of Haishan Changgong Equipment Company shows an upward trend; for Sansan Daten Heavy Industry Company, operating revenue does not show an upward trend (only R&D investment increased). Therefore, it is not the case that both operating revenue and R&D investment of both companies show an upward trend. Conclusion: No."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "2022 data for Haishan Changgong Equipment Company and Sansan Daten Heavy Industry Company (operating revenue amount and R&D investment amount in yuan) as well as year-over-year change in operating revenue and R&D investment were found in company_operation_status.csv."
+  ],
+  "milestone": {
+    "Haishan Changgong Equipment Company year-over-year change in operating revenue (%)": -19.28,
+    "Haishan Changgong Equipment Company year-over-year change in R&D investment (%)": -14.09,
+    "Haishan Changgong Equipment Company trend": "Both operating revenue and R&D investment declined",
+    "Sansan Daten Heavy Industry Company year-over-year change in operating revenue (%)": -9.41,
+    "Sansan Daten Heavy Industry Company year-over-year change in R&D investment (%)": 26.97,
+    "Sansan Daten Heavy Industry Company trend": "Operating revenue declined, R&D investment increased",
+    "Comprehensive conclusion": "Not all showing upward trend"
+  },
+  "answer": "no",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium031_result.json b/assets/qa_raw/comprehensive_decision/medium031_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..af98a2b2971b12c6b0afb9a734c67b7d15975abc
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium031_result.json
@@ -0,0 +1,30 @@
+{
+  "id": "medium031",
+  "question": "In 2022, was Haishan Changgong Equipment Company the only enterprise in its region with both increased operating revenue and R&D investment?",
+  "guidelines": "The answer must be \"yes\" or \"no\". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Look up the record for enterprise name \"Haishan Changgong Equipment Company\" in company_profile.csv, extract the province field, and obtain the region where the enterprise is located as \"Shandong Province\".",
+    "Filter records with enterprise name \"Haishan Changgong Equipment Company\" and year=2022 from company_operation_status.csv, and extract the year-over-year change in operating revenue and year-over-year change in R&D investment fields.",
+    "Analyze Haishan Changgong Equipment Company's data: year-over-year change in operating revenue is -19.28% (decline), year-over-year change in R&D investment is -14.09% (decline); neither operating revenue nor R&D investment increased.",
+    "Since neither operating revenue nor R&D investment of Haishan Changgong Equipment Company increased, it cannot be the only enterprise in its region with both increased operating revenue and R&D investment. Conclusion: No."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "The region where Haishan Changgong Equipment Company is located (Shandong Province) was found in company_profile.csv.",
+    "Year-over-year change in operating revenue and R&D investment data for Haishan Changgong Equipment Company was found in company_operation_status.csv."
+  ],
+  "milestone": {
+    "Region where Haishan Changgong Equipment Company is located": "Shandong Province",
+    "Haishan Changgong Equipment Company year-over-year change in operating revenue (%)": -19.28,
+    "Haishan Changgong Equipment Company year-over-year change in R&D investment (%)": -14.09,
+    "Did operating revenue increase": "No",
+    "Did R&D investment increase": "No",
+    "Conclusion": "No"
+  },
+  "answer": "no",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium032_result.json b/assets/qa_raw/comprehensive_decision/medium032_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..150ddbc8c3eb2aa50c633a954090929526fb18a9
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium032_result.json
@@ -0,0 +1,33 @@
+{
+  "id": "medium032",
+  "question": "In 2022, was Haishan Changgong Equipment Company the only enterprise in its region with both declined operating revenue and R&D investment?",
+  "guidelines": "The answer must be \"yes\" or \"no\". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Look up the record for enterprise name \"Haishan Changgong Equipment Company\" in company_profile.csv, extract the province field, and obtain the region where the enterprise is located as \"Shandong Province\".",
+    "Filter records with enterprise name \"Haishan Changgong Equipment Company\" and year=2022 from company_operation_status.csv, and extract the year-over-year change in operating revenue and year-over-year change in R&D investment fields.",
+    "Analyze Haishan Changgong Equipment Company's data: year-over-year change in operating revenue is -19.28% (decline), year-over-year change in R&D investment is -14.09% (decline); both operating revenue and R&D investment declined.",
+    "Filter all enterprise records with province=\"Shandong Province\" and year=2022 from company_operation_status.csv, and extract the enterprise name, year-over-year change in operating revenue, and year-over-year change in R&D investment fields.",
+    "Count the number of enterprises in Shandong Province with both declined operating revenue and R&D investment (year-over-year change in operating revenue < 0 and year-over-year change in R&D investment < 0); a total of 61 enterprises were found. Determine whether Haishan Changgong Equipment Company is the only one: Since 61 enterprises in Shandong Province have both declined operating revenue and R&D investment, Haishan Changgong Equipment Company is not the only one. Conclusion: No."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "The region where Haishan Changgong Equipment Company is located (Shandong Province) was found in company_profile.csv.",
+    "Year-over-year change in operating revenue and R&D investment data for Haishan Changgong Equipment Company was found in company_operation_status.csv.",
+    "Year-over-year change in operating revenue and R&D investment data for all enterprises in Shandong Province was found in company_operation_status.csv."
+  ],
+  "milestone": {
+    "Region where Haishan Changgong Equipment Company is located": "Shandong Province",
+    "Haishan Changgong Equipment Company year-over-year change in operating revenue (%)": -19.28,
+    "Haishan Changgong Equipment Company year-over-year change in R&D investment (%)": -14.09,
+    "Did operating revenue decline": "Yes",
+    "Did R&D investment decline": "Yes",
+    "Number of enterprises in Shandong Province with both declined operating revenue and R&D investment": 61,
+    "Conclusion": "No"
+  },
+  "answer": "no",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium033_result.json b/assets/qa_raw/comprehensive_decision/medium033_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..44736928f92a9a0b078d56990ce7e91c99454873
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium033_result.json
@@ -0,0 +1,33 @@
+{
+  "id": "medium033",
+  "question": "In the industry with the most invention patent grants in 2022, what is the average asset-liability ratio of enterprises? In which province is this industry concentrated?",
+  "guidelines": "Answer data should retain two decimal places. Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all records with district=\"National\" from national_industry_status.csv, and extract the industry and total cumulative China invention patent grants fields.",
+    "Filter industry records with non-null total cumulative China invention patent grants, sort by total cumulative China invention patent grants in descending order, and identify the industry with the most invention patent grants as \"Consumer Electronics and Electrical Equipment\" with a total of 217,188 cumulative China invention patent grants.",
+    "Filter industry \"Consumer Electronics and Electrical Equipment\" from national_industry_status.csv to obtain the average asset-liability ratio of 85.70%.",
+    "Filter enterprises with industry \"Consumer Electronics and Electrical Equipment\" from company_profile.csv, and count enterprises by region.",
+    "From the regional enterprise counts, determine the province with the maximum number as \"Guangdong Province\" with 150 enterprises."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Total cumulative China invention patent grants for each industry at the national level was found in national_industry_status.csv.",
+    "Industry information for each enterprise was found in company_profile.csv.",
+    "Asset-liability ratio data for enterprises in 2022 was found in company_operation_status.csv."
+  ],
+  "milestone": {
+    "Industry with the most invention patent grants": "Consumer Electronics and Electrical Equipment",
+    "Number of enterprises in this industry": 358,
+    "Average asset-liability ratio (%)": 85.7,
+    "Number of 'Consumer Electronics and Electrical Equipment' enterprises in Guangdong Province": 150
+  },
+  "answer": [
+    85.7,
+    "Guangdong Province"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium034_result.json b/assets/qa_raw/comprehensive_decision/medium034_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..198ca69c7c55ad7990a73494a8ec3807490f5bdc
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium034_result.json
@@ -0,0 +1,39 @@
+{
+  "id": "medium034",
+  "question": "Among the provinces that performed best in cultivating high-tech enterprises in 2022 (measured by the number of high-tech enterprises), which high-tech enterprise has the highest year-over-year growth in average operating revenue?",
+  "guidelines": "The answer must be an enterprise name. Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Identify high-tech enterprises from company_profile.csv: enterprises whose firstClassification and secondClassification fields contain keywords such as \"technology\", \"innovation\", \"R&D\", or whose industry field contains keywords such as \"technology\", \"information\", \"software\", \"electronics\", \"communication\".",
+    "Build a mapping from enterprise name to province from company_profile.csv. Count high-tech enterprises by province and identify the top 3 provinces with the most high-tech enterprises: Guangdong Province (89), Beijing (49), Jiangsu Province (48).",
+    "Filter all records with year=2022 from company_operation_status.csv, and extract the enterprise name and year-over-year change in operating revenue fields.",
+    "Filter enterprise records that are high-tech enterprises and have non-null (notna) year-over-year change in operating revenue. For each province, identify the high-tech enterprise with the highest year-over-year growth in operating revenue: Guangdong Province — \"Kuaike Chuangjin Equipment Company\" (47.76%), Beijing — \"Baoxin Ruanlian Software Company\" (51.65%), Jiangsu Province — \"Langji Ruanchuang Information Technology Company\" (93.78%).",
+    "Compare the enterprises with the highest year-over-year growth in operating revenue across these three provinces, and find the highest as \"Langji Ruanchuang Information Technology Company\" (located in Jiangsu Province, 93.78%)."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Province and classification information (firstClassification, secondClassification, industry) for each enterprise was found in company_profile.csv.",
+    "Year-over-year change in operating revenue data for enterprises in 2022 was found in company_operation_status.csv."
+  ],
+  "milestone": {
+    "Top 3 provinces with the most high-tech enterprises": [
+      "Guangdong Province",
+      "Beijing",
+      "Jiangsu Province"
+    ],
+    "Number of high-tech enterprises in Guangdong Province": 89,
+    "Number of high-tech enterprises in Beijing": 49,
+    "Number of high-tech enterprises in Jiangsu Province": 48,
+    "High-tech enterprise with highest year-over-year growth in operating revenue in Guangdong Province": "Kuaike Chuangjin Equipment Company",
+    "High-tech enterprise with highest year-over-year growth in operating revenue in Beijing": "Baoxin Ruanlian Software Company",
+    "High-tech enterprise with highest year-over-year growth in operating revenue in Jiangsu Province": "Langji Ruanchuang Information Technology Company",
+    "Enterprise with highest year-over-year growth in operating revenue": "Langji Ruanchuang Information Technology Company",
+    "Province where this enterprise is located": "Jiangsu Province",
+    "Year-over-year growth in operating revenue (%)": 93.78
+  },
+  "answer": "Langji Ruanchuang Information Technology Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium035_result.json b/assets/qa_raw/comprehensive_decision/medium035_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..cefea3df186e7c5b74ac88bd8e2f71082b3ca1ee
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium035_result.json
@@ -0,0 +1,42 @@
+{
+  "id": "medium035",
+  "question": "In 2022, which provinces ranked in the top three in terms of market share among enterprises in the cloud computing services sector?",
+  "guidelines": "The answer must be province names. Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Identify cloud computing services enterprises from company_profile.csv: enterprises with secondClassification=\"Internet and cloud computing, big data services\" or containing keywords such as \"cloud computing\", \"cloud services\", \"cloud platform\".",
+    "Build a mapping from enterprise name to province from company_profile.csv. Filter all records with year=2022 from company_operation_status.csv, and extract the enterprise name and operating revenue amount fields.",
+    "Filter enterprise records in the cloud computing services sector with non-null operating revenue amount, and calculate total operating revenue by province.",
+    "Calculate total operating revenue as 81,064,346,518.77 yuan for computing market share percentages. Sort by total operating revenue by province and identify the top 3 provinces by market share: Beijing (93.92%, 3 enterprises), Shanghai (2.43%, 1 enterprise), Guangdong Province (2.13%, 2 enterprises).",
+    "Determine the provinces with leading market share positions as the top 3: Beijing, Shanghai, and Guangdong Province."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Province and classification information (secondClassification) for each enterprise was found in company_profile.csv.",
+    "Operating revenue amount data for enterprises in 2022 was found in company_operation_status.csv."
+  ],
+  "milestone": {
+    "Number of enterprises in cloud computing services sector": 9,
+    "Total operating revenue (yuan)": 81064346518.77,
+    "Top 3 provinces by market share": [
+      "Beijing",
+      "Shanghai",
+      "Guangdong Province"
+    ],
+    "Beijing market share (%)": 93.92,
+    "Number of enterprises in Beijing": 3,
+    "Shanghai market share (%)": 2.43,
+    "Number of enterprises in Shanghai": 1,
+    "Guangdong Province market share (%)": 2.13,
+    "Number of enterprises in Guangdong Province": 2
+  },
+  "answer": [
+    "Beijing",
+    "Shanghai",
+    "Guangdong Province"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium036_result.json b/assets/qa_raw/comprehensive_decision/medium036_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5fac2412a4c9b25a866b51751f7b3ccf9be19b4a
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium036_result.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium036",
+  "question": "In 2022, what percentage of total operating revenue in the market do the top 20% enterprises by R&D investment in Sichuan Province's pharmaceutical manufacturing industry account for?",
+  "guidelines": "The answer must be an exact number with 2 decimal places. Output only the number, do not add units, commas, or any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all enterprise records with province=\"Sichuan Province\" and industry=\"Pharmaceutical Manufacturing\" from company_profile.csv, extract the enterprise name and bmCode fields, and find 15 pharmaceutical manufacturing enterprises in Sichuan Province.",
+    "Filter 2022 data for these enterprises from company_operation_status.csv by enterprise name and bmCode, extract the R&D investment amount and operating revenue amount fields, and find 2022 data for 15 pharmaceutical manufacturing enterprises in Sichuan Province.",
+    "Filter enterprise records with non-null (notna) R&D investment amount and operating revenue amount (15 records total), sort all enterprises by R&D investment amount in descending order, determine R&D investment rankings, and identify the top 20% enterprises as \"Huaren Taize Pharmaceutical Co., Ltd., Fuhe Chenze Biopharmaceutical Company, Yishan Shengchen Medical Technology Company\".",
+    "Extract the operating revenue amount for the 3 enterprises: 19038287881.00, 100067338.00, 3389021585.67 respectively. Calculate the sum of operating revenue for the top 20% enterprises as 22527376804.67.",
+    "Calculate the sum of operating revenue for all 15 pharmaceutical manufacturing enterprises in Sichuan Province as 36382939883.72. Calculate percentage = (sum of operating revenue of top 20% enterprises / sum of operating revenue of all enterprises) × 100%, rounded to two decimal places: 61.92%"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Sum of operating revenue of top 20% enterprises (yuan)": 22527376804.67,
+    "Sum of operating revenue of all enterprises (yuan)": 36382939883.72,
+    "Percentage": 61.92
+  },
+  "answer": 61.92,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium037_result.json b/assets/qa_raw/comprehensive_decision/medium037_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..90573413fccb208c8f0c73f81be96f2d4a317669
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium037_result.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium037",
+  "question": "In 2022, what percentage of total operating revenue in the market do the top enterprises by R&D investment (top 3) in Sichuan Province's pharmaceutical manufacturing industry account for?",
+  "guidelines": "The answer must be an exact number with 2 decimal places. Output only the number, do not add units, commas, or any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all enterprise records with province=\"Sichuan Province\" and industry=\"Pharmaceutical Manufacturing\" from company_profile.csv, extract the enterprise name and bmCode fields, and find 15 pharmaceutical manufacturing enterprises in Sichuan Province.",
+    "Filter 2022 data for these enterprises from company_operation_status.csv by enterprise name and bmCode, extract the R&D investment amount and operating revenue amount fields, and find 2022 data for 15 pharmaceutical manufacturing enterprises in Sichuan Province.",
+    "Filter enterprise records with non-null R&D investment amount and operating revenue amount (15 records total), sort all enterprises by R&D investment amount in descending order, determine R&D investment rankings, and identify the top 20% enterprises as \"Huaren Taize Pharmaceutical Co., Ltd., Fuhe Chenze Biopharmaceutical Company, Yishan Shengchen Medical Technology Company\".",
+    "Extract the operating revenue amount for the 3 enterprises: 19038287881.00, 100067338.00, 3389021585.67 respectively. Calculate the sum of operating revenue for the top 20% enterprises as 22527376804.67.",
+    "Calculate the sum of operating revenue for all 15 pharmaceutical manufacturing enterprises in Sichuan Province as 36382939883.72. Calculate percentage = (sum of operating revenue of top 20% enterprises / sum of operating revenue of all enterprises) × 100%, rounded to two decimal places: 61.92%"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Sum of operating revenue of top 20% enterprises (yuan)": 22527376804.67,
+    "Sum of operating revenue of all enterprises (yuan)": 36382939883.72,
+    "Percentage": 61.92
+  },
+  "answer": 61.92,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium038_result.json b/assets/qa_raw/comprehensive_decision/medium038_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e361c9ab96296fe791f4a0de4aa69cf82eece298
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium038_result.json
@@ -0,0 +1,23 @@
+{
+  "id": "medium038",
+  "question": "In 2022, which enterprise has the highest operating revenue in the same region and industry as Sansan Daten Heavy Industry Company?",
+  "guidelines": "The answer must be an enterprise name. Output only the enterprise name, do not add any other explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Look up the record for Sansan Daten Heavy Industry Company in company_profile.csv, extract the province and industry fields, and determine its region as \"Beijing\" and industry as \"Special Purpose Equipment Manufacturing\".",
+    "Filter all enterprise records with province=\"Beijing\" and industry=\"Special Purpose Equipment Manufacturing\" from company_profile.csv, extract the enterprise name and bmCode fields, and find 38 special purpose equipment manufacturing enterprises in Beijing.",
+    "Filter 2022 data for these enterprises from company_operation_status.csv by enterprise name and bmCode, extract the operating revenue amount field, and find 2022 operating revenue data for 38 special purpose equipment manufacturing enterprises in Beijing.",
+    "Sort all enterprises by operating revenue amount in descending order to determine the operating revenue ranking. Identify the enterprise with the highest operating revenue as \"Lingong Hangteng Heavy Industry Company\" with operating revenue of 80034467972.00 yuan."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Total number of special purpose equipment manufacturing enterprises in Beijing": 38,
+    "Operating revenue of Lingong Hangteng Heavy Industry Company (yuan)": 80034467972.0,
+    "Enterprise with highest operating revenue": "Lingong Hangteng Heavy Industry Company"
+  },
+  "answer": "Lingong Hangteng Heavy Industry Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium039_result.json b/assets/qa_raw/comprehensive_decision/medium039_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..29ce06d374ba83ca82a48a721c8a82f0b50325b2
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium039_result.json
@@ -0,0 +1,56 @@
+{
+  "id": "medium039",
+  "question": "In 2022, list all indicators for which Guangdong Province's Information Transmission, Software and Information Technology Services industry has mean values superior to the national average, and sort them by advantage magnitude from high to low.",
+  "guidelines": "Answer format: [Indicator 1, Indicator 2, Indicator 3, ...]. Output only indicator names, separated by commas and spaces, do not add any other explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter records with industry=\"Information Transmission, Software and Information Technology Services\" and district=\"National\" from national_industry_status.csv, extract all fields containing \"mean\" and their values, and obtain national average data. Total number of indicator fields containing mean is 48.",
+    "Filter records with industry=\"Information Transmission, Software and Information Technology Services\" and province=\"Guangdong Province\" from regional_industry_status.csv, extract all fields containing \"mean\" and their values, and obtain Guangdong Province average data. Total number of indicator fields containing mean is 48.",
+    "Iterate through all fields containing \"mean\", compare Guangdong Province and national values. For each field, if Guangdong Province value is greater than national value, mark it as an advantage indicator. Statistics show 15 indicators where Guangdong Province mean is higher than the national average.",
+    "Sort by advantage magnitude from high to low to obtain 15 advantage indicators: mean of year-over-year change in R&D investment ratio, mean of year-over-year change in net profit, mean of annual PCT patent applications, mean of annual PCT invention patent applications, mean of government incentive funds and subsidies, mean of net profit amount, mean of year-over-year change in capitalized R&D investment, mean of cumulative PCT patent applications, mean of cumulative PCT invention patent applications, mean of cumulative citations of all patents, mean of cumulative China invention patent grants, mean of participation in drafting national standards, mean of R&D personnel ratio, mean of R&D investment ratio, mean of asset-liability ratio."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Total number of national indicator fields containing mean": 48,
+    "Total number of Guangdong Province indicator fields containing mean": 48,
+    "Number of advantage indicators in Guangdong Province": 15,
+    "List of advantage indicators": [
+      "Mean of year-over-year change in R&D investment ratio",
+      "Mean of year-over-year change in net profit",
+      "Mean of annual PCT patent applications",
+      "Mean of annual PCT invention patent applications",
+      "Mean of government incentive funds and subsidies",
+      "Mean of net profit amount",
+      "Mean of year-over-year change in capitalized R&D investment",
+      "Mean of cumulative PCT patent applications",
+      "Mean of cumulative PCT invention patent applications",
+      "Mean of cumulative citations of all patents",
+      "Mean of cumulative China invention patent grants",
+      "Mean of participation in drafting national standards",
+      "Mean of R&D personnel ratio",
+      "Mean of R&D investment ratio",
+      "Mean of asset-liability ratio"
+    ]
+  },
+  "answer": [
+    "Mean of year-over-year change in R&D investment ratio",
+    "Mean of year-over-year change in net profit",
+    "Mean of annual PCT patent applications",
+    "Mean of annual PCT invention patent applications",
+    "Mean of government incentive funds and subsidies",
+    "Mean of net profit amount",
+    "Mean of year-over-year change in capitalized R&D investment",
+    "Mean of cumulative PCT patent applications",
+    "Mean of cumulative PCT invention patent applications",
+    "Mean of cumulative citations of all patents",
+    "Mean of cumulative China invention patent grants",
+    "Mean of participation in drafting national standards",
+    "Mean of R&D personnel ratio",
+    "Mean of R&D investment ratio",
+    "Mean of asset-liability ratio"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium040_result.json b/assets/qa_raw/comprehensive_decision/medium040_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5fc8280ac4ef1a93bd083b2355b76f5a9a1aa640
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium040_result.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium040",
+  "question": "In 2022, in Guangdong Province's Raw Chemical Materials and Chemical Products Manufacturing industry, what share of the market's operating revenue does the company with the highest market capitalization account for?",
+  "guidelines": "The answer must be an exact number, rounded to 2 decimal places. Output only the number, do not add units, commas, or any textual explanation. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from company_profile.csv all enterprise records with province=\"Guangdong Province\" and industry=\"Raw Chemical Materials and Chemical Products Manufacturing\", extract enterprise name and bmCode fields, finding 41 enterprises in Guangdong Province's Raw Chemical Materials and Chemical Products Manufacturing industry.",
+    "Filter from company_operation_status.csv the 2022 data of these enterprises by enterprise name and bmCode fields, extract company market capitalization and operating revenue amount fields, finding 2022 data for 41 enterprises in Guangdong Province's Raw Chemical Materials and Chemical Products Manufacturing industry.",
+    "Filter enterprise records where both company market capitalization and operating revenue amount are not null (notna), totaling 41 records. Sort all enterprises in descending order by company market capitalization to determine market capitalization ranking. The enterprise with the highest market capitalization is \"Hengyi Shengsheng Technology Co., Ltd.\". The operating revenue amount of \"Hengyi Shengsheng Technology Co., Ltd.\" is 22316919995.89 CNY.",
+    "Calculate the total operating revenue of Guangdong Province's Raw Chemical Materials and Chemical Products Manufacturing industry as 101800752670.91 CNY.",
+    "Calculate share = (operating revenue of highest market cap enterprise / total industry operating revenue) × 100%, rounded to 2 decimal places: 21.92%"
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Operating revenue of highest market cap enterprise (CNY)": 22316919995.89,
+    "Total operating revenue of Guangdong Province Raw Chemical Materials and Chemical Products Manufacturing industry (CNY)": 101800752670.91,
+    "Share": 21.92
+  },
+  "answer": 21.92,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium041_result.json b/assets/qa_raw/comprehensive_decision/medium041_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d3f17197a5620de21def64ff9af02862fef76069
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium041_result.json
@@ -0,0 +1,24 @@
+{
+  "id": "medium041",
+  "question": "In 2022, nationwide, is the province with the highest mean market capitalization also the province with the highest total net profit?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, do not add any other explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Extract province information for all enterprises from the province field in company_profile.csv, totaling 6895 enterprise records.",
+    "Extract company market capitalization and net profit amount fields for all enterprises from company_operation_status.csv, totaling 6895 enterprise records. Join company_profile.csv and company_operation_status.csv by enterprise name and bmCode fields to merge the two data sources, resulting in 6895 records.",
+    "Filter enterprise records where company market capitalization is not null and greater than 0, totaling 6885 records. Group by province and calculate the mean market capitalization for each province. Sort by mean market capitalization in descending order. The province with the highest mean market capitalization is Taiwan Province, with a mean market cap of 358.689 billion CNY and 12 enterprises.",
+    "Filter enterprise records where net profit amount is not null, totaling 6895 records. Group by province and calculate the total net profit and mean net profit for each province. Sort by total net profit in descending order. The province with the highest total net profit is Beijing, with a total net profit of 4998709315999.03 CNY and 774 enterprises.",
+    "Compare the province with the highest mean market capitalization (Taiwan Province) and the province with the highest total net profit (Beijing). They are not the same, so the conclusion is: No."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Province with highest mean market capitalization": "Taiwan Province",
+    "Province with highest total net profit": "Beijing",
+    "Comparison result": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium042_result.json b/assets/qa_raw/comprehensive_decision/medium042_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..9dbc9425509388260049295b0bb47956d3083f60
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium042_result.json
@@ -0,0 +1,25 @@
+{
+  "id": "medium042",
+  "question": "In 2022, among enterprises in regions applicable to the \"Notice on Organizing Applications for First Home Purchase Subsidies for Outstanding Young Talents in the Biomedicine Industry\" policy, what is the net profit of the enterprise with the best operating revenue performance?",
+  "guidelines": "The answer must be an exact number, in CNY, rounded to 2 decimal places. Output only the number, do not add units, commas, or any textual explanation. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from policy_release_status.csv for policy name=\"Notice on Organizing Applications for First Home Purchase Subsidies for Outstanding Young Talents in the Biomedicine Industry\", extract province field, and determine that the applicable region is \"Guangdong Province\" and the applicable industry is \"Pharmaceutical Manufacturing\".",
+    "Filter from company_profile.csv all enterprise records with province=\"Guangdong Province\" and industry=\"Pharmaceutical Manufacturing\", extract enterprise name and bmCode fields, finding 51 enterprises in Guangdong Province's Pharmaceutical Manufacturing industry.",
+    "Filter from company_operation_status.csv the 2022 data of the above enterprises by bmCode field, extract operating revenue amount and net profit amount fields, obtaining 51 records of enterprise 2022 operating data.",
+    "Sort all enterprises by operating revenue amount in descending order. The enterprise with the highest operating revenue is \"Yishan Zeyuan Pharmaceutical Co., Ltd.\".",
+    "Extract the net profit amount field of this enterprise, obtaining a net profit of 4253373290.96."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "Applicable region": "Guangdong Province",
+    "Applicable industry": "Pharmaceutical Manufacturing",
+    "Enterprise with highest operating revenue": "Yishan Zeyuan Pharmaceutical Co., Ltd.",
+    "Net profit amount of enterprise with highest operating revenue (CNY)": 4253373290.96
+  },
+  "answer": 4253373290.96,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium043_result.json b/assets/qa_raw/comprehensive_decision/medium043_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..64bcc22625479adad69753dffed999e29cfaa573
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium043_result.json
@@ -0,0 +1,23 @@
+{
+  "id": "medium043",
+  "question": "In 2022, for Pharmaceutical Manufacturing, is the province with the highest total assets also the province with the highest R&D investment?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only \"Yes\" or \"No\", do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from regional_industry_status.csv all provincial records with industry=\"Pharmaceutical Manufacturing\".",
+    "In the filtered results, sort by total assets in descending order. The province with the maximum total assets is \"Beijing\", with total assets of 813813599815.91.",
+    "In the same filtered results, sort by total R&D investment amount in descending order. The province with the maximum total R&D investment amount is \"Beijing\", with total R&D investment of 57722993168.07.",
+    "Compare whether the provinces with the maximum values for both indicators are the same: Beijing == Beijing, they are consistent, so the answer is \"Yes\"."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province with highest total assets in Pharmaceutical Manufacturing": "Beijing",
+    "Province with highest total R&D investment in Pharmaceutical Manufacturing": "Beijing",
+    "Whether the two top provinces are the same": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium044_result.json b/assets/qa_raw/comprehensive_decision/medium044_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..8cc703a1129b358ea3dc84b208d7d88bca88cf51
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium044_result.json
@@ -0,0 +1,25 @@
+{
+  "id": "medium044",
+  "question": "In 2022, in Shandong Province's Pharmaceutical Manufacturing industry, is Haishan Changgong Equipment Company's R&D investment ratio higher than the R&D investment ratio of the 10th-ranked Pharmaceutical Manufacturing enterprise in Hunan Province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Search for records with enterprise name=\"Haishan Changgong Equipment Company\" in company_profile.csv. The enterprise has bmCode=100071, industry=\"General Equipment Manufacturing\", province=\"Shandong Province\".",
+    "Filter from company_operation_status.csv the 2022 data by enterprise name=\"Haishan Changgong Equipment Company\" and bmCode=100071, extract the R&D investment ratio field. Haishan Changgong Equipment Company's R&D investment ratio is 4.33%.",
+    "Filter from company_profile.csv all enterprise records with province=\"Hunan Province\" and industry=\"Pharmaceutical Manufacturing\", extract enterprise name and bmCode fields, finding 11 enterprises in Hunan Province's Pharmaceutical Manufacturing industry.",
+    "Filter from company_operation_status.csv the 2022 data of these enterprises by enterprise name and bmCode fields, extract the R&D investment ratio field. Filter enterprise records where R&D investment ratio is not null, totaling 11 records.",
+    "Sort all enterprises by R&D investment ratio in descending order to determine the R&D investment ranking. The 10th-ranked enterprise is \"Ruiying Taiyuan Medical Equipment Co., Ltd.\" (bmCode=823535), with an R&D investment ratio of 2.87%. Compare Haishan Changgong Equipment Company's R&D investment ratio (4.33%) with the 10th-ranked Pharmaceutical Manufacturing enterprise in Hunan Province's R&D investment ratio (2.87%). Since 4.33% > 2.87%, the conclusion is: Yes."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "R&D investment ratio of Haishan Changgong Equipment Company (%)": 4.33,
+    "10th-ranked enterprise in Hunan Province Pharmaceutical Manufacturing": "Ruiying Taiyuan Medical Equipment Co., Ltd.",
+    "R&D investment ratio of 10th-ranked enterprise in Hunan Province Pharmaceutical Manufacturing (%)": 2.87,
+    "Comparison result": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/comprehensive_decision/medium045_result.json b/assets/qa_raw/comprehensive_decision/medium045_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..26babb9d81192fc548b2c36ef1d774157e7a5126
--- /dev/null
+++ b/assets/qa_raw/comprehensive_decision/medium045_result.json
@@ -0,0 +1,25 @@
+{
+  "id": "medium045",
+  "question": "In 2022, in Sichuan Province, is Zhongbai Jinmao Chain Company's R&D investment higher than the R&D investment of the 15th-ranked enterprise nationwide in its industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Search for records with enterprise name=\"Zhongbai Jinmao Chain Company\" in company_profile.csv. The enterprise has bmCode=100120, industry=\"Wholesale and Retail Trade\", province=\"Sichuan Province\".",
+    "Filter from company_operation_status.csv the 2022 data by enterprise name=\"Zhongbai Jinmao Chain Company\" and bmCode=100120, extract the R&D investment amount field. Zhongbai Jinmao Chain Company's R&D investment amount is 11270987.0 CNY.",
+    "Filter from company_profile.csv all enterprise records with industry=\"Wholesale and Retail Trade\", extract enterprise name and bmCode fields, finding 273 enterprises in the Wholesale and Retail Trade industry.",
+    "Filter from company_operation_status.csv the 2022 data of these enterprises by enterprise name and bmCode fields, extract the R&D investment amount field. Filter enterprise records where R&D investment amount is not null, totaling 143 records.",
+    "Sort all enterprises by R&D investment amount in descending order to determine the R&D investment ranking. The 15th-ranked enterprise is \"Lianhua Tongze Trading Company\", with an R&D investment amount of 265616054.7 CNY. Compare Zhongbai Jinmao Chain Company's R&D investment amount (11270987.0 CNY) with the 15th-ranked nationwide enterprise in Wholesale and Retail Trade (265616054.7 CNY). Since 11270987.0 < 265616054.7, the conclusion is: No."
+  ],
+  "steps_num": 5,
+  "milestone": {
+    "R&D investment amount of Zhongbai Jinmao Chain Company (CNY)": 11270987.0,
+    "15th-ranked enterprise nationwide in Wholesale and Retail Trade": "Lianhua Tongze Trading Company",
+    "R&D investment amount of 15th-ranked enterprise nationwide in Wholesale and Retail Trade (CNY)": 265616054.7,
+    "Comparison result": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "comprehensive_decision"
+  }
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy001_result.json b/assets/qa_raw/enterprise_industry_analysis/easy001_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..dd676a8c1c5e51dff43a3299a31b89d290465433
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy001_result.json
@@ -0,0 +1,39 @@
+{
+  "id": "easy001",
+  "question": "In 2022, which is higher: the total number of citations of all patents of Zhong Ji Da Chang Tong Ye Co., Ltd. or the industry median benchmark?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhong Ji Da Chang Tong Ye Co., Ltd. (2022 dataset value): total citations of all patents = 558",
+    "The company belongs to the Metal Products industry",
+    "Industry benchmark (median) for total citations of all patents in Metal Products = 484"
+  ],
+  "milestone": {
+    "Zhong Ji Da Chang Tong Ye Co., Ltd. total citations of all patents in 2022": 558,
+    "Industry of Zhong Ji Da Chang Tong Ye Co., Ltd.": "Metal Products",
+    "Industry benchmark (median) for total citations of all patents in Metal Products": 484,
+    "Comparison result (whether the company is higher than the industry median)": "Yes"
+  },
+  "answer": "Zhong Ji Da Chang Tong Ye Co., Ltd.",
+  "steps": [
+    "Extracted from company_operation_status.csv: Zhong Ji Da Chang Tong Ye Co., Ltd. total citations of all patents in 2022 = 558",
+    "Extracted from company_profile.csv: the company belongs to Metal Products; extracted from national_industry_status.csv: the industry benchmark (median) is 484",
+    "Compared 558 and 484; since 558 > 484, the company is higher, so output \"Zhong Ji Da Chang Tong Ye Co., Ltd.\""
+  ],
+  "steps_num": 3,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "reference": [
+    {
+      "company_operation_status": "05e7a409-b175-4fb1-8382-54f744f20a46"
+    },
+    {
+      "national_industry_status": "02c91882-76b2-45b4-abcf-3db6194fb2de"
+    },
+    {
+      "company_profile": "b6494485-93ab-4705-ab72-0aa551d2f1d6"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy002_result.json b/assets/qa_raw/enterprise_industry_analysis/easy002_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d7ad7817a1f080fbec4afd5fb5aad9c7c6be4253
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy002_result.json
@@ -0,0 +1,39 @@
+{
+  "id": "easy002",
+  "question": "In 2022, what is the difference between the total liabilities of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. and the median total liabilities of its industry?",
+  "guidelines": "The answer must be a number with three decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. (2022 dataset value): total liabilities = 5309105105.07 CNY",
+    "The company belongs to the Construction industry",
+    "Median total liabilities in the Construction industry = 4815749291.465 CNY"
+  ],
+  "milestone": {
+    "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. total liabilities in 2022 (CNY)": 5309105105.07,
+    "Industry of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd.": "Construction",
+    "Median total liabilities in the Construction industry (CNY)": 4815749291.465,
+    "Difference (company - industry median)": 493355813.605
+  },
+  "answer": 493355813.605,
+  "steps": [
+    "Extracted from company_operation_status.csv: Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. total liabilities in 2022 = 5309105105.07 CNY",
+    "Extracted from company_profile.csv: the company belongs to Construction; extracted from national_industry_status.csv: industry median total liabilities = 4815749291.465 CNY",
+    "Calculated by requirement: difference (company - industry median) = 5309105105.07 - 4815749291.465 = 493355813.605, output with three decimal places."
+  ],
+  "steps_num": 3,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "reference": [
+    {
+      "company_operation_status": "64cdaeda-369d-4d2e-9214-a7e2cb75edb0"
+    },
+    {
+      "national_industry_status": "a3ea618e-b6c2-4ae4-a46d-7d4fafa9aff9"
+    },
+    {
+      "company_profile": "6e6b4e87-8925-4b4b-a3d5-8f4e80099348"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy003_result.json b/assets/qa_raw/enterprise_industry_analysis/easy003_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..203247e15dc4fdcb060075517b44797b2a66a8fb
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy003_result.json
@@ -0,0 +1,39 @@
+{
+  "id": "easy003",
+  "question": "In 2022, which is lower: the year-over-year change rate of operating profit of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. or the average of this indicator in its industry?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. (2022 dataset value): year-over-year operating profit change rate = -263.46 %",
+    "The company belongs to the Construction industry",
+    "Average year-over-year operating profit change rate in the Construction industry = -80.1477027027027 %"
+  ],
+  "milestone": {
+    "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. year-over-year operating profit change rate in 2022": "-263.46 %",
+    "Industry of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd.": "Construction",
+    "Average year-over-year operating profit change rate in the Construction industry": "-80.1477027027027 %",
+    "Comparison result (whether the company is lower than the industry average)": "Yes"
+  },
+  "answer": "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd.",
+  "steps": [
+    "Extracted from company_operation_status.csv: Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. year-over-year operating profit change rate in 2022 = -263.46 %",
+    "Extracted from company_profile.csv: the company belongs to Construction; extracted from national_industry_status.csv: industry average = -80.1477027027027 %",
+    "Compared -263.46 and -80.1477027027027; since -263.46 < -80.1477027027027, the company is lower, so the answer is \"Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd.\"."
+  ],
+  "steps_num": 3,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "reference": [
+    {
+      "company_operation_status": "64cdaeda-369d-4d2e-9214-a7e2cb75edb0"
+    },
+    {
+      "national_industry_status": "a3ea618e-b6c2-4ae4-a46d-7d4fafa9aff9"
+    },
+    {
+      "company_profile": "6e6b4e87-8925-4b4b-a3d5-8f4e80099348"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy004_result.json b/assets/qa_raw/enterprise_industry_analysis/easy004_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..862893b7838c7bf9e90da8357271f90f92a96acb
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy004_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy004",
+  "question": "In 2022, is the year-over-year net profit growth rate of Zhong Tong Jie Tong Yun Shu Co., Ltd. lower than that of Yun Da Hang Chang Kuai Di Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhong Tong Jie Tong Yun Shu Co., Ltd. (2022 dataset value): year-over-year net profit growth rate = -4.46 %",
+    "Yun Da Hang Chang Kuai Di Co., Ltd. (2022 dataset value): year-over-year net profit growth rate = 377.60 %"
+  ],
+  "milestone": {
+    "Year-over-year net profit growth rate of Zhong Tong Jie Tong Yun Shu Co., Ltd. in 2022": "-4.46 %",
+    "Year-over-year net profit growth rate of Yun Da Hang Chang Kuai Di Co., Ltd. in 2022": "377.60 %",
+    "Comparison result (whether Zhong Tong Jie Tong Yun Shu Co., Ltd. is lower than Yun Da Hang Chang Kuai Di Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: year-over-year net profit growth rate of Zhong Tong Jie Tong Yun Shu Co., Ltd. in 2022 = -4.46 %",
+    "Extracted from company_operation_status.csv: year-over-year net profit growth rate of Yun Da Hang Chang Kuai Di Co., Ltd. in 2022 = 377.60 %",
+    "Compared -4.46 and 377.60; since -4.46 < 377.60, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "c6a245c0-c35b-4eab-9338-71e29c7a4ee2"
+    },
+    {
+      "company_operation_status": "621b0592-2646-4b1b-a226-80ae2405cc68"
+    },
+    {
+      "company_profile": "a0149779-8c3a-4625-a26d-e86852928f54"
+    },
+    {
+      "company_profile": "7b91b75d-c41b-4ea8-a396-2f4907e19cf7"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy005_result.json b/assets/qa_raw/enterprise_industry_analysis/easy005_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb877415b1230a752f0648013146210f772adc86
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy005_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy005",
+  "question": "In 2022, is the market capitalization of Zhong Tong Jie Tong Yun Shu Co., Ltd. lower than that of Yun Da Hang Chang Kuai Di Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhong Tong Jie Tong Yun Shu Co., Ltd. (2022 dataset value): market capitalization = 33.0 hundred million CNY",
+    "Yun Da Hang Chang Kuai Di Co., Ltd. (2022 dataset value): market capitalization = 136.0 hundred million CNY"
+  ],
+  "milestone": {
+    "Market capitalization of Zhong Tong Jie Tong Yun Shu Co., Ltd. in 2022 (hundred million CNY)": 33.0,
+    "Market capitalization of Yun Da Hang Chang Kuai Di Co., Ltd. in 2022 (hundred million CNY)": 136.0,
+    "Comparison result (whether Zhong Tong Jie Tong Yun Shu Co., Ltd. is lower than Yun Da Hang Chang Kuai Di Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: market capitalization of Zhong Tong Jie Tong Yun Shu Co., Ltd. in 2022 = 33.0 hundred million CNY",
+    "Extracted from company_operation_status.csv: market capitalization of Yun Da Hang Chang Kuai Di Co., Ltd. in 2022 = 136.0 hundred million CNY",
+    "Compared 33.0 and 136.0; since 33.0 < 136.0, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "c6a245c0-c35b-4eab-9338-71e29c7a4ee2"
+    },
+    {
+      "company_operation_status": "621b0592-2646-4b1b-a226-80ae2405cc68"
+    },
+    {
+      "company_profile": "a0149779-8c3a-4625-a26d-e86852928f54"
+    },
+    {
+      "company_profile": "7b91b75d-c41b-4ea8-a396-2f4907e19cf7"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy006_result.json b/assets/qa_raw/enterprise_industry_analysis/easy006_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c67b88b5f9502141ea1f754941b60c71c24a575f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy006_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy006",
+  "question": "In 2022, is the annual number of authorized Chinese invention patents of Huan Xing Jin Ya Shi Shang Co., Ltd. lower than that of Li Ding Sheng Shang Fang Zhi Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Huan Xing Jin Ya Shi Shang Co., Ltd. (2022 dataset value): annual number of authorized Chinese invention patents = 8",
+    "Li Ding Sheng Shang Fang Zhi Co., Ltd. (2022 dataset value): annual number of authorized Chinese invention patents = 16"
+  ],
+  "milestone": {
+    "Annual number of authorized Chinese invention patents of Huan Xing Jin Ya Shi Shang Co., Ltd. in 2022": 8,
+    "Annual number of authorized Chinese invention patents of Li Ding Sheng Shang Fang Zhi Co., Ltd. in 2022": 16,
+    "Comparison result (whether Huan Xing Jin Ya Shi Shang Co., Ltd. is lower than Li Ding Sheng Shang Fang Zhi Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: annual number of authorized Chinese invention patents of Huan Xing Jin Ya Shi Shang Co., Ltd. in 2022 = 8",
+    "Extracted from company_operation_status.csv: annual number of authorized Chinese invention patents of Li Ding Sheng Shang Fang Zhi Co., Ltd. in 2022 = 16",
+    "Compared 8 and 16; since 8 < 16, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "5d67105a-bf43-417d-90ac-3ede097bafe4"
+    },
+    {
+      "company_operation_status": "6f68c180-c6ce-442f-8974-6544dde16d10"
+    },
+    {
+      "company_profile": "2ce0db8a-6bb6-4faf-964c-e078851990cf"
+    },
+    {
+      "company_profile": "73ee79bd-b25a-42e8-87be-42324114c643"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy007_result.json b/assets/qa_raw/enterprise_industry_analysis/easy007_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f70ef4efd3aff51cdf305677620661a5b2021cd
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy007_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy007",
+  "question": "In 2022, is the number of national standards participated in drafting by Bao Xin Ke Hui Ruan Jian Co., Ltd. the same as that of Zhong Ke Chuang Xin Ruan Jian Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Bao Xin Ke Hui Ruan Jian Co., Ltd. (2022 dataset value): number of national standards participated in drafting = 1",
+    "Zhong Ke Chuang Xin Ruan Jian Co., Ltd. (2022 dataset value): number of national standards participated in drafting = 1"
+  ],
+  "milestone": {
+    "Number of national standards participated in drafting by Bao Xin Ke Hui Ruan Jian Co., Ltd. in 2022": 1,
+    "Number of national standards participated in drafting by Zhong Ke Chuang Xin Ruan Jian Co., Ltd. in 2022": 1,
+    "Comparison result (whether they are the same)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: number of national standards participated in drafting by Bao Xin Ke Hui Ruan Jian Co., Ltd. in 2022 = 1",
+    "Extracted from company_operation_status.csv: number of national standards participated in drafting by Zhong Ke Chuang Xin Ruan Jian Co., Ltd. in 2022 = 1",
+    "Compared 1 and 1; since they are equal, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "8d756dae-ab72-486c-b4b4-051d5c60aded"
+    },
+    {
+      "company_operation_status": "6cd8b136-cff6-4f44-8317-ff6d6d3f5d18"
+    },
+    {
+      "company_profile": "1c1441c7-e95e-4f3e-ac46-37a48236ddbe"
+    },
+    {
+      "company_profile": "64052c09-3d7d-4553-beb2-2f25b7e94648"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy008_result.json b/assets/qa_raw/enterprise_industry_analysis/easy008_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5b1c9736f14ccc23e479ab3e1f665f01d5ce848f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy008_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy008",
+  "question": "In 2022, compared with Zhong Ke Chuang Xin Ruan Jian Co., Ltd., what is the difference in cumulative PCT patent applications of Bao Xin Ke Hui Ruan Jian Co., Ltd.?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number without units, commas, or any text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Bao Xin Ke Hui Ruan Jian Co., Ltd. (2022 dataset value): cumulative PCT patent applications = 7",
+    "Zhong Ke Chuang Xin Ruan Jian Co., Ltd. (2022 dataset value): cumulative PCT patent applications = 490"
+  ],
+  "milestone": {
+    "Cumulative PCT patent applications of Bao Xin Ke Hui Ruan Jian Co., Ltd. in 2022": 7,
+    "Cumulative PCT patent applications of Zhong Ke Chuang Xin Ruan Jian Co., Ltd. in 2022": 490,
+    "Difference (Bao Xin Ke Hui Ruan Jian Co., Ltd. - Zhong Ke Chuang Xin Ruan Jian Co., Ltd.)": -483.0
+  },
+  "answer": -483.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: cumulative PCT patent applications of Bao Xin Ke Hui Ruan Jian Co., Ltd. in 2022 = 7",
+    "Extracted from company_operation_status.csv: cumulative PCT patent applications of Zhong Ke Chuang Xin Ruan Jian Co., Ltd. in 2022 = 490",
+    "Calculated the difference (Bao Xin Ke Hui Ruan Jian Co., Ltd. - Zhong Ke Chuang Xin Ruan Jian Co., Ltd.): 7 - 490 = -483.0, and output with one decimal place as required"
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "8d756dae-ab72-486c-b4b4-051d5c60aded"
+    },
+    {
+      "company_operation_status": "6cd8b136-cff6-4f44-8317-ff6d6d3f5d18"
+    },
+    {
+      "company_profile": "1c1441c7-e95e-4f3e-ac46-37a48236ddbe"
+    },
+    {
+      "company_profile": "64052c09-3d7d-4553-beb2-2f25b7e94648"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy009_result.json b/assets/qa_raw/enterprise_industry_analysis/easy009_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..053503d63e8f5280dd0444f475c5dea57c654df0
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy009_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy009",
+  "question": "In 2022, is the total number of employees of Can Xin Hui Xin Semiconductor Co., Ltd. higher than that of Rui Xin Xin Yao Materials Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Can Xin Hui Xin Semiconductor Co., Ltd. (2022 dataset value): total employees = 7277.0",
+    "Rui Xin Xin Yao Materials Co., Ltd. (2022 dataset value): total employees = 2989.0"
+  ],
+  "milestone": {
+    "Can Xin Hui Xin Semiconductor Co., Ltd. total employees in 2022": 7277.0,
+    "Rui Xin Xin Yao Materials Co., Ltd. total employees in 2022": 2989.0,
+    "Comparison result (whether Can Xin Hui Xin Semiconductor Co., Ltd. is higher than Rui Xin Xin Yao Materials Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Can Xin Hui Xin Semiconductor Co., Ltd. total employees in 2022 = 7277.0",
+    "Extracted from company_operation_status.csv: Rui Xin Xin Yao Materials Co., Ltd. total employees in 2022 = 2989.0",
+    "Compared 7277.0 and 2989.0; since 7277.0 > 2989.0, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "ddbff962-a84a-4b00-ab50-cbd0fd41ab5f"
+    },
+    {
+      "company_operation_status": "7d7e048d-3e3d-4ca6-8a8c-fff1dea07d45"
+    },
+    {
+      "company_profile": "70750f9d-0f29-4e45-acaa-95fdb53e4ec0"
+    },
+    {
+      "company_profile": "4226787a-eeb7-49c5-9b91-57d11e97296e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy010_result.json b/assets/qa_raw/enterprise_industry_analysis/easy010_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7accca55c647b0a7130fbf84c0f68b1788e3bd19
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy010_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy010",
+  "question": "In 2022, is the cumulative number of granted Chinese invention patents of Chuang Wei Yao Yao Dian Qi Co., Ltd. lower than that of Mei Neng Dian Jin Technology Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Chuang Wei Yao Yao Dian Qi Co., Ltd. (2022 dataset value): cumulative granted Chinese invention patents = 59",
+    "Mei Neng Dian Jin Technology Co., Ltd. (2022 dataset value): cumulative granted Chinese invention patents = 140"
+  ],
+  "milestone": {
+    "Chuang Wei Yao Yao Dian Qi Co., Ltd. cumulative granted Chinese invention patents in 2022": 59,
+    "Mei Neng Dian Jin Technology Co., Ltd. cumulative granted Chinese invention patents in 2022": 140,
+    "Comparison result (whether Chuang Wei Yao Yao Dian Qi Co., Ltd. is lower than Mei Neng Dian Jin Technology Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Chuang Wei Yao Yao Dian Qi Co., Ltd. cumulative granted Chinese invention patents in 2022 = 59",
+    "Extracted from company_operation_status.csv: Mei Neng Dian Jin Technology Co., Ltd. cumulative granted Chinese invention patents in 2022 = 140",
+    "Compared 59 and 140; since 59 < 140, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "f1c231b9-3bb2-4ccb-9dea-015c6d03b2ab"
+    },
+    {
+      "company_operation_status": "e8c24716-5362-4b4d-b0fd-7fac6e3faba7"
+    },
+    {
+      "company_profile": "3aae2df1-8ad3-4eed-bba9-92c1e96021af"
+    },
+    {
+      "company_profile": "cfc78306-ad47-4edd-a11f-d441d5797ce0"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy011_result.json b/assets/qa_raw/enterprise_industry_analysis/easy011_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2775668784f7a5a643eded7178a93884f902139f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy011_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy011",
+  "question": "In 2022, is the R&D investment ratio of Yong Feng Xin Ruan Network Co., Ltd. higher than that of Jin Fei Shu Ruan Data Services Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Yong Feng Xin Ruan Network Co., Ltd. (2022 dataset value): R&D investment ratio = 11.7 %",
+    "Jin Fei Shu Ruan Data Services Co., Ltd. (2022 dataset value): R&D investment ratio = 5.85 %"
+  ],
+  "milestone": {
+    "Yong Feng Xin Ruan Network Co., Ltd. R&D investment ratio in 2022": "11.7 %",
+    "Jin Fei Shu Ruan Data Services Co., Ltd. R&D investment ratio in 2022": "5.85 %",
+    "Comparison result (whether Yong Feng Xin Ruan Network Co., Ltd. is higher than Jin Fei Shu Ruan Data Services Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Yong Feng Xin Ruan Network Co., Ltd. R&D investment ratio in 2022 = 11.7 %",
+    "Extracted from company_operation_status.csv: Jin Fei Shu Ruan Data Services Co., Ltd. R&D investment ratio in 2022 = 5.85 %",
+    "Compared 11.7 and 5.85; since 11.7 > 5.85, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "7e4ee9ad-488e-4b1e-957c-ec4ef746b558"
+    },
+    {
+      "company_operation_status": "757b1209-5063-43ce-a251-537667e340a1"
+    },
+    {
+      "company_profile": "5f8d33d8-0528-451e-aaf6-8ea404fd0080"
+    },
+    {
+      "company_profile": "6070c698-de3e-4b37-874c-056225b3214e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy012_result.json b/assets/qa_raw/enterprise_industry_analysis/easy012_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca84792169d01d630c85d9295de5c4c1ecc32c77
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy012_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy012",
+  "question": "In 2022, is the total assets of Yong Feng Xin Ruan Network Co., Ltd. lower than that of Jin Fei Shu Ruan Data Services Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Yong Feng Xin Ruan Network Co., Ltd. (2022 dataset value): total assets = 957348449.13 CNY",
+    "Jin Fei Shu Ruan Data Services Co., Ltd. (2022 dataset value): total assets = 674647621.10 CNY"
+  ],
+  "milestone": {
+    "Yong Feng Xin Ruan Network Co., Ltd. total assets in 2022 (CNY)": 957348449.13,
+    "Jin Fei Shu Ruan Data Services Co., Ltd. total assets in 2022 (CNY)": 674647621.1,
+    "Comparison result (whether Yong Feng Xin Ruan Network Co., Ltd. is lower than Jin Fei Shu Ruan Data Services Co., Ltd.)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Yong Feng Xin Ruan Network Co., Ltd. total assets in 2022 = 957348449.13 CNY",
+    "Extracted from company_operation_status.csv: Jin Fei Shu Ruan Data Services Co., Ltd. total assets in 2022 = 674647621.10 CNY",
+    "Compared 957348449.13 and 674647621.10; since 957348449.13 > 674647621.10, the condition \"lower than\" is false, so the judgment is \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "7e4ee9ad-488e-4b1e-957c-ec4ef746b558"
+    },
+    {
+      "company_operation_status": "757b1209-5063-43ce-a251-537667e340a1"
+    },
+    {
+      "company_profile": "5f8d33d8-0528-451e-aaf6-8ea404fd0080"
+    },
+    {
+      "company_profile": "6070c698-de3e-4b37-874c-056225b3214e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy013_result.json b/assets/qa_raw/enterprise_industry_analysis/easy013_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba351d4772c3212134cdb8e39caccf280ce12ba6
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy013_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy013",
+  "question": "In 2022, is the company market value of Wu Li Chang Yuan Wholesale Co., Ltd. higher than that of Wu Li Hui Jin Retail Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Wu Li Chang Yuan Wholesale Co., Ltd. (2022 dataset value): company market value = 220.0 (100 million CNY)",
+    "Wu Li Hui Jin Retail Co., Ltd. (2022 dataset value): company market value = 0.58 (100 million CNY)"
+  ],
+  "milestone": {
+    "Wu Li Chang Yuan Wholesale Co., Ltd. company market value in 2022 (100 million CNY)": 220.0,
+    "Wu Li Hui Jin Retail Co., Ltd. company market value in 2022 (100 million CNY)": 0.58,
+    "Comparison result (whether Wu Li Chang Yuan Wholesale Co., Ltd. is higher than Wu Li Hui Jin Retail Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Wu Li Chang Yuan Wholesale Co., Ltd. company market value in 2022 = 220.0 (100 million CNY)",
+    "Extracted from company_operation_status.csv: Wu Li Hui Jin Retail Co., Ltd. company market value in 2022 = 0.58 (100 million CNY)",
+    "Compared 220.0 and 0.58; since 220.0 > 0.58, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "328fdefe-18d1-4c41-8462-5fedd4a3fbed"
+    },
+    {
+      "company_operation_status": "e8c22d07-3c1f-4593-9fbf-3a181b520d85"
+    },
+    {
+      "company_profile": "4a50579b-d523-4bca-8dcd-ee5defe6d541"
+    },
+    {
+      "company_profile": "0fb0e21a-ae10-46c2-9c10-20f694a4c9b1"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy014_result.json b/assets/qa_raw/enterprise_industry_analysis/easy014_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..88dfd5fff322d2695315317b3c9393029c55390e
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy014_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy014",
+  "question": "In 2022, is the operating revenue of Wu Li Chang Yuan Pi Fa Co., Ltd. lower than that of Wu Li Hui Jin Ling Shou Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Wu Li Chang Yuan Pi Fa Co., Ltd. (2022 dataset value): operating revenue = 73443148744.17 CNY",
+    "Wu Li Hui Jin Ling Shou Co., Ltd. (2022 dataset value): operating revenue = 1144710941.0 CNY"
+  ],
+  "milestone": {
+    "Operating revenue of Wu Li Chang Yuan Pi Fa Co., Ltd. in 2022 (CNY)": 73443148744.17,
+    "Operating revenue of Wu Li Hui Jin Ling Shou Co., Ltd. in 2022 (CNY)": 1144710941.0,
+    "Comparison result (whether Wu Li Chang Yuan Pi Fa Co., Ltd. is lower than Wu Li Hui Jin Ling Shou Co., Ltd.)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: operating revenue of Wu Li Chang Yuan Pi Fa Co., Ltd. in 2022 = 73443148744.17 CNY",
+    "Extracted from company_operation_status.csv: operating revenue of Wu Li Hui Jin Ling Shou Co., Ltd. in 2022 = 1144710941.0 CNY",
+    "Compared 73443148744.17 and 1144710941.0; since 73443148744.17 > 1144710941.0, the statement \"is lower\" is false, so the judgment is \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "328fdefe-18d1-4c41-8462-5fedd4a3fbed"
+    },
+    {
+      "company_operation_status": "e8c22d07-3c1f-4593-9fbf-3a181b520d85"
+    },
+    {
+      "company_profile": "4a50579b-d523-4bca-8dcd-ee5defe6d541"
+    },
+    {
+      "company_profile": "0fb0e21a-ae10-46c2-9c10-20f694a4c9b1"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy015_result.json b/assets/qa_raw/enterprise_industry_analysis/easy015_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b1cec255782c108233aeb547abbd9baf3b4fead4
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy015_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy015",
+  "question": "In 2022, is the annual number of Chinese patent applications of Mei Neng Xuan Jin Dian Qi Co., Ltd. higher than that of Li Xin Yao Yue Dian Qi Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Mei Neng Xuan Jin Dian Qi Co., Ltd. (2022 dataset value): annual number of Chinese patent applications = 4552",
+    "Li Xin Yao Yue Dian Qi Co., Ltd. (2022 dataset value): annual number of Chinese patent applications = 1203"
+  ],
+  "milestone": {
+    "Annual number of Chinese patent applications of Mei Neng Xuan Jin Dian Qi Co., Ltd. in 2022": 4552,
+    "Annual number of Chinese patent applications of Li Xin Yao Yue Dian Qi Co., Ltd. in 2022": 1203,
+    "Comparison result (whether Mei Neng Xuan Jin Dian Qi Co., Ltd. is higher than Li Xin Yao Yue Dian Qi Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: annual number of Chinese patent applications of Mei Neng Xuan Jin Dian Qi Co., Ltd. in 2022 = 4552",
+    "Extracted from company_operation_status.csv: annual number of Chinese patent applications of Li Xin Yao Yue Dian Qi Co., Ltd. in 2022 = 1203",
+    "Compared 4552 and 1203; since 4552 > 1203, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "22285f57-6f5f-47ec-93d7-1e34cce5fcbc"
+    },
+    {
+      "company_operation_status": "713a02d3-532d-4626-bbf7-8156d9c2c4fc"
+    },
+    {
+      "company_profile": "90e9de1d-a9ed-4432-b6fb-3e488d2ee3c6"
+    },
+    {
+      "company_profile": "485f3a26-1734-471d-b18b-09907d34355b"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy016_result.json b/assets/qa_raw/enterprise_industry_analysis/easy016_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2dbb935d75787feb218bb98cec538bb8851a325e
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy016_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy016",
+  "question": "In 2022, is the annual number of authorized Chinese invention patents of Mei Neng Xuan Jin Dian Qi Co., Ltd. higher than the annual number of Chinese invention patent applications of Hai Li Chuang Yao Jia Dian Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Mei Neng Xuan Jin Dian Qi Co., Ltd. (2022 dataset value): annual number of authorized Chinese invention patents = 3304",
+    "Hai Li Chuang Yao Jia Dian Co., Ltd. (2022 dataset value): annual number of Chinese invention patent applications = 8"
+  ],
+  "milestone": {
+    "Annual number of authorized Chinese invention patents of Mei Neng Xuan Jin Dian Qi Co., Ltd. in 2022": 3304,
+    "Annual number of Chinese invention patent applications of Hai Li Chuang Yao Jia Dian Co., Ltd. in 2022": 8,
+    "Comparison result (whether Mei Neng Xuan Jin Dian Qi Co., Ltd. is higher than Hai Li Chuang Yao Jia Dian Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: annual number of authorized Chinese invention patents of Mei Neng Xuan Jin Dian Qi Co., Ltd. in 2022 = 3304",
+    "Extracted from company_operation_status.csv: annual number of Chinese invention patent applications of Hai Li Chuang Yao Jia Dian Co., Ltd. in 2022 = 8",
+    "Compared 3304 and 8; since 3304 > 8, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "3808263c-c56e-4eeb-b16d-b32302554002"
+    },
+    {
+      "company_operation_status": "1742607a-1367-458d-8bf9-b9850d72a344"
+    },
+    {
+      "company_profile": "90e9de1d-a9ed-4432-b6fb-3e488d2ee3c6"
+    },
+    {
+      "company_profile": "cbcdc5e9-a16e-41ca-9f5c-3b29d0902389"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy017_result.json b/assets/qa_raw/enterprise_industry_analysis/easy017_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5218a4c64483fb90fb6110d4c2980651800affbd
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy017_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy017",
+  "question": "In 2022, is the number of provincial or ministerial science and technology progress awards of Mei Neng Xuan Jin Dian Qi Co., Ltd. lower than that of Hai Li Chuang Yao Jia Dian Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Mei Neng Xuan Jin Dian Qi Co., Ltd. (2022 dataset value): number of provincial or ministerial science and technology progress awards = 2",
+    "Hai Li Chuang Yao Jia Dian Co., Ltd. (2022 dataset value): number of provincial or ministerial science and technology progress awards = 0"
+  ],
+  "milestone": {
+    "Number of provincial or ministerial science and technology progress awards of Mei Neng Xuan Jin Dian Qi Co., Ltd. in 2022": 2,
+    "Number of provincial or ministerial science and technology progress awards of Hai Li Chuang Yao Jia Dian Co., Ltd. in 2022": 0,
+    "Comparison result (whether Mei Neng Xuan Jin Dian Qi Co., Ltd. is lower than Hai Li Chuang Yao Jia Dian Co., Ltd.)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: number of provincial or ministerial science and technology progress awards of Mei Neng Xuan Jin Dian Qi Co., Ltd. in 2022 = 2",
+    "Extracted from company_operation_status.csv: number of provincial or ministerial science and technology progress awards of Hai Li Chuang Yao Jia Dian Co., Ltd. in 2022 = 0",
+    "Compared 2 and 0; since 2 < 0 is false, the judgment is \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "3808263c-c56e-4eeb-b16d-b32302554002"
+    },
+    {
+      "company_operation_status": "1742607a-1367-458d-8bf9-b9850d72a344"
+    },
+    {
+      "company_profile": "90e9de1d-a9ed-4432-b6fb-3e488d2ee3c6"
+    },
+    {
+      "company_profile": "cbcdc5e9-a16e-41ca-9f5c-3b29d0902389"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy018_result.json b/assets/qa_raw/enterprise_industry_analysis/easy018_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f217155fed421b3d7a49e0e45f4908cace76964
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy018_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy018",
+  "question": "In 2022, compared with Jing Neng Dian Re Ran Qi Co., Ltd., what is the difference in total liabilities of San Xia Ze Neng Dian Li Co., Ltd.?",
+  "guidelines": "The answer must be a number with two decimal places. Output only the number without units, commas, or any text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "San Xia Ze Neng Dian Li Co., Ltd. (2022 dataset value): total liabilities = 10344552545.65 CNY",
+    "Jing Neng Dian Re Ran Qi Co., Ltd. (2022 dataset value): total liabilities = 8274232707.73 CNY"
+  ],
+  "milestone": {
+    "Total liabilities of San Xia Ze Neng Dian Li Co., Ltd. in 2022 (CNY)": 10344552545.65,
+    "Total liabilities of Jing Neng Dian Re Ran Qi Co., Ltd. in 2022 (CNY)": 8274232707.73,
+    "Difference (San Xia Ze Neng Dian Li Co., Ltd. - Jing Neng Dian Re Ran Qi Co., Ltd.)": 2070319837.92
+  },
+  "answer": 2070319837.92,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: total liabilities of San Xia Ze Neng Dian Li Co., Ltd. in 2022 = 10344552545.65 CNY",
+    "Extracted from company_operation_status.csv: total liabilities of Jing Neng Dian Re Ran Qi Co., Ltd. in 2022 = 8274232707.73 CNY",
+    "Calculated the difference (San Xia Ze Neng Dian Li Co., Ltd. - Jing Neng Dian Re Ran Qi Co., Ltd.): 10344552545.65 - 8274232707.73 = 2070319837.92, and output with two decimal places as required"
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "77f252db-7ff7-4422-aac6-4db091b00f91"
+    },
+    {
+      "company_operation_status": "78ce6b29-7295-43bb-ab69-473b188f464d"
+    },
+    {
+      "company_profile": "6e04f2f0-42d4-46a2-885c-87d82009b319"
+    },
+    {
+      "company_profile": "af90245d-1d33-4fa9-800a-cabc15e69927"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy019_result.json b/assets/qa_raw/enterprise_industry_analysis/easy019_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb329e4d0a992933aa0100ddc404978ced99bb59
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy019_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy019",
+  "question": "In 2022, is the R&D investment ratio of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. lower than that of Yi Shan Tai Tai Medical Devices Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. (2022 dataset value): R&D investment ratio = 3.96 %",
+    "Yi Shan Tai Tai Medical Devices Co., Ltd. (2022 dataset value): R&D investment ratio = 11.1 %"
+  ],
+  "milestone": {
+    "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. R&D investment ratio in 2022": "3.96 %",
+    "Yi Shan Tai Tai Medical Devices Co., Ltd. R&D investment ratio in 2022": "11.1 %",
+    "Comparison result (whether Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. is lower than Yi Shan Tai Tai Medical Devices Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. R&D investment ratio in 2022 = 3.96 %",
+    "Extracted from company_operation_status.csv: Yi Shan Tai Tai Medical Devices Co., Ltd. R&D investment ratio in 2022 = 11.1 %",
+    "Compared 3.96 and 11.1; since 3.96 < 11.1, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "82b1665a-ac50-4f32-bb39-62e19fce2035"
+    },
+    {
+      "company_operation_status": "a50f3044-957a-4380-8852-f32116951aed"
+    },
+    {
+      "company_profile": "6e6b4e87-8925-4b4b-a3d5-8f4e80099348"
+    },
+    {
+      "company_profile": "14950802-93f3-4237-a0ed-9d1b591fab65"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy020_result.json b/assets/qa_raw/enterprise_industry_analysis/easy020_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..26973a1ce12371314ae00dd6677f3c9fc909fac8
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy020_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy020",
+  "question": "In 2022, is the debt-to-asset ratio of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. higher than that of Yi Shan Tai Tai Medical Devices Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. (2022 dataset value): debt-to-asset ratio = 63.23 %",
+    "Yi Shan Tai Tai Medical Devices Co., Ltd. (2022 dataset value): debt-to-asset ratio = 39.97 %"
+  ],
+  "milestone": {
+    "Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. debt-to-asset ratio in 2022": "63.23 %",
+    "Yi Shan Tai Tai Medical Devices Co., Ltd. debt-to-asset ratio in 2022": "39.97 %",
+    "Comparison result (whether Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. is higher than Yi Shan Tai Tai Medical Devices Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. debt-to-asset ratio in 2022 = 63.23 %",
+    "Extracted from company_operation_status.csv: Yi Shan Tai Tai Medical Devices Co., Ltd. debt-to-asset ratio in 2022 = 39.97 %",
+    "Compared 63.23 and 39.97; since 63.23 > 39.97, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "82b1665a-ac50-4f32-bb39-62e19fce2035"
+    },
+    {
+      "company_operation_status": "a50f3044-957a-4380-8852-f32116951aed"
+    },
+    {
+      "company_profile": "6e6b4e87-8925-4b4b-a3d5-8f4e80099348"
+    },
+    {
+      "company_profile": "14950802-93f3-4237-a0ed-9d1b591fab65"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy021_result.json b/assets/qa_raw/enterprise_industry_analysis/easy021_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e8975bed45a3b8e409ca837e28bd5cbbb3803d9
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy021_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy021",
+  "question": "In 2022, is the total citations of all patents of Guang Sheng Chang Ze Group Co., Ltd. higher than the corresponding indicator of Lang Ji Yun Hui Technology Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Guang Sheng Chang Ze Group Co., Ltd. (2022 dataset value): total citations of all patents = 199",
+    "Lang Ji Yun Hui Technology Co., Ltd. (2022 dataset value): total citations of all patents = 2107"
+  ],
+  "milestone": {
+    "Guang Sheng Chang Ze Group Co., Ltd. total citations of all patents in 2022": 199,
+    "Lang Ji Yun Hui Technology Co., Ltd. total citations of all patents in 2022": 2107,
+    "Comparison result (whether Guang Sheng Chang Ze Group Co., Ltd. is higher than Lang Ji Yun Hui Technology Co., Ltd.)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Guang Sheng Chang Ze Group Co., Ltd. total citations of all patents in 2022 = 199",
+    "Extracted from company_operation_status.csv: Lang Ji Yun Hui Technology Co., Ltd. total citations of all patents in 2022 = 2107",
+    "Compared 199 and 2107; since 199 < 2107, the judgment is \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "edcad9e4-7732-4626-be17-221470971633"
+    },
+    {
+      "company_operation_status": "65e4ecc7-a9f7-4b79-8e68-371a69e09fbc"
+    },
+    {
+      "company_profile": "604aa624-c515-41ee-b7b2-7b253cdfb094"
+    },
+    {
+      "company_profile": "9c26f228-f68a-4696-8ec6-a9def8807013"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy022_result.json b/assets/qa_raw/enterprise_industry_analysis/easy022_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..61f4de9cbecb85c4da46266ac990de290186eac6
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy022_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy022",
+  "question": "In 2022, is the annual number of Chinese patent applications of Pu Ge Jian Chen Pharmaceutical Co., Ltd. higher than that of Jin Hu Real Estate Construction Development Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Pu Ge Jian Chen Pharmaceutical Co., Ltd. (2022 dataset value): annual Chinese patent applications = 11",
+    "Jin Hu Real Estate Construction Development Co., Ltd. (2022 dataset value): annual Chinese patent applications = 6"
+  ],
+  "milestone": {
+    "Pu Ge Jian Chen Pharmaceutical Co., Ltd. annual Chinese patent applications in 2022": 11,
+    "Jin Hu Real Estate Construction Development Co., Ltd. annual Chinese patent applications in 2022": 6,
+    "Comparison result (whether Pu Ge Jian Chen Pharmaceutical Co., Ltd. is higher than Jin Hu Real Estate Construction Development Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Pu Ge Jian Chen Pharmaceutical Co., Ltd. annual Chinese patent applications in 2022 = 11",
+    "Extracted from company_operation_status.csv: Jin Hu Real Estate Construction Development Co., Ltd. annual Chinese patent applications in 2022 = 6",
+    "Compared 11 and 6; since 11 > 6, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "6b625166-a050-4409-961c-ac618ab953a1"
+    },
+    {
+      "company_operation_status": "3cb33706-a1b1-455f-8576-c81fef20c0dc"
+    },
+    {
+      "company_profile": "d2100a92-abee-4ae0-bc48-dd4086571095"
+    },
+    {
+      "company_profile": "93890202-7e17-4984-8ee9-30c130437075"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy023_result.json b/assets/qa_raw/enterprise_industry_analysis/easy023_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..1c57727e184e14ef1e6f127b1555371db3c673ad
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy023_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy023",
+  "question": "In 2022, which is larger: the R&D investment ratio of Hua Cheng Sheng Yuan Integrated Development Co., Ltd. or the year-over-year net profit change rate of Long He Zhi Jin Real Estate Co., Ltd.?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Hua Cheng Sheng Yuan Integrated Development Co., Ltd. (2022 dataset value): R&D investment ratio = 4.79 %",
+    "Long He Zhi Jin Real Estate Co., Ltd. (2022 dataset value): year-over-year net profit change rate = -32.50 %"
+  ],
+  "milestone": {
+    "Hua Cheng Sheng Yuan Integrated Development Co., Ltd. R&D investment ratio in 2022": "4.79 %",
+    "Long He Zhi Jin Real Estate Co., Ltd. year-over-year net profit change rate in 2022": "-32.50 %",
+    "Comparison result (which is larger)": "Hua Cheng Sheng Yuan Integrated Development Co., Ltd."
+  },
+  "answer": "Hua Cheng Sheng Yuan Integrated Development Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Hua Cheng Sheng Yuan Integrated Development Co., Ltd. R&D investment ratio in 2022 = 4.79 %",
+    "Extracted from company_operation_status.csv: Long He Zhi Jin Real Estate Co., Ltd. year-over-year net profit change rate in 2022 = -32.50 %",
+    "Compared 4.79 and -32.50; since 4.79 > -32.50, the larger one is \"Hua Cheng Sheng Yuan Integrated Development Co., Ltd.\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "42462fed-6dbc-400e-b9da-c46efc7fcf28"
+    },
+    {
+      "company_operation_status": "e42fabec-4192-4d35-8382-c3f2c2575de0"
+    },
+    {
+      "company_profile": "41fc40f8-35ba-4a45-9c72-56aa081d8caa"
+    },
+    {
+      "company_profile": "26b455a0-8084-4ea6-a2d9-b08d9305f26b"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy024_result.json b/assets/qa_raw/enterprise_industry_analysis/easy024_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a88aad627605b04cdabb132b4457f3f4846b938b
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy024_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy024",
+  "question": "In 2022, is the cumulative number of PCT patent applications of Shi Yang Zhi Guang Dian Qi Co., Ltd. lower than the corresponding value of Xu Ye Zhi Gong Technology Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Shi Yang Zhi Guang Dian Qi Co., Ltd. (2022 dataset value): cumulative number of PCT patent applications = 93",
+    "Xu Ye Zhi Gong Technology Co., Ltd. (2022 dataset value): cumulative number of PCT patent applications = 10"
+  ],
+  "milestone": {
+    "Cumulative number of PCT patent applications of Shi Yang Zhi Guang Dian Qi Co., Ltd. in 2022": 93,
+    "Cumulative number of PCT patent applications of Xu Ye Zhi Gong Technology Co., Ltd. in 2022": 10,
+    "Comparison result (whether Shi Yang Zhi Guang Dian Qi Co., Ltd. is lower than Xu Ye Zhi Gong Technology Co., Ltd.)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: cumulative number of PCT patent applications of Shi Yang Zhi Guang Dian Qi Co., Ltd. in 2022 = 93",
+    "Extracted from company_operation_status.csv: cumulative number of PCT patent applications of Xu Ye Zhi Gong Technology Co., Ltd. in 2022 = 10",
+    "Compared 93 and 10; since 93 < 10 is false, the judgment is \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "ce42b047-ea53-4a54-9cd5-10e1e29ac434"
+    },
+    {
+      "company_operation_status": "0a456d79-14fd-443e-9e71-c9df70663a82"
+    },
+    {
+      "company_profile": "345dea86-b238-41e5-8187-01d30579f033"
+    },
+    {
+      "company_profile": "0a786746-192d-4c1a-899a-be3590508d6b"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy025_result.json b/assets/qa_raw/enterprise_industry_analysis/easy025_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f5962386ea0237c4030dee308d8240d241c4e32
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy025_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy025",
+  "question": "In 2022, is the government award funding or subsidy of Shi Yang Zhi Guang Dian Qi Co., Ltd. lower than that of Xu Ye Zhi Gong Technology Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Shi Yang Zhi Guang Dian Qi Co., Ltd. (2022 dataset value): government award funding or subsidy = 14071208.03 CNY",
+    "Xu Ye Zhi Gong Technology Co., Ltd. (2022 dataset value): government award funding or subsidy = 100476261.0 CNY"
+  ],
+  "milestone": {
+    "Government award funding or subsidy of Shi Yang Zhi Guang Dian Qi Co., Ltd. in 2022 (CNY)": 14071208.03,
+    "Government award funding or subsidy of Xu Ye Zhi Gong Technology Co., Ltd. in 2022 (CNY)": 100476261.0,
+    "Comparison result (whether Shi Yang Zhi Guang Dian Qi Co., Ltd. is lower than Xu Ye Zhi Gong Technology Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: government award funding or subsidy of Shi Yang Zhi Guang Dian Qi Co., Ltd. in 2022 = 14071208.03 CNY",
+    "Extracted from company_operation_status.csv: government award funding or subsidy of Xu Ye Zhi Gong Technology Co., Ltd. in 2022 = 100476261.0 CNY",
+    "Compared 14071208.03 and 100476261.0; since 14071208.03 < 100476261.0, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "ce42b047-ea53-4a54-9cd5-10e1e29ac434"
+    },
+    {
+      "company_operation_status": "0a456d79-14fd-443e-9e71-c9df70663a82"
+    },
+    {
+      "company_profile": "345dea86-b238-41e5-8187-01d30579f033"
+    },
+    {
+      "company_profile": "0a786746-192d-4c1a-899a-be3590508d6b"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy026_result.json b/assets/qa_raw/enterprise_industry_analysis/easy026_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..94151f96951aa321843446f88bea33b4673fcacb
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy026_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy026",
+  "question": "In 2022, is the total number of employees of Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. higher than that of Shen Zhou Wu Jin Zi Xun Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. (2022 dataset value): total number of employees = 1066.0",
+    "Shen Zhou Wu Jin Zi Xun Co., Ltd. (2022 dataset value): total number of employees = 133.0"
+  ],
+  "milestone": {
+    "Total number of employees of Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. in 2022": 1066.0,
+    "Total number of employees of Shen Zhou Wu Jin Zi Xun Co., Ltd. in 2022": 133.0,
+    "Comparison result (whether Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. is higher than Shen Zhou Wu Jin Zi Xun Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: total number of employees of Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. in 2022 = 1066.0",
+    "Extracted from company_operation_status.csv: total number of employees of Shen Zhou Wu Jin Zi Xun Co., Ltd. in 2022 = 133.0",
+    "Compared 1066.0 and 133.0; since 1066.0 > 133.0, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "67aa7a78-b1ee-4a1b-9528-b28368c8deeb"
+    },
+    {
+      "company_operation_status": "81d91eb1-1f50-4c56-a4e4-a2c803a02c24"
+    },
+    {
+      "company_profile": "9d65552c-1b38-4861-bbbd-fa673cc1b93c"
+    },
+    {
+      "company_profile": "06a4b49b-2d95-424a-bc5a-a2a44d235ca2"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy027_result.json b/assets/qa_raw/enterprise_industry_analysis/easy027_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0e55f7887afd992fb3779c7698f24fff6345a72
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy027_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy027",
+  "question": "In 2022, is the cumulative number of invalid Chinese invention patents of Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. lower than the same indicator of Shen Zhou Wu Jin Zi Xun Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. (2022 dataset value): cumulative number of invalid Chinese invention patents = 1",
+    "Shen Zhou Wu Jin Zi Xun Co., Ltd. (2022 dataset value): cumulative number of invalid Chinese invention patents = 8"
+  ],
+  "milestone": {
+    "Cumulative number of invalid Chinese invention patents of Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. in 2022": 1,
+    "Cumulative number of invalid Chinese invention patents of Shen Zhou Wu Jin Zi Xun Co., Ltd. in 2022": 8,
+    "Comparison result (whether Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. is lower than Shen Zhou Wu Jin Zi Xun Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: cumulative number of invalid Chinese invention patents of Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. in 2022 = 1",
+    "Extracted from company_operation_status.csv: cumulative number of invalid Chinese invention patents of Shen Zhou Wu Jin Zi Xun Co., Ltd. in 2022 = 8",
+    "Compared 1 and 8; since 1 < 8, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "67aa7a78-b1ee-4a1b-9528-b28368c8deeb"
+    },
+    {
+      "company_operation_status": "81d91eb1-1f50-4c56-a4e4-a2c803a02c24"
+    },
+    {
+      "company_profile": "9d65552c-1b38-4861-bbbd-fa673cc1b93c"
+    },
+    {
+      "company_profile": "06a4b49b-2d95-424a-bc5a-a2a44d235ca2"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy028_result.json b/assets/qa_raw/enterprise_industry_analysis/easy028_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c8d72f9779742a40c587ec2fe8c86c763aa27e6f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy028_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy028",
+  "question": "In 2022, is the government reward fund or subsidy of Guang Sheng Chang Ze Group Co., Ltd. higher than that of Zhong You Zheng Yun Yun Port Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Guang Sheng Chang Ze Group Co., Ltd. (2022 dataset value): government reward fund or subsidy = 8527155.34 CNY",
+    "Zhong You Zheng Yun Yun Port Co., Ltd. (2022 dataset value): government reward fund or subsidy = 36491108.19 CNY"
+  ],
+  "milestone": {
+    "Guang Sheng Chang Ze Group Co., Ltd. government reward fund or subsidy in 2022 (CNY)": 8527155.34,
+    "Zhong You Zheng Yun Yun Port Co., Ltd. government reward fund or subsidy in 2022 (CNY)": 36491108.19,
+    "Comparison result (whether Guang Sheng Chang Ze Group Co., Ltd. is higher than Zhong You Zheng Yun Yun Port Co., Ltd.)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Guang Sheng Chang Ze Group Co., Ltd. government reward fund or subsidy in 2022 = 8527155.34 CNY",
+    "Extracted from company_operation_status.csv: Zhong You Zheng Yun Yun Port Co., Ltd. government reward fund or subsidy in 2022 = 36491108.19 CNY",
+    "Compared 8527155.34 and 36491108.19; since 8527155.34 < 36491108.19, the judgment is \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "991d4848-80a3-44e7-a032-e3cf54278e83"
+    },
+    {
+      "company_operation_status": "51152e33-2b18-4711-bfda-d4a1fc5b16e0"
+    },
+    {
+      "company_profile": "604aa624-c515-41ee-b7b2-7b253cdfb094"
+    },
+    {
+      "company_profile": "bc15ef65-603d-4fea-8d76-65d5708c8459"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy029_result.json b/assets/qa_raw/enterprise_industry_analysis/easy029_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2862dd924f38a392fc31caec4860f4ed325ca5c6
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy029_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy029",
+  "question": "In 2022, is the year-over-year change rate of R&D investment of Yong Feng Ke Lian Software Co., Ltd. higher than that of Hai Li Xuan Yue Electric Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Yong Feng Ke Lian Software Co., Ltd. (2022 dataset value): year-over-year change rate of R&D investment = -27.60 %",
+    "Hai Li Xuan Yue Electric Co., Ltd. (2022 dataset value): year-over-year change rate of R&D investment = -32.38 %"
+  ],
+  "milestone": {
+    "Yong Feng Ke Lian Software Co., Ltd. year-over-year change rate of R&D investment in 2022": "-27.60 %",
+    "Hai Li Xuan Yue Electric Co., Ltd. year-over-year change rate of R&D investment in 2022": "-32.38 %",
+    "Comparison result (whether Yong Feng Ke Lian Software Co., Ltd. is higher than Hai Li Xuan Yue Electric Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Yong Feng Ke Lian Software Co., Ltd. year-over-year change rate of R&D investment in 2022 = -27.60 %",
+    "Extracted from company_operation_status.csv: Hai Li Xuan Yue Electric Co., Ltd. year-over-year change rate of R&D investment in 2022 = -32.38 %",
+    "Compared -27.60 and -32.38; since -27.60 > -32.38, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "44e9ab9f-a44b-4aff-8e7d-daf1378262a1"
+    },
+    {
+      "company_operation_status": "49347141-510a-48e9-9eae-aeaa66708434"
+    },
+    {
+      "company_profile": "0b047bf5-e0ad-4a2d-86f2-aa973ef81724"
+    },
+    {
+      "company_profile": "2a4e9e8a-629f-4e35-b0b6-73e1f611409a"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy030_result.json b/assets/qa_raw/enterprise_industry_analysis/easy030_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..83605db059a5ca68ea05f6ae3034734a42e0390f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy030_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy030",
+  "question": "In 2022, is the total assets of Yong Feng Ke Lian Software Co., Ltd. higher than that of Hai Li Xuan Yue Electric Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Yong Feng Ke Lian Software Co., Ltd. (2022 dataset value): total assets = 17094624520.40 CNY",
+    "Hai Li Xuan Yue Electric Co., Ltd. (2022 dataset value): total assets = 2678955419.08 CNY"
+  ],
+  "milestone": {
+    "Yong Feng Ke Lian Software Co., Ltd. total assets in 2022 (CNY)": 17094624520.4,
+    "Hai Li Xuan Yue Electric Co., Ltd. total assets in 2022 (CNY)": 2678955419.08,
+    "Comparison result (whether Yong Feng Ke Lian Software Co., Ltd. is higher than Hai Li Xuan Yue Electric Co., Ltd.)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Yong Feng Ke Lian Software Co., Ltd. total assets in 2022 = 17094624520.40 CNY",
+    "Extracted from company_operation_status.csv: Hai Li Xuan Yue Electric Co., Ltd. total assets in 2022 = 2678955419.08 CNY",
+    "Compared 17094624520.40 and 2678955419.08; since 17094624520.40 > 2678955419.08, the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "44e9ab9f-a44b-4aff-8e7d-daf1378262a1"
+    },
+    {
+      "company_operation_status": "49347141-510a-48e9-9eae-aeaa66708434"
+    },
+    {
+      "company_profile": "0b047bf5-e0ad-4a2d-86f2-aa973ef81724"
+    },
+    {
+      "company_profile": "2a4e9e8a-629f-4e35-b0b6-73e1f611409a"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy031_result.json b/assets/qa_raw/enterprise_industry_analysis/easy031_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e81af41d64b96da47f29fa9a7ad5e0e4aab8b895
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy031_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy031",
+  "question": "In 2022, which is higher: the operating revenue amount of Heng Yi Run Heng Technology Co., Ltd. or the total assets of Lian Ji Ji Jin Ji Chuang Co., Ltd.?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Heng Yi Run Heng Technology Co., Ltd. (2022 dataset value): operating revenue amount = 4259181531.38 CNY",
+    "Lian Ji Ji Jin Ji Chuang Co., Ltd. (2022 dataset value): total assets = 6102473889.29 CNY"
+  ],
+  "milestone": {
+    "Heng Yi Run Heng Technology Co., Ltd. operating revenue amount in 2022 (CNY)": 4259181531.38,
+    "Lian Ji Ji Jin Ji Chuang Co., Ltd. total assets in 2022 (CNY)": 6102473889.29,
+    "Comparison result (which is higher)": "Lian Ji Ji Jin Ji Chuang Co., Ltd."
+  },
+  "answer": "Lian Ji Ji Jin Ji Chuang Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Heng Yi Run Heng Technology Co., Ltd. operating revenue amount in 2022 = 4259181531.38 CNY",
+    "Extracted from company_operation_status.csv: Lian Ji Ji Jin Ji Chuang Co., Ltd. total assets in 2022 = 6102473889.29 CNY",
+    "Compared 4259181531.38 and 6102473889.29; since 6102473889.29 > 4259181531.38, the higher one is \"Lian Ji Ji Jin Ji Chuang Co., Ltd.\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "445b18a4-fbae-4a49-b1f3-0c83f3557f04"
+    },
+    {
+      "company_operation_status": "c3537b48-5679-48c0-b912-9c181360ffdf"
+    },
+    {
+      "company_profile": "66aa1d69-d99c-4c11-8393-cd832d4ac729"
+    },
+    {
+      "company_profile": "8ea3eaf8-c762-4525-a573-ed3e5ac71842"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy032_result.json b/assets/qa_raw/enterprise_industry_analysis/easy032_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d286d75f924b5e98b7b332fe86f49c1fdafd6413
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy032_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy032",
+  "question": "In 2022, is the number of R&D personnel of Heng Yi Run Heng Technology Co., Ltd. lower than the annual number of Chinese patent applications of Lian Ji Ji Jin Ji Chuang Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Heng Yi Run Heng Technology Co., Ltd. (2022 dataset value): number of R&D personnel = 257.0",
+    "Lian Ji Ji Jin Ji Chuang Co., Ltd. (2022 dataset value): annual Chinese patent applications = 75"
+  ],
+  "milestone": {
+    "Heng Yi Run Heng Technology Co., Ltd. number of R&D personnel in 2022": 257.0,
+    "Lian Ji Ji Jin Ji Chuang Co., Ltd. annual Chinese patent applications in 2022": 75,
+    "Comparison result (whether Heng Yi Run Heng Technology Co., Ltd. is lower than Lian Ji Ji Jin Ji Chuang Co., Ltd.)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Heng Yi Run Heng Technology Co., Ltd. number of R&D personnel in 2022 = 257.0",
+    "Extracted from company_operation_status.csv: Lian Ji Ji Jin Ji Chuang Co., Ltd. annual Chinese patent applications in 2022 = 75",
+    "Compared 257.0 and 75; since 257.0 < 75 is false, the judgment is \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "445b18a4-fbae-4a49-b1f3-0c83f3557f04"
+    },
+    {
+      "company_operation_status": "c3537b48-5679-48c0-b912-9c181360ffdf"
+    },
+    {
+      "company_profile": "66aa1d69-d99c-4c11-8393-cd832d4ac729"
+    },
+    {
+      "company_profile": "8ea3eaf8-c762-4525-a573-ed3e5ac71842"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy033_result.json b/assets/qa_raw/enterprise_industry_analysis/easy033_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f6c020cd2ddf89dec88135f56abf4e897fcab1da
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy033_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy033",
+  "question": "In 2022, compared with the total liabilities of Wei Xing Run Jin Ke Ji Co., Ltd., what is the difference in total assets of Ping Ru Gang Tong Yun Wu Liu Co., Ltd.?",
+  "guidelines": "The answer must be a number with two decimal places. Output only the number without units, commas, or any text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Ping Ru Gang Tong Yun Wu Liu Co., Ltd. (2022 dataset value): total assets = 78458358656.04 CNY",
+    "Wei Xing Run Jin Ke Ji Co., Ltd. (2022 dataset value): total liabilities = 40508066253.17 CNY"
+  ],
+  "milestone": {
+    "Total assets of Ping Ru Gang Tong Yun Wu Liu Co., Ltd. in 2022 (CNY)": 78458358656.04,
+    "Total liabilities of Wei Xing Run Jin Ke Ji Co., Ltd. in 2022 (CNY)": 40508066253.17,
+    "Difference (Ping Ru Gang Tong Yun Wu Liu Co., Ltd. - Wei Xing Run Jin Ke Ji Co., Ltd.)": 37950292402.87
+  },
+  "answer": 37950292402.87,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: total assets of Ping Ru Gang Tong Yun Wu Liu Co., Ltd. in 2022 = 78458358656.04 CNY",
+    "Extracted from company_operation_status.csv: total liabilities of Wei Xing Run Jin Ke Ji Co., Ltd. in 2022 = 40508066253.17 CNY",
+    "Calculated the difference: 78458358656.04 - 40508066253.17 = 37950292402.87"
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "fa2375e5-202c-4f7f-bae3-5d96cbeeaaf8"
+    },
+    {
+      "company_operation_status": "0ec89f27-4b21-43c2-ae26-69d789b9e8ac"
+    },
+    {
+      "company_profile": "e0583755-6bf9-4c80-a2cd-5d51dc51e05c"
+    },
+    {
+      "company_profile": "6d6e1959-c844-4469-acce-840c7c7f2599"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy034_result.json b/assets/qa_raw/enterprise_industry_analysis/easy034_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e029a3e76e30af711c2d3fd03ff4bddd7a01a38c
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy034_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy034",
+  "question": "In 2022, between the year-over-year net profit growth rate of Ping Ru Gang Tong Yun Wu Liu Co., Ltd. and the net profit amount of Wei Xing Run Jin Ke Ji Co., Ltd., which value is larger?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Ping Ru Gang Tong Yun Wu Liu Co., Ltd. (2022 dataset value): year-over-year net profit growth rate = -9.38 %",
+    "Wei Xing Run Jin Ke Ji Co., Ltd. (2022 dataset value): net profit amount = 388147978.44 CNY"
+  ],
+  "milestone": {
+    "Year-over-year net profit growth rate of Ping Ru Gang Tong Yun Wu Liu Co., Ltd. in 2022": "-9.38 %",
+    "Net profit amount of Wei Xing Run Jin Ke Ji Co., Ltd. in 2022 (CNY)": 388147978.44,
+    "Comparison result (which value is larger)": "Wei Xing Run Jin Ke Ji Co., Ltd."
+  },
+  "answer": "Wei Xing Run Jin Ke Ji Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: year-over-year net profit growth rate of Ping Ru Gang Tong Yun Wu Liu Co., Ltd. in 2022 = -9.38 %",
+    "Extracted from company_operation_status.csv: net profit amount of Wei Xing Run Jin Ke Ji Co., Ltd. in 2022 = 388147978.44 CNY",
+    "Compared -9.38 and 388147978.44 as numeric values; since 388147978.44 > -9.38, the larger one is \"Wei Xing Run Jin Ke Ji Co., Ltd.\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_operation_status": "fa2375e5-202c-4f7f-bae3-5d96cbeeaaf8"
+    },
+    {
+      "company_operation_status": "0ec89f27-4b21-43c2-ae26-69d789b9e8ac"
+    },
+    {
+      "company_profile": "e0583755-6bf9-4c80-a2cd-5d51dc51e05c"
+    },
+    {
+      "company_profile": "6d6e1959-c844-4469-acce-840c7c7f2599"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy035_result.json b/assets/qa_raw/enterprise_industry_analysis/easy035_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d84717d8158aaa113695c9958d1719a5f8689898
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy035_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy035",
+  "question": "In 2022, are the industries of Gao Yin Ze Tong Pi Fa Co., Ltd. and Yong Hui Ze Hui Pi Fa Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Gao Yin Ze Tong Pi Fa Co., Ltd. (2022 dataset value): industry = Wholesale and Retail",
+    "Yong Hui Ze Hui Pi Fa Co., Ltd. (2022 dataset value): industry = Wholesale and Retail"
+  ],
+  "milestone": {
+    "Industry of Gao Yin Ze Tong Pi Fa Co., Ltd.": "Wholesale and Retail",
+    "Industry of Yong Hui Ze Hui Pi Fa Co., Ltd.": "Wholesale and Retail",
+    "Comparison result (whether the industries are the same)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Gao Yin Ze Tong Pi Fa Co., Ltd. = Wholesale and Retail",
+    "Extracted from company_profile.csv: industry of Yong Hui Ze Hui Pi Fa Co., Ltd. = Wholesale and Retail",
+    "Compared the industry text of both companies; they are the same, so the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "a9b797ac-f74d-4a35-acfa-01b494fe2b3f"
+    },
+    {
+      "company_profile": "c02c6c02-5aa8-45ac-a752-7feb7538bf2c"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy036_result.json b/assets/qa_raw/enterprise_industry_analysis/easy036_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..97dde7ee058a5ea65cd338a8e027ec1d61b35168
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy036_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy036",
+  "question": "Compared with the listing date of Yong Hui Ze Hui Pi Fa Co., Ltd., which listing date is earlier for Gao Yin Ze Tong Pi Fa Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Gao Yin Ze Tong Pi Fa Co., Ltd. (2022 dataset value): listing date = 1992-08-17",
+    "Yong Hui Ze Hui Pi Fa Co., Ltd. (2022 dataset value): listing date = 1993-02-02"
+  ],
+  "milestone": {
+    "Listing date of Gao Yin Ze Tong Pi Fa Co., Ltd.": "1992-08-17",
+    "Listing date of Yong Hui Ze Hui Pi Fa Co., Ltd.": "1993-02-02",
+    "Comparison result (which one is earlier)": "Gao Yin Ze Tong Pi Fa Co., Ltd."
+  },
+  "answer": "Gao Yin Ze Tong Pi Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: listing date of Gao Yin Ze Tong Pi Fa Co., Ltd. = 1992-08-17",
+    "Extracted from company_profile.csv: listing date of Yong Hui Ze Hui Pi Fa Co., Ltd. = 1993-02-02",
+    "Compared the dates; 1992-08-17 is earlier than 1993-02-02, so the earlier one is \"Gao Yin Ze Tong Pi Fa Co., Ltd.\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "a9b797ac-f74d-4a35-acfa-01b494fe2b3f"
+    },
+    {
+      "company_profile": "c02c6c02-5aa8-45ac-a752-7feb7538bf2c"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy037_result.json b/assets/qa_raw/enterprise_industry_analysis/easy037_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4b11836cf6b7195d8c2df058650dd7d65e3e5d2b
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy037_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy037",
+  "question": "In 2022, are the industries of Ma Gang Tai Jin Cai Liao Co., Ltd. and Ma Gang Gang Sheng Bu Xiu Gang Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Ma Gang Tai Jin Cai Liao Co., Ltd. (2022 dataset value): industry = Metal Products",
+    "Ma Gang Gang Sheng Bu Xiu Gang Co., Ltd. (2022 dataset value): industry = Metal Products"
+  ],
+  "milestone": {
+    "Industry of Ma Gang Tai Jin Cai Liao Co., Ltd.": "Metal Products",
+    "Industry of Ma Gang Gang Sheng Bu Xiu Gang Co., Ltd.": "Metal Products",
+    "Comparison result (whether the industries are the same)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Ma Gang Tai Jin Cai Liao Co., Ltd. = Metal Products",
+    "Extracted from company_profile.csv: industry of Ma Gang Gang Sheng Bu Xiu Gang Co., Ltd. = Metal Products",
+    "Compared the industry text of both companies; they are the same, so the judgment is \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "7d643897-2fa1-4382-b948-8b984738888c"
+    },
+    {
+      "company_profile": "9da5bd9f-2707-4ea0-9bf8-ae269b9f71e7"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy038_result.json b/assets/qa_raw/enterprise_industry_analysis/easy038_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b252125aa8861a709b51c8fc4aef4672211f0e87
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy038_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy038",
+  "question": "In 2022, is the enterprise type of Magang Taijin Materials Co., Ltd. the same as that of Magang Gangsheng Stainless Steel Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Magang Taijin Materials Co., Ltd. (2022 dataset value): enterprise type = Shanghai and Shenzhen",
+    "Magang Gangsheng Stainless Steel Co., Ltd. (2022 dataset value): enterprise type = Shanghai and Shenzhen"
+  ],
+  "milestone": {
+    "Enterprise type of Magang Taijin Materials Co., Ltd.": "Shanghai and Shenzhen",
+    "Enterprise type of Magang Gangsheng Stainless Steel Co., Ltd.": "Shanghai and Shenzhen",
+    "Comparison result (whether the enterprise types are the same)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: enterprise type of Magang Taijin Materials Co., Ltd. = Shanghai and Shenzhen",
+    "Extracted from company_profile.csv: enterprise type of Magang Gangsheng Stainless Steel Co., Ltd. = Shanghai and Shenzhen",
+    "Compared the two enterprise type texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "7d643897-2fa1-4382-b948-8b984738888c"
+    },
+    {
+      "company_profile": "9da5bd9f-2707-4ea0-9bf8-ae269b9f71e7"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy039_result.json b/assets/qa_raw/enterprise_industry_analysis/easy039_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a6a1d2c8aca078e80aa551c1387ccda38bb23fd
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy039_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy039",
+  "question": "In 2022, is the industry of Chuangwei Yaosheng Electric Co., Ltd. the same as that of Lixin Zhichuang Home Appliances Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Chuangwei Yaosheng Electric Co., Ltd. (2022 dataset value): industry = Consumer Electronics and Electrical Industry",
+    "Lixin Zhichuang Home Appliances Co., Ltd. (2022 dataset value): industry = Consumer Electronics and Electrical Industry"
+  ],
+  "milestone": {
+    "Industry of Chuangwei Yaosheng Electric Co., Ltd.": "Consumer Electronics and Electrical Industry",
+    "Industry of Lixin Zhichuang Home Appliances Co., Ltd.": "Consumer Electronics and Electrical Industry",
+    "Comparison result (whether the industries are the same)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Chuangwei Yaosheng Electric Co., Ltd. = Consumer Electronics and Electrical Industry",
+    "Extracted from company_profile.csv: industry of Lixin Zhichuang Home Appliances Co., Ltd. = Consumer Electronics and Electrical Industry",
+    "Compared the two industry texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "1f9ed871-a876-4084-b817-62445a01ae19"
+    },
+    {
+      "company_profile": "f216eb5c-96da-4edd-8659-6cf8e17b783d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy040_result.json b/assets/qa_raw/enterprise_industry_analysis/easy040_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f090f54b3f9c4270955b0328e99b961114b07f2
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy040_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy040",
+  "question": "In 2022, is the stock exchange of Chuangwei Yaosheng Electric Co., Ltd. the same as that of Lixin Zhichuang Home Appliances Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Chuangwei Yaosheng Electric Co., Ltd. (2022 dataset value): stock exchange = Shenzhen Stock Exchange",
+    "Lixin Zhichuang Home Appliances Co., Ltd. (2022 dataset value): stock exchange = Shenzhen Stock Exchange"
+  ],
+  "milestone": {
+    "Stock exchange of Chuangwei Yaosheng Electric Co., Ltd.": "Shenzhen Stock Exchange",
+    "Stock exchange of Lixin Zhichuang Home Appliances Co., Ltd.": "Shenzhen Stock Exchange",
+    "Comparison result (whether the stock exchanges are the same)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: stock exchange of Chuangwei Yaosheng Electric Co., Ltd. = Shenzhen Stock Exchange",
+    "Extracted from company_profile.csv: stock exchange of Lixin Zhichuang Home Appliances Co., Ltd. = Shenzhen Stock Exchange",
+    "Compared the two stock exchange texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "1f9ed871-a876-4084-b817-62445a01ae19"
+    },
+    {
+      "company_profile": "f216eb5c-96da-4edd-8659-6cf8e17b783d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy041_result.json b/assets/qa_raw/enterprise_industry_analysis/easy041_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b0c6166b084e85e38b4135fe95e36ec7ad36127
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy041_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy041",
+  "question": "Comparing the incorporation dates of Biyuan Chanhua Real Estate Holdings Co., Ltd. and Huarun Zhijin Construction Development Co., Ltd., which one was established earlier?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Biyuan Chanhua Real Estate Holdings Co., Ltd. (2022 dataset value): incorporation date = 1989-11-22",
+    "Huarun Zhijin Construction Development Co., Ltd. (2022 dataset value): incorporation date = 1991-03-30"
+  ],
+  "milestone": {
+    "Incorporation date of Biyuan Chanhua Real Estate Holdings Co., Ltd.": "1989-11-22",
+    "Incorporation date of Huarun Zhijin Construction Development Co., Ltd.": "1991-03-30",
+    "Comparison result (which one is earlier)": "Biyuan Chanhua Real Estate Holdings Co., Ltd."
+  },
+  "answer": "Biyuan Chanhua Real Estate Holdings Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: incorporation date of Biyuan Chanhua Real Estate Holdings Co., Ltd. = 1989-11-22",
+    "Extracted from company_profile.csv: incorporation date of Huarun Zhijin Construction Development Co., Ltd. = 1991-03-30",
+    "Compared dates 1989-11-22 and 1991-03-30; since 1989-11-22 is earlier, the judgment is \"Biyuan Chanhua Real Estate Holdings Co., Ltd.\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "5f766e43-74ee-4181-8792-be1de51a721e"
+    },
+    {
+      "company_profile": "f2e93357-9c4a-4b4c-9090-3679117f651a"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy042_result.json b/assets/qa_raw/enterprise_industry_analysis/easy042_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..34bb3f2e3db041fa42219ffc9b8162b425133e20
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy042_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy042",
+  "question": "In 2022, are the industries of Biyuan Chanhua Real Estate Holdings Co., Ltd. and Huarun Zhijin Construction Development Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Biyuan Chanhua Real Estate Holdings Co., Ltd. (2022 dataset value): industry = Real Estate",
+    "Huarun Zhijin Construction Development Co., Ltd. (2022 dataset value): industry = Real Estate"
+  ],
+  "milestone": {
+    "Industry of Biyuan Chanhua Real Estate Holdings Co., Ltd.": "Real Estate",
+    "Industry of Huarun Zhijin Construction Development Co., Ltd.": "Real Estate",
+    "Comparison result (whether the industries are the same)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Biyuan Chanhua Real Estate Holdings Co., Ltd. = Real Estate",
+    "Extracted from company_profile.csv: industry of Huarun Zhijin Construction Development Co., Ltd. = Real Estate",
+    "Compared the two industry texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "5f766e43-74ee-4181-8792-be1de51a721e"
+    },
+    {
+      "company_profile": "f2e93357-9c4a-4b4c-9090-3679117f651a"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy043_result.json b/assets/qa_raw/enterprise_industry_analysis/easy043_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c142841b16173a772d98876d1189a169b760b07
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy043_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy043",
+  "question": "Comparing the incorporation dates of Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd. and Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd., which one was established earlier?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd. (2022 dataset value): incorporation date = 1994-04-20",
+    "Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd. (2022 dataset value): incorporation date = 2004-10-09"
+  ],
+  "milestone": {
+    "Incorporation date of Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd.": "1994-04-20",
+    "Incorporation date of Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd.": "2004-10-09",
+    "Comparison result (which one is earlier)": "Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd."
+  },
+  "answer": "Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: incorporation date of Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd. = 1994-04-20",
+    "Extracted from company_profile.csv: incorporation date of Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd. = 2004-10-09",
+    "Compared dates 1994-04-20 and 2004-10-09; since 1994-04-20 is earlier, the judgment is \"Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd.\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "b3d11f26-f7b7-4a0f-bfb7-049ceb95f179"
+    },
+    {
+      "company_profile": "d57c27e6-c71c-478d-8a29-b860c4006dad"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy044_result.json b/assets/qa_raw/enterprise_industry_analysis/easy044_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..22ddcfdd6c1349097c6450acd550505473d92a44
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy044_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy044",
+  "question": "In 2022, are the stock exchanges of Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd. and Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd. (2022 dataset value): stock exchange = Shenzhen Stock Exchange",
+    "Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd. (2022 dataset value): stock exchange = Shenzhen Stock Exchange"
+  ],
+  "milestone": {
+    "Stock exchange of Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd.": "Shenzhen Stock Exchange",
+    "Stock exchange of Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd.": "Shenzhen Stock Exchange",
+    "Comparison result (whether the stock exchanges are the same)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: stock exchange of Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd. = Shenzhen Stock Exchange",
+    "Extracted from company_profile.csv: stock exchange of Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd. = Shenzhen Stock Exchange",
+    "Compared the two stock exchange texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "b3d11f26-f7b7-4a0f-bfb7-049ceb95f179"
+    },
+    {
+      "company_profile": "d57c27e6-c71c-478d-8a29-b860c4006dad"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy045_result.json b/assets/qa_raw/enterprise_industry_analysis/easy045_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..81abd1390915a9e9b7cd5d12e3ec522412504d10
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy045_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy045",
+  "question": "In 2022, is the industry of Chuang Wei Jin Zhi Electrical Appliances Co., Ltd. the same as that of Long He Chan Zhi Di Chan Holdings Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Chuang Wei Jin Zhi Electrical Appliances Co., Ltd. (2022 dataset value): industry = Consumer Electronics and Electrical Industry",
+    "Long He Chan Zhi Di Chan Holdings Co., Ltd. (2022 dataset value): industry = Real Estate"
+  ],
+  "milestone": {
+    "Industry of Chuang Wei Jin Zhi Electrical Appliances Co., Ltd.": "Consumer Electronics and Electrical Industry",
+    "Industry of Long He Chan Zhi Di Chan Holdings Co., Ltd.": "Real Estate",
+    "Comparison result (whether the industries are the same)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Chuang Wei Jin Zhi Electrical Appliances Co., Ltd. = Consumer Electronics and Electrical Industry",
+    "Extracted from company_profile.csv: industry of Long He Chan Zhi Di Chan Holdings Co., Ltd. = Real Estate",
+    "Compared the two industry texts and judged \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "50554e97-eb8a-47a7-b8c3-bd642ed71cfd"
+    },
+    {
+      "company_profile": "75cba6a7-770b-4270-a8c5-e378847b0017"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy046_result.json b/assets/qa_raw/enterprise_industry_analysis/easy046_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..587659ffe889e3ef46064766610495ffc334da95
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy046_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy046",
+  "question": "In 2022, are the listing boards of Chuang Wei Jin Zhi Electrical Appliances Co., Ltd. and Long He Chan Zhi Di Chan Holdings Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Chuang Wei Jin Zhi Electrical Appliances Co., Ltd. (2022 dataset value): listing board = Main Board",
+    "Long He Chan Zhi Di Chan Holdings Co., Ltd. (2022 dataset value): listing board = Main Board"
+  ],
+  "milestone": {
+    "Listing board of Chuang Wei Jin Zhi Electrical Appliances Co., Ltd.": "Main Board",
+    "Listing board of Long He Chan Zhi Di Chan Holdings Co., Ltd.": "Main Board",
+    "Comparison result (whether the listing boards are the same)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: listing board of Chuang Wei Jin Zhi Electrical Appliances Co., Ltd. = Main Board",
+    "Extracted from company_profile.csv: listing board of Long He Chan Zhi Di Chan Holdings Co., Ltd. = Main Board",
+    "Compared the two listing board texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "50554e97-eb8a-47a7-b8c3-bd642ed71cfd"
+    },
+    {
+      "company_profile": "75cba6a7-770b-4270-a8c5-e378847b0017"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy047_result.json b/assets/qa_raw/enterprise_industry_analysis/easy047_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..618e99d28cd48154f5b3901634a3d020452b144c
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy047_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy047",
+  "question": "In 2022, are the industries of Hua Ying Tai Sheng Wealth Management Co., Ltd. and Yong Feng Lian Chuang Xi Tong Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Hua Ying Tai Sheng Wealth Management Co., Ltd. (2022 dataset value): industry = Financial Industry",
+    "Yong Feng Lian Chuang Xi Tong Co., Ltd. (2022 dataset value): industry = Information Transmission, Software and IT Services"
+  ],
+  "milestone": {
+    "Industry of Hua Ying Tai Sheng Wealth Management Co., Ltd.": "Financial Industry",
+    "Industry of Yong Feng Lian Chuang Xi Tong Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Comparison result (whether the industries are the same)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Hua Ying Tai Sheng Wealth Management Co., Ltd. = Financial Industry",
+    "Extracted from company_profile.csv: industry of Yong Feng Lian Chuang Xi Tong Co., Ltd. = Information Transmission, Software and IT Services",
+    "Compared the two industry texts and judged \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "b3ffe2d5-c5b8-4499-b7e7-c6955e84e897"
+    },
+    {
+      "company_profile": "8214fdfd-e320-4e2e-bdcf-a622940758af"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy048_result.json b/assets/qa_raw/enterprise_industry_analysis/easy048_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e55a6dcd82f07ec14e3db16df431c67b25a3805d
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy048_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy048",
+  "question": "In 2022, is the stock exchange of Huaying Taisheng Wealth Management Co., Ltd. the same as that of Yongfeng Lianchuang Systems Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Huaying Taisheng Wealth Management Co., Ltd. (2022 dataset value): stock exchange = Shenzhen Stock Exchange",
+    "Yongfeng Lianchuang Systems Co., Ltd. (2022 dataset value): stock exchange = Shenzhen Stock Exchange"
+  ],
+  "milestone": {
+    "Stock exchange of Huaying Taisheng Wealth Management Co., Ltd.": "Shenzhen Stock Exchange",
+    "Stock exchange of Yongfeng Lianchuang Systems Co., Ltd.": "Shenzhen Stock Exchange",
+    "Comparison result (whether the stock exchanges are the same)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: stock exchange of Huaying Taisheng Wealth Management Co., Ltd. = Shenzhen Stock Exchange",
+    "Extracted from company_profile.csv: stock exchange of Yongfeng Lianchuang Systems Co., Ltd. = Shenzhen Stock Exchange",
+    "Compared the two stock exchange texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "b3ffe2d5-c5b8-4499-b7e7-c6955e84e897"
+    },
+    {
+      "company_profile": "8214fdfd-e320-4e2e-bdcf-a622940758af"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy049_result.json b/assets/qa_raw/enterprise_industry_analysis/easy049_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0a3695ccbcaf94bcc04847c71a4e1ce26034083
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy049_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy049",
+  "question": "In 2022, is the board segment of Zhongbai Damao Wholesale Co., Ltd. the same as that of Luxi Runheng Chemical Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhongbai Damao Wholesale Co., Ltd. (2022 dataset value): board segment = Main Board",
+    "Luxi Runheng Chemical Co., Ltd. (2022 dataset value): board segment = Main Board"
+  ],
+  "milestone": {
+    "Board segment of Zhongbai Damao Wholesale Co., Ltd.": "Main Board",
+    "Board segment of Luxi Runheng Chemical Co., Ltd.": "Main Board",
+    "Comparison result (whether the board segments are the same)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: board segment of Zhongbai Damao Wholesale Co., Ltd. = Main Board",
+    "Extracted from company_profile.csv: board segment of Luxi Runheng Chemical Co., Ltd. = Main Board",
+    "Compared the two board segment texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "b8d58e29-5641-4ec0-a2e7-e0b658003f1a"
+    },
+    {
+      "company_profile": "707eb2a3-a701-4ec3-9cf9-3df0efc1246b"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy050_result.json b/assets/qa_raw/enterprise_industry_analysis/easy050_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4bdd751db0d1e856d868bbe2b716512664d1957a
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy050_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy050",
+  "question": "In 2022, is the ownership type of Zhongbai Damao Wholesale Co., Ltd. the same as that of Luxi Runheng Chemical Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhongbai Damao Wholesale Co., Ltd. (2022 dataset value): ownership type = Private Enterprise",
+    "Luxi Runheng Chemical Co., Ltd. (2022 dataset value): ownership type = Private Enterprise"
+  ],
+  "milestone": {
+    "Ownership type of Zhongbai Damao Wholesale Co., Ltd.": "Private Enterprise",
+    "Ownership type of Luxi Runheng Chemical Co., Ltd.": "Private Enterprise",
+    "Comparison result (whether the ownership types are the same)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: ownership type of Zhongbai Damao Wholesale Co., Ltd. = Private Enterprise",
+    "Extracted from company_profile.csv: ownership type of Luxi Runheng Chemical Co., Ltd. = Private Enterprise",
+    "Compared the two ownership type texts and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "b8d58e29-5641-4ec0-a2e7-e0b658003f1a"
+    },
+    {
+      "company_profile": "707eb2a3-a701-4ec3-9cf9-3df0efc1246b"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy051_result.json b/assets/qa_raw/enterprise_industry_analysis/easy051_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..170c04f11f7deb78b2a2aa0a80f8438a9a049efe
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy051_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy051",
+  "question": "In 2022, is the enterprise type of Meineng Xuanyue Electric Co., Ltd. the same as that of Baotie Yuanchang Metal Products Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Meineng Xuanyue Electric Co., Ltd. (2022 dataset value): enterprise type = Shanghai and Shenzhen",
+    "Baotie Yuanchang Metal Products Co., Ltd. (2022 dataset value): enterprise type = Hong Kong Stocks"
+  ],
+  "milestone": {
+    "Enterprise type of Meineng Xuanyue Electric Co., Ltd.": "Shanghai and Shenzhen",
+    "Enterprise type of Baotie Yuanchang Metal Products Co., Ltd.": "Hong Kong Stocks",
+    "Comparison result (whether the enterprise types are the same)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: enterprise type of Meineng Xuanyue Electric Co., Ltd. = Shanghai and Shenzhen",
+    "Extracted from company_profile.csv: enterprise type of Baotie Yuanchang Metal Products Co., Ltd. = Hong Kong Stocks",
+    "Compared the two enterprise type texts and judged \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "5ae7c38c-e9f3-47d0-972d-d7d00e7cd04b"
+    },
+    {
+      "company_profile": "8b773f98-e603-4b07-b235-fc1e71e2ea0d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy052_result.json b/assets/qa_raw/enterprise_industry_analysis/easy052_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0169546bafc5ac93a9d417f9d97953d9ece6e7e
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy052_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy052",
+  "question": "In 2022, are the stock exchanges of Mei Neng Xuan Yue Dian Qi Co., Ltd. and Bao Tie Yuan Chang Jin Shu Zhi Pin Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Mei Neng Xuan Yue Dian Qi Co., Ltd. (2022 dataset value): stock exchange = Shenzhen Stock Exchange",
+    "Bao Tie Yuan Chang Jin Shu Zhi Pin Co., Ltd. (2022 dataset value): stock exchange = Gang Jiao Suo"
+  ],
+  "milestone": {
+    "Stock exchange of Mei Neng Xuan Yue Dian Qi Co., Ltd.": "Shenzhen Stock Exchange",
+    "Stock exchange of Bao Tie Yuan Chang Jin Shu Zhi Pin Co., Ltd.": "Gang Jiao Suo",
+    "Comparison result (whether the stock exchanges are the same)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: stock exchange of Mei Neng Xuan Yue Dian Qi Co., Ltd. = Shenzhen Stock Exchange",
+    "Extracted from company_profile.csv: stock exchange of Bao Tie Yuan Chang Jin Shu Zhi Pin Co., Ltd. = Gang Jiao Suo",
+    "Compared the two stock exchange texts and judged \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "5ae7c38c-e9f3-47d0-972d-d7d00e7cd04b"
+    },
+    {
+      "company_profile": "8b773f98-e603-4b07-b235-fc1e71e2ea0d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy053_result.json b/assets/qa_raw/enterprise_industry_analysis/easy053_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..28fd513362c4a4dc30fcb1837efdeb570939b171
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy053_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy053",
+  "question": "In 2022, does the secondary industry of Bao He Hua Chang Jian She Kai Fa Co., Ltd. belong to the same industry as that of Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Bao He Hua Chang Jian She Kai Fa Co., Ltd. (2022 dataset value): secondary industry = Residential Real Estate",
+    "Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd. (2022 dataset value): secondary industry = Retail"
+  ],
+  "milestone": {
+    "Secondary industry of Bao He Hua Chang Jian She Kai Fa Co., Ltd.": "Residential Real Estate",
+    "Secondary industry of Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd.": "Retail",
+    "Comparison result (whether they belong to the same industry)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: secondary industry of Bao He Hua Chang Jian She Kai Fa Co., Ltd. = Residential Real Estate",
+    "Extracted from company_profile.csv: secondary industry of Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd. = Retail",
+    "Compared the two secondary industry texts and judged \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "78c678c3-306c-452a-9b32-dea8f9856975"
+    },
+    {
+      "company_profile": "3a4ca92d-5116-414e-a67c-2e2699774abc"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy054_result.json b/assets/qa_raw/enterprise_industry_analysis/easy054_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f519992bc3b2f5e4da4ce117d6f271c4803a5000
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy054_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy054",
+  "question": "In 2022, are the industries of Bao He Hua Chang Jian She Kai Fa Co., Ltd. and Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd. the same?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Bao He Hua Chang Jian She Kai Fa Co., Ltd. (2022 dataset value): industry = Real Estate",
+    "Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd. (2022 dataset value): industry = Wholesale and Retail"
+  ],
+  "milestone": {
+    "Industry of Bao He Hua Chang Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "Industry of Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd.": "Wholesale and Retail",
+    "Comparison result (whether the industries are the same)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Bao He Hua Chang Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from company_profile.csv: industry of Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd. = Wholesale and Retail",
+    "Compared the two industry texts and judged \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_profile": "78c678c3-306c-452a-9b32-dea8f9856975"
+    },
+    {
+      "company_profile": "3a4ca92d-5116-414e-a67c-2e2699774abc"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy055_result.json b/assets/qa_raw/enterprise_industry_analysis/easy055_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5afc432756a5351a842bf9fb4f2637f197184994
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy055_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy055",
+  "question": "In 2022, compared with Chuang Xin Yao Rui Integrated Circuit Co., Ltd., did the core competitiveness of Ya Wei Ze Zhi Technology Co., Ltd. also emphasize technological innovation and high-quality customer resources?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Ya Wei Ze Zhi Technology Co., Ltd. (based on publicly disclosed 2022 annual report) emphasizes \"technological innovation capability\" and \"high-quality customer resources\" in its core competitiveness description",
+    "Chuang Xin Yao Rui Integrated Circuit Co., Ltd. (based on publicly disclosed 2022 annual report) includes \"R&D innovation advantages\" and \"high-quality customer and brand advantages (customer resources)\" in its core competitiveness description"
+  ],
+  "milestone": {
+    "Core competitiveness keywords of Ya Wei Ze Zhi Technology Co., Ltd.": "Technological innovation capability, high-quality customer resources",
+    "Core competitiveness keywords of Chuang Xin Yao Rui Integrated Circuit Co., Ltd.": "R&D innovation advantages, high-quality customer and brand advantages",
+    "Comparison result (whether both emphasize technological innovation and high-quality customer resources)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted the core competitiveness description of Ya Wei Ze Zhi Technology Co., Ltd. from company_core.csv and identified statements about \"technological innovation capability\" and \"high-quality customer resources\"",
+    "Extracted the core competitiveness description of Chuang Xin Yao Rui Integrated Circuit Co., Ltd. from company_core.csv and identified statements about \"R&D innovation advantages\" and \"high-quality customer and brand advantages (customer resources)\"",
+    "Determined that both companies emphasize technological innovation and high-quality customer resources, and judged \"Yes\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_Core": "4bbcac1c-822c-4e8c-af0e-857826f415ac"
+    },
+    {
+      "company_Core": "7a675a13-5765-4b8c-a98b-da01b53f0de8"
+    },
+    {
+      "company_profile": "26f0fbb2-8470-43a7-8bb7-ff0a4e3893f8"
+    },
+    {
+      "company_profile": "347385e0-b8db-4cb4-834e-deec934c5390"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy056_result.json b/assets/qa_raw/enterprise_industry_analysis/easy056_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..942212d0c71910b022a9c2a0da290fb402b55b59
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy056_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy056",
+  "question": "In 2022, did the core competitiveness of Zhong Ke Sheng Ke Ji Shu Yan Jiu Yuan Co., Ltd. and that of Lian Ji Zhi Sheng Ji Xie Co., Ltd. in the same province show a competitive relationship in terms of technical capabilities?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhong Ke Sheng Ke Ji Shu Yan Jiu Yuan Co., Ltd. (based on publicly disclosed 2022 annual report) emphasizes a biopharmaceutical R&D and production technology platform, including recombinant proteins and mRNA raw-material enzymes",
+    "Lian Ji Zhi Sheng Ji Xie Co., Ltd. (based on publicly disclosed 2022 annual report) emphasizes R&D and manufacturing capabilities for fastening tools such as gas nail guns and construction hardware"
+  ],
+  "milestone": {
+    "Technical focus of Zhong Ke Sheng Ke Ji Shu Yan Jiu Yuan Co., Ltd.": "Biopharmaceutical-related R&D and production technology platform",
+    "Technical focus of Lian Ji Zhi Sheng Ji Xie Co., Ltd.": "R&D and manufacturing of fastening tools and construction hardware",
+    "Comparison result (whether there is a competitive relationship in technical capabilities)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted the core competitiveness description of Zhong Ke Sheng Ke Ji Shu Yan Jiu Yuan Co., Ltd. from company_core.csv and identified that its technical capability focuses on biopharmaceutical R&D and production technology (such as recombinant proteins and mRNA raw-material enzymes)",
+    "Extracted the core competitiveness description of Lian Ji Zhi Sheng Ji Xie Co., Ltd. from company_core.csv and identified that its technical capability focuses on manufacturing R&D for fastening tools (such as gas nail guns) and construction hardware",
+    "Determined that the two companies have different technical directions and the descriptions provide no direct evidence of competition, and judged \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_Core": "5e4de797-ac10-4bb9-ad63-0ec3a6bfc2f8"
+    },
+    {
+      "company_Core": "6c324075-e4cd-4b89-a492-d8f969171bb6"
+    },
+    {
+      "company_profile": "6e79c455-f092-4c77-bb31-e879a67e16dc"
+    },
+    {
+      "company_profile": "c61e9286-d51c-4999-b828-cf5afdd9bcf8"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy057_result.json b/assets/qa_raw/enterprise_industry_analysis/easy057_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba6ba44a29daf6b08681c4f3f00b8ab3ac1b00e9
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy057_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy057",
+  "question": "In 2022, did the products of He Lian Chuang Hang She Bei Co., Ltd. and Shan La Da Chuang Zhi Neng Zhuang Bei Co., Ltd. have a competitive relationship?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "He Lian Chuang Hang She Bei Co., Ltd. (based on publicly disclosed 2022 annual report) shows in its core competitiveness description that its products focus on medical rehabilitation devices and care ecosystems such as pelvic-floor and obstetric rehabilitation",
+    "Shan La Da Chuang Zhi Neng Zhuang Bei Co., Ltd. (based on publicly disclosed 2022 annual report) shows in its core competitiveness description that its products/business include precision mold manufacturing such as lithium-battery equipment cutting dies (and also mentions transportation business)"
+  ],
+  "milestone": {
+    "Product focus of He Lian Chuang Hang She Bei Co., Ltd.": "Medical rehabilitation devices and care ecosystem",
+    "Product focus of Shan La Da Chuang Zhi Neng Zhuang Bei Co., Ltd.": "Precision mold manufacturing such as lithium-battery equipment cutting dies",
+    "Comparison result (whether there is a competitive relationship between products)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted the core competitiveness description of He Lian Chuang Hang She Bei Co., Ltd. from company_core.csv and identified its main product focus as medical rehabilitation devices such as pelvic-floor and obstetric rehabilitation",
+    "Extracted the core competitiveness description of Shan La Da Chuang Zhi Neng Zhuang Bei Co., Ltd. from company_core.csv and identified its main product/business focus as precision mold manufacturing such as lithium-battery equipment cutting dies (with other business segments)",
+    "Determined that the two companies have different product directions and the descriptions provide no direct evidence of direct competition in similar products, and judged \"No\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_Core": "612c090a-8fb3-4b97-a39b-00b18ee5f31e"
+    },
+    {
+      "company_Core": "10a60422-1372-44f3-a720-fb2032f16d0d"
+    },
+    {
+      "company_profile": "d9dc0097-153c-4181-a39d-a177e4de5f78"
+    },
+    {
+      "company_profile": "3c052c5e-2633-477c-bbe9-2dfa3ca75ab1"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy058_result.json b/assets/qa_raw/enterprise_industry_analysis/easy058_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b7bb5f10e49adcf67990d19bd2d98697776e3de1
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy058_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy058",
+  "question": "In 2022, compared with the products of Jiejie Dahang Equipment Co., Ltd., are the products of Sansan Gongzhi Technology Co., Ltd. in a competitive relationship?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "According to the publicly disclosed 2022 annual report of Sansan Gongzhi Technology Co., Ltd., its core competitiveness description shows that its products cover medical diagnostic and monitoring devices such as blood oxygen, monitoring, ECG, ultrasound, and blood pressure.",
+    "According to the publicly disclosed 2022 annual report of Jiejie Dahang Equipment Co., Ltd., its core competitiveness description shows that its products are precision automation equipment for the electronics industry (such as printing equipment, dispensing equipment, and die-bonding equipment)."
+  ],
+  "milestone": {
+    "Product focus of Sansan Gongzhi Technology Co., Ltd.": "Medical diagnostic and monitoring devices",
+    "Product focus of Jiejie Dahang Equipment Co., Ltd.": "Precision automation equipment for the electronics industry",
+    "Comparison result (whether there is a competitive relationship between the products)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_core.csv the core competitiveness description of Sansan Gongzhi Technology Co., Ltd. and identified its product focus as medical diagnostic and monitoring devices (blood oxygen, ECG, ultrasound, etc.).",
+    "Extracted from company_core.csv the core competitiveness description of Jiejie Dahang Equipment Co., Ltd. and identified its product focus as precision automation equipment for the electronics industry (printing, dispensing, die-bonding, etc.).",
+    "Judged that the two product directions are different, and no direct evidence of head-to-head competition in similar products is provided in the descriptions, so the result is \"No\"."
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_Core": "2124e52b-570e-4892-ba89-1870bfa0103d"
+    },
+    {
+      "company_Core": "1141129a-d648-4d67-b840-632fad2f44e8"
+    },
+    {
+      "company_profile": "0217f60b-ffb8-4df3-b87e-a2f9e42418a9"
+    },
+    {
+      "company_profile": "86a75b5f-6cee-47f2-964e-0114659fe4cd"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy059_result.json b/assets/qa_raw/enterprise_industry_analysis/easy059_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..374d1538389072fec614d2ca7797318a1d5d00a5
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy059_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy059",
+  "question": "In 2022, is Haomei Company's core competitiveness in R&D and technology more focused on innovation and diversification than that of Zhongjin Yeye Resources Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "According to the publicly disclosed 2022 annual report of Haomei Company, its R&D and technology strengths include multiple R&D platforms, multidisciplinary R&D teams, and innovation in new materials and new processes, and involve diversified application directions (aluminum profiles, automotive lightweighting, system doors and windows, etc.).",
+    "According to the publicly disclosed 2022 annual report of Zhongjin Yeye Resources Co., Ltd., its core competitiveness description mainly focuses on region/transportation, product quality, skilled workforce, and management systems, without emphasizing \"innovation and diversification\" as the core R&D and technology theme."
+  ],
+  "milestone": {
+    "R&D and technology focus of Haomei Company": "Innovation and diversified application layout",
+    "R&D and technology focus of Zhongjin Yeye Resources Co., Ltd.": "Region/quality/workforce and management systems (not centered on innovation and diversification)",
+    "Comparison result (whether Haomei Company is more focused on innovation and diversification)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_core.csv the core competitiveness description of Haomei Company and identified statements on R&D platform construction, technological innovation, and multi-field application layout (diversification).",
+    "Extracted from company_core.csv the core competitiveness description of Zhongjin Yeye Resources Co., Ltd. and checked whether it takes R&D innovation and diversification as its core expression.",
+    "Compared the two companies' focus on \"innovation and diversification\" and judged that Haomei Company is more focused, so the output is \"Yes\"."
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_Core": "5cd646e6-54db-4e70-ac59-8b525724cde6"
+    },
+    {
+      "company_Core": "f0abfcb4-f2c9-4204-8c77-cdd1548db7e3"
+    },
+    {
+      "company_profile": "cb910366-6699-44b7-889a-00005de99dd4"
+    },
+    {
+      "company_profile": "572fda82-fcac-4396-b536-f63ab245db9e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy060_result.json b/assets/qa_raw/enterprise_industry_analysis/easy060_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b8c6150ec1733ad77fe99a873fdd8e0d6f3294f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy060_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy060",
+  "question": "In 2022, compared with Sansong Shijin Condiment Co., Ltd.'s strengths in technology development and independent innovation, whose strengths are more focused on market influence: Haishan Weixiang Catering Management Co., Ltd. or Sansong Shijin Condiment Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "According to the publicly disclosed 2022 annual report of Haishan Weixiang Catering Management Co., Ltd., its brand strengths emphasize market influence factors such as brand awareness, honors, and increased brand influence after listing.",
+    "According to the publicly disclosed 2022 annual report of Sansong Shijin Condiment Co., Ltd., its core competitiveness description emphasizes technology development and independent innovation (R&D investment, technology platforms, patents, and product R&D), focusing on technical capability rather than market influence."
+  ],
+  "milestone": {
+    "Strength focus of Haishan Weixiang Catering Management Co., Ltd.": "Brand influence/awareness/honors (market influence)",
+    "Strength focus of Sansong Shijin Condiment Co., Ltd.": "Technology development and independent innovation (technical capability)",
+    "Comparison result (which one is more focused on market influence)": "Haishan Weixiang Catering Management Co., Ltd."
+  },
+  "answer": "Haishan Weixiang Catering Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_core.csv the core competitiveness description of Haishan Weixiang Catering Management Co., Ltd. and identified market influence expressions such as \"brand influence/honors/awareness\".",
+    "Extracted from company_core.csv the core competitiveness description of Sansong Shijin Condiment Co., Ltd. and identified that it mainly emphasizes \"technology development and independent innovation/R&D investment/technology platform\".",
+    "Compared the focus of both companies and judged that the one more focused on market influence is \"Haishan Weixiang Catering Management Co., Ltd.\"."
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_Core": "3e31c4b1-d9da-45a5-bd5b-888b4a722dcc"
+    },
+    {
+      "company_Core": "0289794b-89e2-4b25-83c3-bb07574cefb9"
+    },
+    {
+      "company_profile": "46ca80ac-fd45-4dfe-bbc3-60ed1441a1d3"
+    },
+    {
+      "company_profile": "9f8442b2-cb57-40b4-bcf2-2596934346b2"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy061_result.json b/assets/qa_raw/enterprise_industry_analysis/easy061_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e72b60cc9aac691f561f0e07f19f6b6bb35a0e3f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy061_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "easy061",
+  "question": "In 2022, comparing the R&D and technology of Meineng Dianguang Home Appliances Co., Ltd. with the technological innovation of Lixin Shengyue Intelligent Technology Co., Ltd., which company is more competitive in the connector field?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "According to the publicly disclosed 2022 annual report of Meineng Dianguang Home Appliances Co., Ltd., its core competitiveness description clearly focuses on connector R&D, precision mold design and manufacturing for connectors, and connector product customization.",
+    "According to the publicly disclosed 2022 annual report of Lixin Shengyue Intelligent Technology Co., Ltd., its core competitiveness description focuses on piezoelectric quartz crystal components (resonators/TCXO/TSX, etc.), as well as lithography processes and mass production capability."
+  ],
+  "milestone": {
+    "Technology/product focus of Meineng Dianguang Home Appliances Co., Ltd.": "Connector R&D, precision molds, and connector customization",
+    "Technology/product focus of Lixin Shengyue Intelligent Technology Co., Ltd.": "Piezoelectric quartz crystal components and process mass-production capability",
+    "Comparison result (more competitive in the connector field)": "Meineng Dianguang Home Appliances Co., Ltd."
+  },
+  "answer": "Meineng Dianguang Home Appliances Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_core.csv the core competitiveness description of Meineng Dianguang Home Appliances Co., Ltd. and confirmed its expressions on R&D, mold, and product capabilities in the connector field.",
+    "Extracted from company_core.csv the core competitiveness description of Lixin Shengyue Intelligent Technology Co., Ltd. and confirmed that its main technology and product focus is on piezoelectric quartz crystal components rather than connectors.",
+    "Compared relevance and competitive focus around the \"connector field\" and judged that \"Meineng Dianguang Home Appliances Co., Ltd.\" is more competitive."
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "company_Core": "14a2f376-2c6e-4296-8ec5-384bbe5e2704"
+    },
+    {
+      "company_Core": "55a986f9-771d-435c-92fe-a6feee28b84a"
+    },
+    {
+      "company_profile": "dba435cb-5b1f-4224-a961-60a4d66c06b6"
+    },
+    {
+      "company_profile": "78e10a76-df18-4b17-a8cd-a23bcafee7a9"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy062_result.json b/assets/qa_raw/enterprise_industry_analysis/easy062_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f12253034b906e33df15a3fb5211a041fdd7b9de
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy062_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy062",
+  "question": "In 2022, what is the difference between the number of SSE-listed enterprises in the education industry in Beijing and the number of SSE-listed central state-owned enterprises in the national consumer electronics and electrical industry?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number without units, commas, or any explanatory text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Beijing (2022 dataset value): number of SSE-listed enterprises in the education industry = 2",
+    "Nationwide (2022 dataset value): number of SSE-listed central state-owned enterprises in the consumer electronics and electrical industry = 4"
+  ],
+  "milestone": {
+    "Number of SSE-listed enterprises in Beijing's education industry": 2,
+    "Number of SSE-listed central state-owned enterprises in the national consumer electronics and electrical industry": 4,
+    "Difference (Beijing education SSE-listed count - national consumer electronics central SOE SSE-listed count)": -2.0
+  },
+  "answer": -2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the number of SSE-listed enterprises for Beijing in the education industry is 2.",
+    "Extracted from national_industry_status.csv that the number of SSE-listed central state-owned enterprises in the consumer electronics and electrical industry is 4.",
+    "Calculated the difference: 2 - 4 = -2.0."
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "c6b206c0-40ee-4ae8-ab1d-37903bfb3cd1"
+    },
+    {
+      "national_industry_status": "51401b53-3d2a-4dc5-bda7-8a5fc02dd831"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy063_result.json b/assets/qa_raw/enterprise_industry_analysis/easy063_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..6209c8ec85e1ff7bb438af7c8c2649e2af0e02e9
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy063_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy063",
+  "question": "In 2022, which is lower: the average year-on-year net profit growth rate of Beijing's furniture manufacturing industry or that of the nationwide chemical fiber manufacturing industry?",
+  "guidelines": "The answer must be either \"Beijing Furniture Manufacturing\" or \"Nationwide Chemical Fiber Manufacturing\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Beijing (2022 dataset value): average year-on-year net profit growth rate of furniture manufacturing = -80.49%",
+    "Nationwide (2022 dataset value): average year-on-year net profit growth rate of chemical fiber manufacturing = -12.5773529411765%"
+  ],
+  "milestone": {
+    "Average YoY net profit growth rate of Beijing furniture manufacturing": "-80.49%",
+    "Average YoY net profit growth rate of nationwide chemical fiber manufacturing": "-12.5773529411765%",
+    "Comparison result (which is lower)": "Beijing Furniture Manufacturing"
+  },
+  "answer": "Beijing Furniture Manufacturing",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: average YoY net profit growth rate of Beijing furniture manufacturing = -80.49%",
+    "Extracted from national_industry_status.csv: average YoY net profit growth rate of nationwide chemical fiber manufacturing = -12.5773529411765%",
+    "Compared -80.49 and -12.5773529411765; since -80.49 is lower, output \"Beijing Furniture Manufacturing\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "e863bf3c-22d2-49af-b750-ab620b9a1f6f"
+    },
+    {
+      "national_industry_status": "e53c0704-dbd1-451f-82b2-2a0df9488daf"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy064_result.json b/assets/qa_raw/enterprise_industry_analysis/easy064_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..aaecb6bfaa146213146e729100636da3555511c6
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy064_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy064",
+  "question": "In 2022, which is higher: the average year-on-year operating profit growth rate of Beijing's real estate industry or that of the nationwide information transmission, software, and IT services industry?",
+  "guidelines": "The answer must be either \"Beijing Real Estate\" or \"Nationwide Information Transmission, Software and IT Services\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Beijing (2022 dataset value): average YoY operating profit growth rate of real estate = -45.385%",
+    "Nationwide (2022 dataset value): average YoY operating profit growth rate of information transmission, software and IT services = -137.175403726708%"
+  ],
+  "milestone": {
+    "Average YoY operating profit growth rate of Beijing real estate": "-45.385%",
+    "Average YoY operating profit growth rate of nationwide information transmission, software and IT services": "-137.175403726708%",
+    "Comparison result (which is higher)": "Beijing Real Estate"
+  },
+  "answer": "Beijing Real Estate",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: average YoY operating profit growth rate of Beijing real estate = -45.385%",
+    "Extracted from national_industry_status.csv: average YoY operating profit growth rate of nationwide information transmission, software and IT services = -137.175403726708%",
+    "Compared -45.385 and -137.175403726708; since -45.385 is higher, output \"Beijing Real Estate\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "e89ca2d9-ede0-4a1c-aeaa-977be2dc7e53"
+    },
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy065_result.json b/assets/qa_raw/enterprise_industry_analysis/easy065_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..daf269ca963cc0b6b0b938fb47fc836feadbbcfb
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy065_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy065",
+  "question": "In 2022, what is the difference between the number of Shenzhen Stock Exchange-listed local state-owned enterprises in Beijing's information transmission, software, and IT services industry and the number of Shanghai Stock Exchange-listed foreign-funded enterprises in the nationwide real estate industry?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number without units, commas, or any text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Beijing (2022 dataset value): number of Shenzhen Stock Exchange-listed local state-owned enterprises in information transmission, software and IT services = 4",
+    "Nationwide (2022 dataset value): number of Shanghai Stock Exchange-listed foreign-funded enterprises in real estate = 6"
+  ],
+  "milestone": {
+    "Number of SZSE-listed local state-owned enterprises in Beijing information transmission, software and IT services": 4,
+    "Number of SSE-listed foreign-funded enterprises in nationwide real estate": 6,
+    "Difference (Beijing information services local SOE SZSE count - nationwide real estate foreign-funded SSE count)": -2.0
+  },
+  "answer": -2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: number of SZSE-listed local state-owned enterprises in Beijing information transmission, software and IT services = 4",
+    "Extracted from national_industry_status.csv: number of SSE-listed foreign-funded enterprises in nationwide real estate = 6",
+    "Calculated difference: 4 - 6 = -2.0"
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "e7d170ae-b276-465b-b737-971c81e11168"
+    },
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy066_result.json b/assets/qa_raw/enterprise_industry_analysis/easy066_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5d448db6c948124a752710f4ebeac36d2887414a
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy066_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy066",
+  "question": "In 2022, which value is higher: the minimum cumulative number of invalidated PCT invention patents in Beijing's comprehensive industry, or the same metric in the nationwide pharmaceutical manufacturing industry?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Beijing (2022 dataset value): minimum cumulative invalidated PCT invention patents in comprehensive industry = 0",
+    "Nationwide (2022 dataset value): minimum cumulative invalidated PCT invention patents in pharmaceutical manufacturing = 0"
+  ],
+  "milestone": {
+    "Minimum cumulative invalidated PCT invention patents in Beijing comprehensive industry": 0,
+    "Minimum cumulative invalidated PCT invention patents in nationwide pharmaceutical manufacturing": 0,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: minimum cumulative invalidated PCT invention patents in Beijing comprehensive industry = 0",
+    "Extracted from national_industry_status.csv: minimum cumulative invalidated PCT invention patents in nationwide pharmaceutical manufacturing = 0",
+    "Compared the two values; they are equal, so output \"Equal\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "a27f88c6-ce3c-444f-bc4c-c8fd073db2a9"
+    },
+    {
+      "national_industry_status": "9c71278e-97a3-4867-826d-e139f1dffc24"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy067_result.json b/assets/qa_raw/enterprise_industry_analysis/easy067_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..bcd20d0365c1a4e094774267ed564b4918ef6465
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy067_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy067",
+  "question": "In 2022, which is larger: the number of Shenzhen Stock Exchange-listed foreign-funded enterprises in Guangdong's scientific research and technical services industry, or the number of Shanghai Stock Exchange-listed state-owned institute enterprises in the same industry nationwide?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Guangdong Province (2022 dataset value): number of SZSE-listed foreign-funded enterprises in scientific research and technical services = 1",
+    "Nationwide (2022 dataset value): number of SSE-listed state-owned institute enterprises in scientific research and technical services = 1"
+  ],
+  "milestone": {
+    "Number of SZSE-listed foreign-funded enterprises in Guangdong scientific research and technical services": 1,
+    "Number of SSE-listed state-owned institute enterprises in nationwide scientific research and technical services": 1,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: number of SZSE-listed foreign-funded enterprises in Guangdong scientific research and technical services = 1",
+    "Extracted from national_industry_status.csv: number of SSE-listed state-owned institute enterprises in nationwide scientific research and technical services = 1",
+    "Compared 1 and 1; they are equal, so output \"Equal\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "d3b2b4cc-9f44-422e-a3f1-5dcc6c3906c0"
+    },
+    {
+      "national_industry_status": "83fd9785-8bdc-4a18-ad0a-83f92b9aef7c"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy068_result.json b/assets/qa_raw/enterprise_industry_analysis/easy068_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e8a31e35dbad99208eca5415a4e1027a371ff24
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy068_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy068",
+  "question": "In 2022, which is higher: the average year-over-year change rate of R&D personnel in Guangdong Province's scientific research and technical services industry, or that of the same industry nationwide?",
+  "guidelines": "The answer must be \"National\" or \"Guangdong\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Guangdong Province (2022 dataset value): average year-over-year change rate of R&D personnel in the scientific research and technical services industry = 5.80304347826087 %",
+    "Nationwide (2022 dataset value): average year-over-year change rate of R&D personnel in the scientific research and technical services industry = 18.2512727272727 %"
+  ],
+  "milestone": {
+    "Average year-over-year change rate of R&D personnel in Guangdong Province's scientific research and technical services industry": "5.80304347826087 %",
+    "Average year-over-year change rate of R&D personnel in the nationwide scientific research and technical services industry": "18.2512727272727 %",
+    "Comparison result (which is higher)": "National"
+  },
+  "answer": "National",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the average year-over-year change rate of R&D personnel for Guangdong Province in the scientific research and technical services industry is 5.80304347826087 %.",
+    "Extracted from national_industry_status.csv that the average year-over-year change rate of R&D personnel in the scientific research and technical services industry is 18.2512727272727 %.",
+    "Compared 5.80304347826087 and 18.2512727272727; since 18.2512727272727 is higher, output \"National\"."
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "d3b2b4cc-9f44-422e-a3f1-5dcc6c3906c0"
+    },
+    {
+      "national_industry_status": "83fd9785-8bdc-4a18-ad0a-83f92b9aef7c"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy069_result.json b/assets/qa_raw/enterprise_industry_analysis/easy069_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..da4c040f4f592aa34ad931930cd8e2d316327910
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy069_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy069",
+  "question": "In 2022, what percentage does the total number of employees in Jilin Province's comprehensive industry represent relative to the total number of employees in the same industry nationwide?",
+  "guidelines": "The answer must be \"National\" or \"Jilin\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Jilin Province (2022 dataset value): total number of employees in the comprehensive industry = 0 persons",
+    "Nationwide (2022 dataset value): total number of employees in the comprehensive industry = 439485 persons"
+  ],
+  "milestone": {
+    "Total number of employees in Jilin Province's comprehensive industry (persons)": 0,
+    "Total number of employees in the nationwide comprehensive industry (persons)": 439485,
+    "Percentage (Jilin/National, %)": 0
+  },
+  "answer": 0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the total number of employees in Jilin Province's comprehensive industry is 0 persons.",
+    "Extracted from national_industry_status.csv that the total number of employees in the comprehensive industry is 439485 persons.",
+    "Calculated Jilin's percentage of the national total: 0 / 439485 * 100 = 0."
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "fcca2f07-d872-4e0c-8486-f009d9acf1f8"
+    },
+    {
+      "national_industry_status": "fee7bf00-36e3-4f1a-a296-f5612ad8cedd"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy070_result.json b/assets/qa_raw/enterprise_industry_analysis/easy070_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..64da5ec8af54ead236663b3b188d111491bde4f9
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy070_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy070",
+  "question": "In 2022, comparing the median number of participation in drafting industry standards for Jilin Province's comprehensive industry and the same indicator for the nationwide comprehensive industry, which one is lower?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or a company name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Jilin Province (2022 dataset value): median number of participation in drafting industry standards in the comprehensive industry = 0",
+    "Nationwide (2022 dataset value): median number of participation in drafting industry standards in the comprehensive industry = 0"
+  ],
+  "milestone": {
+    "Median number of participation in drafting industry standards in Jilin Province's comprehensive industry": 0,
+    "Median number of participation in drafting industry standards in the nationwide comprehensive industry": 0,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the median number of participation in drafting industry standards for Jilin Province's comprehensive industry is 0.",
+    "Extracted from national_industry_status.csv that the median number of participation in drafting industry standards for the comprehensive industry is 0.",
+    "Compared 0 and 0, and determined the result is \"Equal\"."
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "fcca2f07-d872-4e0c-8486-f009d9acf1f8"
+    },
+    {
+      "national_industry_status": "fee7bf00-36e3-4f1a-a296-f5612ad8cedd"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy071_result.json b/assets/qa_raw/enterprise_industry_analysis/easy071_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a30f841ca5af3d08450942190e1680b628375dc8
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy071_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy071",
+  "question": "In 2022, which is higher: the maximum asset-liability ratio in Jilin Province's commercial electrical machinery and equipment manufacturing industry, or the same indicator nationwide?",
+  "guidelines": "The answer must be \"National\" or \"Jilin\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Jilin Province (2022 dataset value): maximum asset-liability ratio in the commercial electrical machinery and equipment manufacturing industry = 65.21 %",
+    "Nationwide (2022 dataset value): maximum asset-liability ratio in the commercial electrical machinery and equipment manufacturing industry = 205.31 %"
+  ],
+  "milestone": {
+    "Maximum asset-liability ratio in Jilin Province's commercial electrical machinery and equipment manufacturing industry": "65.21 %",
+    "Maximum asset-liability ratio in the nationwide commercial electrical machinery and equipment manufacturing industry": "205.31 %",
+    "Comparison result (which is higher)": "National"
+  },
+  "answer": "National",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the maximum asset-liability ratio for Jilin Province in the commercial electrical machinery and equipment manufacturing industry is 65.21 %.",
+    "Extracted from national_industry_status.csv that the maximum asset-liability ratio in the commercial electrical machinery and equipment manufacturing industry is 205.31 %.",
+    "Compared 65.21 and 205.31; since 205.31 is higher, output \"National\"."
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "495b3b2e-7e1b-463a-b69e-0c358f052a29"
+    },
+    {
+      "national_industry_status": "15263bd4-fe7b-43f3-bd79-b697b63a42b7"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy072_result.json b/assets/qa_raw/enterprise_industry_analysis/easy072_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c7037d96dbf3428837b98e97d3e3a6a9051ca19c
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy072_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy072",
+  "question": "In 2022, which is larger: the number of Shanghai Stock Exchange-listed Sino-foreign joint venture enterprises in Guangdong's semiconductor industry, or the number of Shenzhen Stock Exchange-listed state-owned institute enterprises in the nationwide semiconductor industry?",
+  "guidelines": "The answer must be \"Equal\" or the comparison conclusion term specified in the question. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Guangdong Province (2022 dataset value): number of SSE-listed Sino-foreign joint venture enterprises in the semiconductor industry = 1",
+    "Nationwide (2022 dataset value): number of SZSE-listed state-owned institute enterprises in the semiconductor industry = 1"
+  ],
+  "milestone": {
+    "Number of SSE-listed Sino-foreign joint venture enterprises in Guangdong semiconductor industry": 1,
+    "Number of SZSE-listed state-owned institute enterprises in nationwide semiconductor industry": 1,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: number of SSE-listed Sino-foreign joint venture enterprises in Guangdong semiconductor industry = 1",
+    "Extracted from national_industry_status.csv: number of SZSE-listed state-owned institute enterprises in nationwide semiconductor industry = 1",
+    "Compared 1 and 1, and judged \"Equal\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "4e5494a2-c0e7-4e31-a2ba-540b2abcf794"
+    },
+    {
+      "national_industry_status": "c6c62a37-7d79-4959-9891-195fcadfbb02"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy073_result.json b/assets/qa_raw/enterprise_industry_analysis/easy073_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..6026a028ab7fee3008054b499e5d81d20e68e905
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy073_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy073",
+  "question": "In 2022, which is larger: the total number of enterprises in Jilin's transportation, warehousing, and postal industry, or the number of Shenzhen Stock Exchange-listed central state-owned enterprises in the same industry nationwide?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Jilin Province (2022 dataset value): total enterprises in transportation, warehousing and postal industry = 1",
+    "Nationwide (2022 dataset value): number of SZSE-listed central state-owned enterprises in transportation, warehousing and postal industry = 5"
+  ],
+  "milestone": {
+    "Total enterprises in Jilin transportation, warehousing and postal industry": 1,
+    "Number of SZSE-listed central state-owned enterprises in nationwide transportation, warehousing and postal industry": 5,
+    "Comparison result (which is larger)": "Nationwide"
+  },
+  "answer": "Nationwide",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: total enterprises in Jilin transportation, warehousing and postal industry = 1",
+    "Extracted from national_industry_status.csv: number of SZSE-listed central state-owned enterprises in nationwide transportation, warehousing and postal industry = 5",
+    "Compared 1 and 5; since 5 is larger, the nationwide-side metric is larger, so output \"Nationwide\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "73d46d5c-7faa-481a-b05a-bd1688f3b472"
+    },
+    {
+      "national_industry_status": "c5e98e13-e208-406f-aa43-9e029e2c4de6"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy074_result.json b/assets/qa_raw/enterprise_industry_analysis/easy074_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..44fa5afe3d5332d459f34831fa045273c603342d
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy074_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy074",
+  "question": "In 2022, which is higher: the maximum government award funding or subsidy value in Jilin's transportation, warehousing, and postal industry, or the same metric in the nationwide industry?",
+  "guidelines": "The answer must be \"Nationwide\" or \"Jilin\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Jilin Province (2022 dataset value): maximum government award funding or subsidy in transportation, warehousing and postal industry = 1244449.8 yuan",
+    "Nationwide (2022 dataset value): maximum government award funding or subsidy in transportation, warehousing and postal industry = 4688008342 yuan"
+  ],
+  "milestone": {
+    "Maximum government award funding or subsidy in Jilin transportation, warehousing and postal industry (yuan)": 1244449.8,
+    "Maximum government award funding or subsidy in nationwide transportation, warehousing and postal industry (yuan)": 4688008342,
+    "Comparison result (which is higher)": "Nationwide"
+  },
+  "answer": "Nationwide",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: maximum government award funding or subsidy in Jilin transportation, warehousing and postal industry = 1244449.8 yuan",
+    "Extracted from national_industry_status.csv: maximum government award funding or subsidy in nationwide transportation, warehousing and postal industry = 4688008342 yuan",
+    "Compared 1244449.8 and 4688008342; since 4688008342 is higher, output \"Nationwide\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "73d46d5c-7faa-481a-b05a-bd1688f3b472"
+    },
+    {
+      "national_industry_status": "c5e98e13-e208-406f-aa43-9e029e2c4de6"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy075_result.json b/assets/qa_raw/enterprise_industry_analysis/easy075_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..60a7e807840973048aed58d878ade11adfeafcc0
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy075_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy075",
+  "question": "In 2022, which is larger in the Tibet Autonomous Region: the total number of enterprises in the electricity, heat, gas, and water production and supply industry, or in the construction industry?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Tibet Autonomous Region (2022 dataset value): total enterprises in electricity, heat, gas, and water production and supply industry = 0",
+    "Tibet Autonomous Region (2022 dataset value): total enterprises in construction industry = 0"
+  ],
+  "milestone": {
+    "Total enterprises in Tibet electricity, heat, gas, and water production and supply industry": 0,
+    "Total enterprises in Tibet construction industry": 0,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: total enterprises in Tibet electricity, heat, gas, and water production and supply industry = 0",
+    "Extracted from regional_industry_status.csv: total enterprises in Tibet construction industry = 0",
+    "Compared 0 and 0, and judged \"Equal\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "6cfae52b-5907-4f04-bbae-fe07d0314711"
+    },
+    {
+      "regional_industry_status": "7ce8065a-1f12-4137-a29f-4c4c60a31e44"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy076_result.json b/assets/qa_raw/enterprise_industry_analysis/easy076_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e87bf364d546103eeafb26745651294f990a32da
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy076_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy076",
+  "question": "In 2022, is there any difference between the total number of enterprises in the Tibet Autonomous Region's electricity, heat, gas, and water production and supply industry and that in the construction industry?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Tibet Autonomous Region (2022 dataset value): total enterprises in electricity, heat, gas, and water production and supply industry = 0",
+    "Tibet Autonomous Region (2022 dataset value): total enterprises in construction industry = 0"
+  ],
+  "milestone": {
+    "Total enterprises in Tibet electricity, heat, gas, and water production and supply industry": 0,
+    "Total enterprises in Tibet construction industry": 0,
+    "Comparison result (whether there is a difference)": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: total enterprises in Tibet electricity, heat, gas, and water production and supply industry = 0",
+    "Extracted from regional_industry_status.csv: total enterprises in Tibet construction industry = 0",
+    "Compared whether the two values are the same, and judged \"Equal\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "6cfae52b-5907-4f04-bbae-fe07d0314711"
+    },
+    {
+      "regional_industry_status": "7ce8065a-1f12-4137-a29f-4c4c60a31e44"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy077_result.json b/assets/qa_raw/enterprise_industry_analysis/easy077_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c2c4a095d7a6d883ce15d534c086c21f2e4e451
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy077_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy077",
+  "question": "In 2022, which is larger in the Tibet Autonomous Region: the total number of enterprises in the electricity, heat, gas, and water production and supply industry, or in the leasing and business services industry?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Tibet Autonomous Region (2022 dataset value): total enterprises in electricity, heat, gas, and water production and supply industry = 0",
+    "Tibet Autonomous Region (2022 dataset value): total enterprises in leasing and business services industry = 0"
+  ],
+  "milestone": {
+    "Total enterprises in Tibet electricity, heat, gas, and water production and supply industry": 0,
+    "Total enterprises in Tibet leasing and business services industry": 0,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: total enterprises in Tibet electricity, heat, gas, and water production and supply industry = 0",
+    "Extracted from regional_industry_status.csv: total enterprises in Tibet leasing and business services industry = 0",
+    "Compared 0 and 0, and judged \"Equal\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "6cfae52b-5907-4f04-bbae-fe07d0314711"
+    },
+    {
+      "regional_industry_status": "0e9da0e4-0204-41ab-b90c-e92983120b7c"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy078_result.json b/assets/qa_raw/enterprise_industry_analysis/easy078_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f01e417ecc2161b7e650382a5c6d993879c9f229
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy078_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy078",
+  "question": "In 2022, is there a difference between the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry and that in the commercial electrical machinery and equipment manufacturing industry?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or a company name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Tibet Autonomous Region (2022 dataset value): total number of enterprises in the electricity, heat, gas and water production and supply industry = 0",
+    "Tibet Autonomous Region (2022 dataset value): total number of enterprises in the commercial electrical machinery and equipment manufacturing industry = 0"
+  ],
+  "milestone": {
+    "Total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry": 0,
+    "Total number of enterprises in Tibet Autonomous Region's commercial electrical machinery and equipment manufacturing industry": 0,
+    "Comparison result (whether there is a difference)": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry is 0.",
+    "Extracted from regional_industry_status.csv that the total number of enterprises in Tibet Autonomous Region's commercial electrical machinery and equipment manufacturing industry is 0.",
+    "Compared whether the two values are consistent, and determined \"Equal\"."
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "6cfae52b-5907-4f04-bbae-fe07d0314711"
+    },
+    {
+      "regional_industry_status": "aa3787d1-ec43-4243-ab22-72afa293068b"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy079_result.json b/assets/qa_raw/enterprise_industry_analysis/easy079_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f417df144849e2a9f64691ef28a0612e07c4ef92
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy079_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy079",
+  "question": "In 2022, is the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry the same as that in the metal products industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Tibet Autonomous Region (2022 dataset value): total number of enterprises in the electricity, heat, gas and water production and supply industry = 0",
+    "Tibet Autonomous Region (2022 dataset value): total number of enterprises in the metal products industry = 0"
+  ],
+  "milestone": {
+    "Total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry": 0,
+    "Total number of enterprises in Tibet Autonomous Region's metal products industry": 0,
+    "Comparison result (whether the counts are the same)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry is 0.",
+    "Extracted from regional_industry_status.csv that the total number of enterprises in Tibet Autonomous Region's metal products industry is 0.",
+    "Compared the two counts and determined \"Yes\" because they are the same."
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "6cfae52b-5907-4f04-bbae-fe07d0314711"
+    },
+    {
+      "regional_industry_status": "fe9ea1bf-ccc2-42a0-a9d1-1240d3e24000"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy080_result.json b/assets/qa_raw/enterprise_industry_analysis/easy080_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..8ab3f8ba033ab9ffbc4c3e0dc96cea4710b3f331
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy080_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy080",
+  "question": "In 2022, comparing the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry with that in the general equipment manufacturing industry, which industry has more enterprises?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or a company name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Tibet Autonomous Region (2022 dataset value): total number of enterprises in the electricity, heat, gas and water production and supply industry = 0",
+    "Tibet Autonomous Region (2022 dataset value): total number of enterprises in the general equipment manufacturing industry = 0"
+  ],
+  "milestone": {
+    "Total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry": 0,
+    "Total number of enterprises in Tibet Autonomous Region's general equipment manufacturing industry": 0,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry is 0.",
+    "Extracted from regional_industry_status.csv that the total number of enterprises in Tibet Autonomous Region's general equipment manufacturing industry is 0.",
+    "Compared 0 and 0, and determined \"Equal\"."
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "6cfae52b-5907-4f04-bbae-fe07d0314711"
+    },
+    {
+      "regional_industry_status": "b673d26a-c5dd-45f6-99ee-61d9d90bb07c"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy081_result.json b/assets/qa_raw/enterprise_industry_analysis/easy081_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..91a25901b26f95428f5f43cc773cc54e69d383e7
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy081_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy081",
+  "question": "In 2022, which is greater: the total number of enterprises in Jilin Province's chemical raw materials and chemical products manufacturing industry, or the number of SSE-listed enterprises in the same industry in Qinghai Province?",
+  "guidelines": "The answer must be \"Equal\", \"Qinghai\", or \"Jilin\" (as specified by the question). If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Jilin Province (2022 dataset value): total number of enterprises in the chemical raw materials and chemical products manufacturing industry = 1",
+    "Qinghai Province (2022 dataset value): number of SSE-listed enterprises in the chemical raw materials and chemical products manufacturing industry = 1"
+  ],
+  "milestone": {
+    "Total number of enterprises in Jilin Province's chemical raw materials and chemical products manufacturing industry": 1,
+    "Number of SSE-listed enterprises in Qinghai Province's chemical raw materials and chemical products manufacturing industry": 1,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv that the total number of enterprises in Jilin Province's chemical raw materials and chemical products manufacturing industry is 1.",
+    "Extracted from regional_industry_status.csv that the number of SSE-listed enterprises in Qinghai Province's chemical raw materials and chemical products manufacturing industry is 1.",
+    "Compared 1 and 1, and determined \"Equal\"."
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "f24fc7c6-6169-41bb-9ad0-f5a00950662d"
+    },
+    {
+      "regional_industry_status": "6f976482-4bdd-4b06-a4ef-24f45ec2b096"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy082_result.json b/assets/qa_raw/enterprise_industry_analysis/easy082_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..aed9fa2293b3deb4a0d5b3557a1009a5160cf452
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy082_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy082",
+  "question": "In 2022, which is higher: the number of SZSE-listed private enterprises in Jilin's chemical raw materials and chemical products manufacturing industry, or the number of SSE-listed local state-owned enterprises in the same industry in Qinghai?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or company name without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Jilin Province (2022 dataset value): number of SZSE-listed private enterprises in chemical raw materials and chemical products manufacturing = 1",
+    "Qinghai Province (2022 dataset value): number of SSE-listed local state-owned enterprises in chemical raw materials and chemical products manufacturing = 1"
+  ],
+  "milestone": {
+    "Number of SZSE-listed private enterprises in Jilin chemical raw materials and chemical products manufacturing": 1,
+    "Number of SSE-listed local state-owned enterprises in Qinghai chemical raw materials and chemical products manufacturing": 1,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: SZSE-listed private enterprise count in Jilin chemical raw materials and chemical products manufacturing = 1",
+    "Extracted from regional_industry_status.csv: SSE-listed local state-owned enterprise count in Qinghai chemical raw materials and chemical products manufacturing = 1",
+    "Compared 1 and 1, and judged \"Equal\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "f24fc7c6-6169-41bb-9ad0-f5a00950662d"
+    },
+    {
+      "regional_industry_status": "6f976482-4bdd-4b06-a4ef-24f45ec2b096"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy083_result.json b/assets/qa_raw/enterprise_industry_analysis/easy083_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..972d393e30ba12c5bf75ab41410eae00810d7621
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy083_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy083",
+  "question": "In 2022, which is higher: the average capitalized R&D investment in Jilin's petroleum processing, coking, and nuclear fuel processing industry, or the same metric in Xinjiang?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or company name without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Jilin Province (2022 dataset value): average capitalized R&D investment in petroleum processing, coking and nuclear fuel processing = 0 yuan",
+    "Xinjiang Uygur Autonomous Region (2022 dataset value): average capitalized R&D investment in petroleum processing, coking and nuclear fuel processing = 0 yuan"
+  ],
+  "milestone": {
+    "Average capitalized R&D investment in Jilin petroleum processing, coking and nuclear fuel processing (yuan)": 0,
+    "Average capitalized R&D investment in Xinjiang petroleum processing, coking and nuclear fuel processing (yuan)": 0,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from regional_industry_status.csv: average capitalized R&D investment in Jilin petroleum processing, coking and nuclear fuel processing = 0 yuan",
+    "Extracted from regional_industry_status.csv: average capitalized R&D investment in Xinjiang petroleum processing, coking and nuclear fuel processing = 0 yuan",
+    "Compared 0 and 0, and judged \"Equal\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "regional_industry_status": "4b3f72ad-ca2e-4e4a-9f57-6c2ca75fd374"
+    },
+    {
+      "regional_industry_status": "bdb752fb-d068-4df6-a617-41ae6fddfcfe"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy084_result.json b/assets/qa_raw/enterprise_industry_analysis/easy084_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..94be5b0a74c66b1e2d7f03dc666123008ebb781b
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy084_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy084",
+  "question": "In 2022, which is larger: the number of HKEX-listed central state-owned enterprises in China's petroleum processing, coking, and nuclear fuel processing industry, or the total number of enterprises in the textiles, footwear, and apparel industry?",
+  "guidelines": "The answer must be either \"Petroleum Processing, Coking and Nuclear Fuel Processing\" or \"Textiles, Footwear and Apparel\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Nationwide (2022 dataset value): number of HKEX-listed central state-owned enterprises in petroleum processing, coking and nuclear fuel processing = 2",
+    "Nationwide (2022 dataset value): enterprise count in textiles, footwear and apparel = 177"
+  ],
+  "milestone": {
+    "Number of HKEX-listed central state-owned enterprises in nationwide petroleum processing, coking and nuclear fuel processing": 2,
+    "Enterprise count in nationwide textiles, footwear and apparel": 177,
+    "Comparison result (which is larger)": "Textiles, Footwear and Apparel"
+  },
+  "answer": "Textiles, Footwear and Apparel",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from national_industry_status.csv: HKEX-listed central state-owned enterprise count in petroleum processing, coking and nuclear fuel processing = 2",
+    "Extracted from national_industry_status.csv: enterprise count in textiles, footwear and apparel = 177",
+    "Compared 2 and 177; since 177 is larger, output \"Textiles, Footwear and Apparel\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "national_industry_status": "655b2fa1-a92d-49c6-9093-92c5a222ba75"
+    },
+    {
+      "national_industry_status": "84366a3d-230e-40ab-9da7-166ad7ac948a"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy085_result.json b/assets/qa_raw/enterprise_industry_analysis/easy085_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3e64c4ea80bd66ad5114c72b10270b88f5c99e03
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy085_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy085",
+  "question": "In 2022, which is higher: the median number of National Technological Invention Awards in China's accommodation and catering industry, or in the real estate industry?",
+  "guidelines": "The answer must be \"Equal\", a company name, or \"industry\". Output only one word or company name without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Nationwide (2022 dataset value): median number of National Technological Invention Awards in accommodation and catering industry = 0",
+    "Nationwide (2022 dataset value): median number of National Technological Invention Awards in real estate industry = 0"
+  ],
+  "milestone": {
+    "Median National Technological Invention Awards in nationwide accommodation and catering industry": 0,
+    "Median National Technological Invention Awards in nationwide real estate industry": 0,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from national_industry_status.csv: median National Technological Invention Awards in accommodation and catering industry = 0",
+    "Extracted from national_industry_status.csv: median National Technological Invention Awards in real estate industry = 0",
+    "Compared 0 and 0, and judged \"Equal\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "national_industry_status": "9f31ba81-0925-47ba-93e1-fdd88dd6c638"
+    },
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy086_result.json b/assets/qa_raw/enterprise_industry_analysis/easy086_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..51052a578ec618afb03a7df9c9fb0ea60d9f0ef8
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy086_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy086",
+  "question": "In 2022, which is lower: the minimum cumulative number of Chinese invention patent applications in China's automobile manufacturing industry, or in the metal smelting and rolling processing industry?",
+  "guidelines": "The answer must be an industry name. Output only one term without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Nationwide (2022 dataset value): minimum cumulative Chinese invention patent applications in automobile manufacturing = 2",
+    "Nationwide (2022 dataset value): minimum cumulative Chinese invention patent applications in metal smelting and rolling processing = 1"
+  ],
+  "milestone": {
+    "Minimum cumulative Chinese invention patent applications in nationwide automobile manufacturing": 2,
+    "Minimum cumulative Chinese invention patent applications in nationwide metal smelting and rolling processing": 1,
+    "Comparison result (which is lower)": "Metal Smelting and Rolling Processing"
+  },
+  "answer": "Metal Smelting and Rolling Processing",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from national_industry_status.csv: minimum cumulative Chinese invention patent applications in automobile manufacturing = 2",
+    "Extracted from national_industry_status.csv: minimum cumulative Chinese invention patent applications in metal smelting and rolling processing = 1",
+    "Compared 2 and 1; since 1 is lower, output \"Metal Smelting and Rolling Processing\""
+  ],
+  "steps_num": 3,
+  "reference": [
+    {
+      "national_industry_status": "253f34de-05e4-42e4-959f-746cedefc598"
+    },
+    {
+      "national_industry_status": "e12e3329-0ab1-4bf1-b19d-0a5c9610b5ce"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy088_result.json b/assets/qa_raw/enterprise_industry_analysis/easy088_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce234ad65a354bca669b14473039e578a36a11b2
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy088_result.json
@@ -0,0 +1,33 @@
+{
+  "id": "easy088",
+  "question": "Were the policy \"Notice on Qualification Recognition Matters for Relevant R&D Institutions in Pudong New Area of Shanghai Applicable to Import Tax Policies\" and the policy \"Notice of the General Office of the Shanghai Municipal People's Government on Issuing the Action Plan for Cultivating the New Metaverse Track\" issued by the same department?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Issuing authorities of the policy \"Notice on Qualification Recognition Matters for Relevant R&D Institutions in Pudong New Area of Shanghai Applicable to Import Tax Policies\": Ministry of Finance, Ministry of Science and Technology, Ministry of Civil Affairs, Ministry of Commerce, General Administration of Customs, State Taxation Administration",
+    "Issuing authority of the policy \"Notice of the General Office of the Shanghai Municipal People's Government on Issuing the Action Plan for Cultivating the New Metaverse Track\": General Office of the Shanghai Municipal People's Government"
+  ],
+  "milestone": {
+    "Issuing authorities of Policy 1": "Ministry of Finance, Ministry of Science and Technology, Ministry of Civil Affairs, Ministry of Commerce, General Administration of Customs, State Taxation Administration",
+    "Issuing authority of Policy 2": "General Office of the Shanghai Municipal People's Government",
+    "Comparison result (whether issued by the same department)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted the issuing authority fields of the two policies from policy_resource.csv.",
+    "Compared the issuing authority texts of the two policies, found them inconsistent, and judged \"No\"."
+  ],
+  "steps_num": 2,
+  "reference": [
+    {
+      "policy_resource": "905442dc-7648-4b5b-bf47-3f4bb4316606"
+    },
+    {
+      "policy_resource": "7a245521-9eb6-4da0-ad68-8ae98ffb7103"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy090_result.json b/assets/qa_raw/enterprise_industry_analysis/easy090_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e2e679c2d9b265dc5f6d002b814dbd0c7c709a81
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy090_result.json
@@ -0,0 +1,37 @@
+{
+  "id": "easy090",
+  "question": "Between the minimum change in the R&D expenditure ratio of the Information Transmission, Software and IT Services industry and that of Other Manufacturing in China, which one is smaller?",
+  "guidelines": "The answer must be a single number with two decimal places. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Information Transmission, Software and IT Services: minimum change in R&D expenditure ratio -472 %",
+    "Other Manufacturing (China): minimum change in R&D expenditure ratio -19.89 %"
+  ],
+  "steps": [
+    "Extract from national_industry_status.csv that the minimum change in the R&D expenditure ratio for Information Transmission, Software and IT Services is -472 %.",
+    "Extract from national_industry_status.csv that the minimum change in the R&D expenditure ratio for Other Manufacturing is -19.89 %.",
+    "Compare the two minimum values: -472 % < -19.89 %, so Information Transmission, Software and IT Services is smaller."
+  ],
+  "steps_num": 3,
+  "answer": "Information Transmission, Software and IT Services",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Information Transmission, Software and IT Services minimum change in R&D expenditure ratio": -472,
+    "Other Manufacturing minimum change in R&D expenditure ratio": -19.89,
+    "Comparison result (smaller one)": "Information Transmission, Software and IT Services"
+  },
+  "reference": [
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "national_industry_status": "15b9907e-9b45-4636-bb57-dee26f48b2d3"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy091_result.json b/assets/qa_raw/enterprise_industry_analysis/easy091_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a5b751a66307f148b63bde1f67358f3f4e385b44
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy091_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy091",
+  "question": "Are Wuli Changyuan Wholesale Company and Xinhua Yuantong Chain Company in a competitive relationship?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Wuli Changyuan Wholesale Company (data in this dataset) belongs to the Wholesale and Retail industry",
+    "Xinhua Yuantong Chain Company (data in this dataset) belongs to the Wholesale and Retail industry"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Wuli Changyuan Wholesale Company is in the Wholesale and Retail industry.",
+    "Extract from company_profile.csv that Xinhua Yuantong Chain Company is in the Wholesale and Retail industry.",
+    "Since the two companies are in the same industry, they are in a competitive relationship; output \"Yes\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Wuli Changyuan Wholesale Company": "Wholesale and Retail",
+    "Industry of Xinhua Yuantong Chain Company": "Wholesale and Retail",
+    "Competitive relationship": "Yes"
+  },
+  "reference": [
+    {
+      "company_profile": "4a50579b-d523-4bca-8dcd-ee5defe6d541"
+    },
+    {
+      "company_profile": "9c9676ca-abd3-43a3-b4cb-1a845480f73a"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy092_result.json b/assets/qa_raw/enterprise_industry_analysis/easy092_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..141137d9a4150abbac202f25a3dd33a3b7719813
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy092_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy092",
+  "question": "Are Huadianeng Jin Hydropower Company and Huaneng Zeze New Energy Company in a competitive relationship?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Huadianeng Jin Hydropower Company (data in this dataset) belongs to the Electricity, Heat, Gas and Water Production and Supply industry",
+    "Huaneng Zeze New Energy Company (data in this dataset) belongs to the Electricity, Heat, Gas and Water Production and Supply industry"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Huadianeng Jin Hydropower Company is in the Electricity, Heat, Gas and Water Production and Supply industry.",
+    "Extract from company_profile.csv that Huaneng Zeze New Energy Company is in the Electricity, Heat, Gas and Water Production and Supply industry.",
+    "Since the two companies are in the same industry, a competitive relationship exists; output \"Yes\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Huadianeng Jin Hydropower Company": "Electricity, Heat, Gas and Water Production and Supply",
+    "Industry of Huaneng Zeze New Energy Company": "Electricity, Heat, Gas and Water Production and Supply",
+    "Competitive relationship": "Yes"
+  },
+  "reference": [
+    {
+      "company_profile": "34cc539e-12b3-49ae-aecf-bc295add8fba"
+    },
+    {
+      "company_profile": "3e48d3e7-d49c-4dba-8a1f-e6288cea6a0b"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy093_result.json b/assets/qa_raw/enterprise_industry_analysis/easy093_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c213518dc9546e069583541dc2c0ae8d02028c4
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy093_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy093",
+  "question": "Are Run Hui Shu Ke Technology Co., Ltd. and Hang Fa Tie Chuan Hang Kong Technology Co., Ltd. in the same industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Hang Fa Tie Chuan Hang Kong Technology Co., Ltd. (data in this dataset) industry: Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "Run Hui Shu Ke Technology Co., Ltd. (data in this dataset) industry: Information Transmission, Software and IT Services"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Hang Fa Tie Chuan Hang Kong Technology Co., Ltd.'s industry is Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "From company_profile.csv, extract that Run Hui Shu Ke Technology Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "The two companies are in different industries; they are not peers; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Hang Fa Tie Chuan Hang Kong Technology Co., Ltd. industry": "Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "Run Hui Shu Ke Technology Co., Ltd. industry": "Information Transmission, Software and IT Services",
+    "Whether same industry": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "a85ad273-d7ab-409c-ac1d-2c5f5cb20bb6"
+    },
+    {
+      "company_profile": "c23a6b65-7292-462b-807e-188f077858f8"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy094_result.json b/assets/qa_raw/enterprise_industry_analysis/easy094_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d4645f88c9b29fae0c16bd28483209bbc9551091
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy094_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy094",
+  "question": "Are Zhong Fang Chang Da Zhong Gong Co., Ltd. and Bao Xin Zhi Zhi Xi Tong Co., Ltd. in a competitive relationship?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhong Fang Chang Da Zhong Gong Co., Ltd. (data in this dataset) industry: Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "Bao Xin Zhi Zhi Xi Tong Co., Ltd. (data in this dataset) industry: Information Transmission, Software and IT Services"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong Fang Chang Da Zhong Gong Co., Ltd.'s industry is Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "From company_profile.csv, extract that Bao Xin Zhi Zhi Xi Tong Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "The two companies are in different industries; they do not constitute a competitive relationship; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Zhong Fang Chang Da Zhong Gong Co., Ltd. industry": "Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "Bao Xin Zhi Zhi Xi Tong Co., Ltd. industry": "Information Transmission, Software and IT Services",
+    "Competitive relationship": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "e0736bdb-d048-4166-bf35-d2288eb8fd36"
+    },
+    {
+      "company_profile": "e5323046-007d-480d-ad3c-8c2dd4d536da"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy095_result.json b/assets/qa_raw/enterprise_industry_analysis/easy095_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5b7250140fb0d85f61c5e7f477117b0b3b9cc78e
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy095_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy095",
+   "question": "Are Da Zu Jin Jing She Bei Co., Ltd. and Xi Fen Ye Jin Jin Shu Co., Ltd. competitors?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Da Zu Jin Jing She Bei Co., Ltd. (data in this dataset) industry: Other Manufacturing",
+    "Xi Fen Ye Jin Jin Shu Co., Ltd. (data in this dataset) industry: Metal Smelting and Rolling Processing"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Da Zu Jin Jing She Bei Co., Ltd.'s industry is Other Manufacturing",
+    "From company_profile.csv, extract that Xi Fen Ye Jin Jin Shu Co., Ltd.'s industry is Metal Smelting and Rolling Processing",
+    "The two companies are in different industries; they are not competitors; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Da Zu Jin Jing She Bei Co., Ltd. industry": "Other Manufacturing",
+    "Xi Fen Ye Jin Jin Shu Co., Ltd. industry": "Metal Smelting and Rolling Processing",
+    "Whether competitors": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "2b8f66b2-53b1-4f40-9afe-43c043a9d88e"
+    },
+    {
+      "company_profile": "2b7eab91-ea44-479a-9247-44821a7adc95"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy096_result.json b/assets/qa_raw/enterprise_industry_analysis/easy096_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..be4cc79fe11eccb531f9055f22aaa099dc70a9d0
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy096_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy096",
+  "question": "Are Lv Tai Jie Xun Huan Bao Technology Co., Ltd. and Feng Huo Chuang Ze Wang Luo She Bei Co., Ltd. competitors?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Lv Tai Jie Xun Huan Bao Technology Co., Ltd. (data in this dataset) industry: Comprehensive Utilization of Waste Resources",
+    "Feng Huo Chuang Ze Wang Luo She Bei Co., Ltd. (data in this dataset) industry: Communication Transmission Equipment"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Lv Tai Jie Xun Huan Bao Technology Co., Ltd.'s industry is Comprehensive Utilization of Waste Resources",
+    "From company_profile.csv, extract that Feng Huo Chuang Ze Wang Luo She Bei Co., Ltd.'s industry is Communication Transmission Equipment",
+    "The two companies are in different industries; they are not competitors; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Lv Tai Jie Xun Huan Bao Technology Co., Ltd. industry": "Comprehensive Utilization of Waste Resources",
+    "Feng Huo Chuang Ze Wang Luo She Bei Co., Ltd. industry": "Communication Transmission Equipment",
+    "Whether competitors": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "b4f31c28-8425-469f-92a3-951c94513333"
+    },
+    {
+      "company_profile": "daa110b2-59d6-4541-a13b-0f00a61028d0"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy097_result.json b/assets/qa_raw/enterprise_industry_analysis/easy097_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..82c8f1a3b97a59d2548fe1c18de3559c83be6c9d
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy097_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy097",
+  "question": "Are Lv Shan Zhi Jin Real Estate Development Co., Ltd. and Huan Qiu Tai Jin Zhi Neng Dian Qi Co., Ltd. in the same industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Huan Qiu Tai Jin Zhi Neng Dian Qi Co., Ltd. (data in this dataset) industry: Commercial Electrical Machinery and Equipment Manufacturing",
+    "Lv Shan Zhi Jin Real Estate Development Co., Ltd. (data in this dataset) industry: Real Estate"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Huan Qiu Tai Jin Zhi Neng Dian Qi Co., Ltd.'s industry is Commercial Electrical Machinery and Equipment Manufacturing",
+    "From company_profile.csv, extract that Lv Shan Zhi Jin Real Estate Development Co., Ltd.'s industry is Real Estate",
+    "The two companies are in different industries; they are not peers; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Huan Qiu Tai Jin Zhi Neng Dian Qi Co., Ltd. industry": "Commercial Electrical Machinery and Equipment Manufacturing",
+    "Lv Shan Zhi Jin Real Estate Development Co., Ltd. industry": "Real Estate",
+    "Whether same industry": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "296c69b8-155e-46d1-991f-f3751bd2f624"
+    },
+    {
+      "company_profile": "d8514a50-5969-4852-9af8-5c1d268d4f31"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy098_result.json b/assets/qa_raw/enterprise_industry_analysis/easy098_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..bd987c3bdf269d065ba941bbdc7105dd892b5a23
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy098_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy098",
+  "question": "Are Jingxin Ruihui Microelectronics Company and Ruixin Yaolan Integrated Circuit Company in the same industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Ruixin Yaolan Integrated Circuit Company (data in this dataset) belongs to the Semiconductor industry",
+    "Jingxin Ruihui Microelectronics Company (data in this dataset) belongs to the Semiconductor industry"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Ruixin Yaolan Integrated Circuit Company is in the Semiconductor industry.",
+    "Extract from company_profile.csv that Jingxin Ruihui Microelectronics Company is in the Semiconductor industry.",
+    "Since the two companies are in the same industry, they are peers; output \"Yes\"."
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Industry of Ruixin Yaolan Integrated Circuit Company": "Semiconductor industry",
+    "Industry of Jingxin Ruihui Microelectronics Company": "Semiconductor industry",
+    "Whether same industry": "Yes"
+  },
+  "reference": [
+    {
+      "company_profile": "e642bedf-c8c5-4350-8ff7-805ecf2d59e4"
+    },
+    {
+      "company_profile": "36bb520d-3ac5-47d6-9a7c-215997de0451"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy099_result.json b/assets/qa_raw/enterprise_industry_analysis/easy099_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..be3c61073bcbb7e0300e31524086d8e5e8888ea5
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy099_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy099",
+  "question": "Will a downturn in the rubber and plastic products industry directly affect the operating conditions of Yao Shi Yuan Ze Sheng Wu Yi Yao Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Yao Shi Yuan Ze Sheng Wu Yi Yao Co., Ltd. (data in this dataset) industry: Pharmaceutical Manufacturing",
+    "Rubber and Plastic Products and Pharmaceutical Manufacturing are different industry categories"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Yao Shi Yuan Ze Sheng Wu Yi Yao Co., Ltd.'s industry is Pharmaceutical Manufacturing",
+    "Conclude that a downturn in the rubber and plastic products industry is not directly equivalent to the operating performance measure of pharmaceutical manufacturing and does not constitute a direct impact",
+    "Output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Yao Shi Yuan Ze Sheng Wu Yi Yao Co., Ltd. industry": "Pharmaceutical Manufacturing",
+    "Comparison industry": "Rubber and Plastic Products",
+    "Whether direct impact": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "b1359959-96f4-4f78-b803-7d92de8c80ba"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy100_result.json b/assets/qa_raw/enterprise_industry_analysis/easy100_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ef09ff7a5c74cd45eac79ec8ea4246a4aaa01ce
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy100_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy100",
+  "question": "Will a downturn in the Information Transmission, Software and IT Services industry directly affect the operating conditions of Zhong Ji Chang Yuan Gang Tie Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhong Ji Chang Yuan Gang Tie Co., Ltd. (data in this dataset) industry: Metal Products",
+    "Information Transmission, Software and IT Services and Metal Products are different industry categories"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong Ji Chang Yuan Gang Tie Co., Ltd.'s industry is Metal Products",
+    "Conclude that Information Transmission, Software and IT Services and Metal Products do not belong to the same industry category",
+    "Therefore a downturn in the former is not directly equivalent to a direct impact on the company's operating conditions; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Zhong Ji Chang Yuan Gang Tie Co., Ltd. industry": "Metal Products",
+    "Comparison industry": "Information Transmission, Software and IT Services",
+    "Whether direct impact": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "3cb448ab-b13e-4078-8442-555fec5b53a3"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy101_result.json b/assets/qa_raw/enterprise_industry_analysis/easy101_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0ef7c608c4d8e788f37582a8fa3058d9a63cebec
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy101_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy101",
+  "question": "Does the non-metallic mineral products industry include Lang Ji Lian Chuang Xin Xi Ji Shu Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Lang Ji Lian Chuang Xin Xi Ji Shu Co., Ltd. (data in this dataset) industry: Information Transmission, Software and IT Services",
+    "Non-metallic Mineral Products and Information Transmission, Software and IT Services are different industry categories"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Lang Ji Lian Chuang Xin Xi Ji Shu Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Compare with the industry named in the question, \"Non-metallic Mineral Products\"; the two are not the same",
+    "Therefore the company does not belong to non-metallic mineral products; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Lang Ji Lian Chuang Xin Xi Ji Shu Co., Ltd. industry": "Information Transmission, Software and IT Services",
+    "Comparison industry": "Non-metallic Mineral Products",
+    "Whether belongs to this industry": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "26c5db34-0cab-4cf9-ad78-0e1d9b6b4410"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy102_result.json b/assets/qa_raw/enterprise_industry_analysis/easy102_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4ad7d36de03d59339ff102b6a5e836f715954b62
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy102_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy102",
+  "question": "Does Lu An Fu Chang Mei Tan Co., Ltd. belong to the Comprehensive Utilization of Waste Resources industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Lu An Fu Chang Mei Tan Co., Ltd. (data in this dataset) industry: Mining",
+    "Comprehensive Utilization of Waste Resources and Mining are different industry categories"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Lu An Fu Chang Mei Tan Co., Ltd.'s industry is Mining",
+    "Compare with the industry named in the question, \"Comprehensive Utilization of Waste Resources\"; the two are not the same",
+    "Therefore the company does not belong to comprehensive utilization of waste resources; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Lu An Fu Chang Mei Tan Co., Ltd. industry": "Mining",
+    "Comparison industry": "Comprehensive Utilization of Waste Resources",
+    "Whether belongs to this industry": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "cb0354e4-911b-4caf-b75e-f60d5a49c9cc"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy103_result.json b/assets/qa_raw/enterprise_industry_analysis/easy103_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1c4641d68eca485fc3d48dadaef65ea02bb18d6
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy103_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy103",
+  "question": "Does Heng Li Yun Chuang Xin Xi Ji Shu Co., Ltd. belong to the Automobile Manufacturing industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Heng Li Yun Chuang Xin Xi Ji Shu Co., Ltd. (data in this dataset) industry: Information Transmission, Software and IT Services",
+    "Automobile Manufacturing and Information Transmission, Software and IT Services are different industry categories"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Heng Li Yun Chuang Xin Xi Ji Shu Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Compare with the industry named in the question, \"Automobile Manufacturing\"; the two are not the same",
+    "Therefore the company does not belong to automobile manufacturing; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Heng Li Yun Chuang Xin Xi Ji Shu Co., Ltd. industry": "Information Transmission, Software and IT Services",
+    "Comparison industry": "Automobile Manufacturing",
+    "Whether belongs to this industry": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "96da6315-0eb7-472f-b876-13530d1c77c6"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy104_result.json b/assets/qa_raw/enterprise_industry_analysis/easy104_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5844998247502981a16123dba2e4a970ad8afbf3
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy104_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy104",
+  "question": "Does Wan Hui Sheng Zhi Construction Development Co., Ltd. belong to the Real Estate industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Wan Hui Sheng Zhi Construction Development Co., Ltd. (data in this dataset) industry: Real Estate",
+    "The industry named in the question is Real Estate"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Wan Hui Sheng Zhi Construction Development Co., Ltd.'s industry is Real Estate",
+    "This matches the industry named in the question, \"Real Estate\"",
+    "Therefore the company belongs to real estate; output \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Wan Hui Sheng Zhi Construction Development Co., Ltd. industry": "Real Estate",
+    "Comparison industry": "Real Estate",
+    "Whether belongs to this industry": "Yes"
+  },
+  "reference": [
+    {
+      "company_profile": "122b543a-480e-41ed-b821-37e3559b2a9a"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy105_result.json b/assets/qa_raw/enterprise_industry_analysis/easy105_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5d6f0b0f8b8515ed84edf0f39b95ba25ddc41756
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy105_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy105",
+  "question": "Does Hang Fa Yuan Jin Hang Kong Technology Co., Ltd. belong to Education?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Hang Fa Yuan Jin Hang Kong Technology Co., Ltd. (data in this dataset) industry: Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "Education and Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing are different industry categories"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Hang Fa Yuan Jin Hang Kong Technology Co., Ltd.'s industry is Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "Compare with the industry named in the question, \"Education\"; the two are not the same",
+    "Therefore the company does not belong to education; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Hang Fa Yuan Jin Hang Kong Technology Co., Ltd. industry": "Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing",
+    "Comparison industry": "Education",
+    "Whether belongs to this industry": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "0faeadd3-eaa1-4e7a-b038-fed69e679c62"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy106_result.json b/assets/qa_raw/enterprise_industry_analysis/easy106_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..05829092b2b114903caa94c3146fcc00f592fa8a
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy106_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy106",
+  "question": "Will a downturn in the Water Conservancy, Environment and Public Facilities Management industry directly affect the operating conditions of Hua Lu Rong Rong Hua Xue Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Hua Lu Rong Rong Hua Xue Co., Ltd. (data in this dataset) industry: Chemical Raw Materials and Chemical Products Manufacturing",
+    "Water Conservancy, Environment and Public Facilities Management and Chemical Raw Materials and Chemical Products Manufacturing are different industry categories"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Hua Lu Rong Rong Hua Xue Co., Ltd.'s industry is Chemical Raw Materials and Chemical Products Manufacturing",
+    "Conclude that Water Conservancy, Environment and Public Facilities Management and Chemical Raw Materials and Chemical Products Manufacturing do not belong to the same industry category",
+    "Therefore a downturn in the former is not directly equivalent to a direct impact on the company's operating conditions; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Hua Lu Rong Rong Hua Xue Co., Ltd. industry": "Chemical Raw Materials and Chemical Products Manufacturing",
+    "Comparison industry": "Water Conservancy, Environment and Public Facilities Management",
+    "Whether direct impact": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "8998d572-1f7e-47a8-9692-45a30d2b9173"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy107_result.json b/assets/qa_raw/enterprise_industry_analysis/easy107_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..df9b67dc45b9a7c217d6f4454ee70efcab723159
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy107_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy107",
+  "question": "Does San San Gong Ji Technology Co., Ltd. belong to the Non-metallic Mineral Products industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "San San Gong Ji Technology Co., Ltd. (data in this dataset) industry: Specialized Equipment Manufacturing",
+    "Non-metallic Mineral Products and Specialized Equipment Manufacturing are different industry categories"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that San San Gong Ji Technology Co., Ltd.'s industry is Specialized Equipment Manufacturing",
+    "Compare with the industry named in the question, \"Non-metallic Mineral Products\"; the two are not the same",
+    "Therefore the company does not belong to non-metallic mineral products; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "San San Gong Ji Technology Co., Ltd. industry": "Specialized Equipment Manufacturing",
+    "Comparison industry": "Non-metallic Mineral Products",
+    "Whether belongs to this industry": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "aaf578b6-882b-4eb5-98bf-fc7717e16d45"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy108_result.json b/assets/qa_raw/enterprise_industry_analysis/easy108_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a4fae0a4d093f6a02ef429740351efc97be4c098
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy108_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy108",
+  "question": "Is Zhong Ju Yue Yin Shi Pin Co., Ltd. registered in Chongqing Municipality?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhong Ju Yue Yin Shi Pin Co., Ltd. (data in this dataset) registration province: Hainan Province",
+    "The province named in the question is Chongqing Municipality"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong Ju Yue Yin Shi Pin Co., Ltd.'s registration province is Hainan Province",
+    "Compare Hainan Province with the province named in the question, \"Chongqing Municipality\"; the two are not the same",
+    "Therefore the company is not registered in Chongqing Municipality; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Zhong Ju Yue Yin Shi Pin Co., Ltd. registration province": "Hainan Province",
+    "Comparison province": "Chongqing Municipality",
+    "Whether registered in Chongqing Municipality": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "3fa7a55f-b108-450a-96b7-0fab5a72735a"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy109_result.json b/assets/qa_raw/enterprise_industry_analysis/easy109_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c3639c5375706c0717d2f7c0e64138b0938a08a7
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy109_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy109",
+  "question": "Does Hua Cheng Jin Jin Zong He Kai Fa Co., Ltd. contribute to the development of Zhejiang Province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Hua Cheng Jin Jin Zong He Kai Fa Co., Ltd. (data in this dataset) registration region: Hong Kong SAR",
+    "The province named in the question is Zhejiang Province"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Hua Cheng Jin Jin Zong He Kai Fa Co., Ltd.'s registration region is Hong Kong SAR",
+    "Compare with the province named in the question, \"Zhejiang Province\"; the two are not the same",
+    "Therefore the condition stated in the question is not met; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Hua Cheng Jin Jin Zong He Kai Fa Co., Ltd. registration region": "Hong Kong SAR",
+    "Comparison province": "Zhejiang Province",
+    "Whether contributes to development of Zhejiang Province": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "f391b306-37a7-41d8-be1a-44dfb51a4d65"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy110_result.json b/assets/qa_raw/enterprise_industry_analysis/easy110_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d9ff741fb1fbd37d854903297a7a05840a561b58
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy110_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy110",
+   "question": "Does Guangdong Province have the enterprise Lv Shan Chan Jin Zhi Ye Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Lv Shan Chan Jin Zhi Ye Co., Ltd. (data in this dataset) registration province: Guangdong Province",
+    "The province named in the question is Guangdong Province"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Lv Shan Chan Jin Zhi Ye Co., Ltd.'s registration province is Guangdong Province",
+    "This matches the province named in the question, \"Guangdong Province\"",
+    "Therefore Guangdong Province has this enterprise; output \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Lv Shan Chan Jin Zhi Ye Co., Ltd. registration province": "Guangdong Province",
+    "Comparison province": "Guangdong Province",
+    "Whether Guangdong Province has this enterprise": "Yes"
+  },
+  "reference": [
+    {
+      "company_profile": "80cf5c52-b78f-4450-a812-d2c6c00a913a"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy111_result.json b/assets/qa_raw/enterprise_industry_analysis/easy111_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4b0470990f40189c0133a27ac543709824af2a7a
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy111_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy111",
+  "question": "Is Hua Xin Ze Chang Xin Cai Liao Co., Ltd. registered in Hebei Province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Hua Xin Ze Chang Xin Cai Liao Co., Ltd. (data in this dataset) registration province: Guangdong Province",
+    "The province named in the question is Hebei Province"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Hua Xin Ze Chang Xin Cai Liao Co., Ltd.'s registration province is Guangdong Province",
+    "Compare Guangdong Province with the province named in the question, \"Hebei Province\"; the two are not the same",
+    "Therefore the company is not registered in Hebei Province; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Hua Xin Ze Chang Xin Cai Liao Co., Ltd. registration province": "Guangdong Province",
+    "Comparison province": "Hebei Province",
+    "Whether registered in Hebei Province": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "7fc42f28-4a93-4620-8dd5-97f6435325f2"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy112_result.json b/assets/qa_raw/enterprise_industry_analysis/easy112_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..77ca1b219de038aeb7c4b6f386aff5c8c42a82fb
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy112_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy112",
+  "question": "Does Wan Hui Jin Sheng Real Estate Development Co., Ltd. contribute to the development of Guangdong Province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Wan Hui Jin Sheng Real Estate Development Co., Ltd. (data in this dataset) registration province: Guangdong Province",
+    "The province named in the question is Guangdong Province"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Wan Hui Jin Sheng Real Estate Development Co., Ltd.'s registration province is Guangdong Province",
+    "This matches the province named in the question, \"Guangdong Province\"",
+    "Therefore it satisfies the condition stated in the question; output \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Wan Hui Jin Sheng Real Estate Development Co., Ltd. registration province": "Guangdong Province",
+    "Comparison province": "Guangdong Province",
+    "Whether contributes to development of Guangdong Province": "Yes"
+  },
+  "reference": [
+    {
+      "company_profile": "3ef2095b-702e-465f-a901-fdc23ba53048"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy113_result.json b/assets/qa_raw/enterprise_industry_analysis/easy113_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..55bfc8230f14b3122164f3a9c9c0ddab996c4ceb
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy113_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy113",
+  "question": "Will changes in Anhui Province's economic environment affect Bao Jin Jin Chang Tong Ye Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Bao Jin Jin Chang Tong Ye Co., Ltd. (data in this dataset) registration province: Jiangsu Province",
+    "The province named in the question is Anhui Province"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Bao Jin Jin Chang Tong Ye Co., Ltd.'s registration province is Jiangsu Province",
+    "Compare Jiangsu Province with the province named in the question, \"Anhui Province\"; the two are not the same",
+    "Therefore changes in Anhui Province's economic environment are not directly relevant to this enterprise; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Bao Jin Jin Chang Tong Ye Co., Ltd. registration province": "Jiangsu Province",
+    "Comparison province": "Anhui Province",
+    "Whether affected by Anhui Province economic environment changes": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "d23c45ba-53ee-4ccf-b1e2-755feb2113c3"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy114_result.json b/assets/qa_raw/enterprise_industry_analysis/easy114_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..96037fb8d54e50528a1f2a2f385e6cbbfe3a4b61
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy114_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy114",
+  "question": "Will changes in Guizhou Province's economic environment affect Zhong You Zheng Da Jin Yun Shu Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhong You Zheng Da Jin Yun Shu Co., Ltd. (data in this dataset) registration province: Shanghai Municipality",
+    "The province named in the question is Guizhou Province"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong You Zheng Da Jin Yun Shu Co., Ltd.'s registration province is Shanghai Municipality",
+    "Compare Shanghai Municipality with the province named in the question, \"Guizhou Province\"; the two are not the same",
+    "Therefore changes in Guizhou Province's economic environment are not directly relevant to this enterprise; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Zhong You Zheng Da Jin Yun Shu Co., Ltd. registration province": "Shanghai Municipality",
+    "Comparison province": "Guizhou Province",
+    "Whether affected by Guizhou Province economic environment changes": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "c0d7672e-a099-4780-874d-29e8425af1c7"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy115_result.json b/assets/qa_raw/enterprise_industry_analysis/easy115_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..1345e8cd200388b3a7b744ec06d155d152811523
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy115_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy115",
+  "question": "Does Huan Xing Jin Ya Apparel Co., Ltd. contribute to the development of Jiangxi Province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Huan Xing Jin Ya Apparel Co., Ltd. (dataset value): registered province = Jiangsu Province",
+    "Province specified in the question: Jiangxi Province"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: registered province of Huan Xing Jin Ya Apparel Co., Ltd. = Jiangsu Province",
+    "Compared Jiangsu Province with the question's province, Jiangxi Province; they do not match",
+    "Therefore, the condition in the question is not satisfied; output \"No\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Registered province of Huan Xing Jin Ya Apparel Co., Ltd.": "Jiangsu Province",
+    "Compared province": "Jiangxi Province",
+    "Whether it contributes to Jiangxi Province's development": "No"
+  },
+  "reference": [
+    {
+      "company_profile": "de5c2a0f-d617-40b5-8fe1-8626a58c1101"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy116_result.json b/assets/qa_raw/enterprise_industry_analysis/easy116_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e1f716de24a1295bd248b0559fb8ab01365b654b
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy116_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy116",
+  "question": "Would changes in Guangdong Province's economic environment affect Hua Dian Neng Jin Hydropower Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Hua Dian Neng Jin Hydropower Co., Ltd. (dataset value): registered province = Guangdong Province",
+    "Province specified in the question: Guangdong Province"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: registered province of Hua Dian Neng Jin Hydropower Co., Ltd. = Guangdong Province",
+    "Compared Guangdong Province with the question's province, Guangdong Province; they match",
+    "Therefore, provincial economic changes described in the question can affect the company's operating environment; output \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Registered province of Hua Dian Neng Jin Hydropower Co., Ltd.": "Guangdong Province",
+    "Compared province": "Guangdong Province",
+    "Whether affected by economic changes in Guangdong Province": "Yes"
+  },
+  "reference": [
+    {
+      "company_profile": "34cc539e-12b3-49ae-aecf-bc295add8fba"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/easy117_result.json b/assets/qa_raw/enterprise_industry_analysis/easy117_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4841301fb3fe8905076240941e2063200d6c26fe
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/easy117_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "easy117",
+  "question": "Is Zhongke Zhiyun Data Services Co., Ltd. registered in Guangdong Province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhongke Zhiyun Data Services Co., Ltd. (dataset value): registered province = Guangdong Province",
+    "Province specified in the question: Guangdong Province"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: registered province of Zhongke Zhiyun Data Services Co., Ltd. = Guangdong Province",
+    "Compared Guangdong Province with the question's province, Guangdong Province; they match",
+    "Therefore, the company is registered in Guangdong Province; output \"Yes\""
+  ],
+  "steps_num": 3,
+  "milestone": {
+    "Registered province of Zhongke Zhiyun Data Services Co., Ltd.": "Guangdong Province",
+    "Compared province": "Guangdong Province",
+    "Whether registered in Guangdong Province": "Yes"
+  },
+  "reference": [
+    {
+      "company_profile": "d6ab2642-2691-4f63-a197-6a4c29026c0d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium001_result.json b/assets/qa_raw/enterprise_industry_analysis/medium001_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c45deed67d52ee8a489dc924c15ec2956d535c3
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium001_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium001",
+  "question": "In 2022, what is the difference between the year-over-year employee change rate of Kangsheng Kangjian Pharmaceutical Co., Ltd. and the minimum level of its industry?",
+  "guidelines": "The answer must be a number with two decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Kangsheng Kangjian Pharmaceutical Co., Ltd. year-over-year employee change rate: 28.63 %",
+    "Industry of Kangsheng Kangjian Pharmaceutical Co., Ltd.: Pharmaceutical Manufacturing",
+    "Minimum year-over-year employee change rate in Pharmaceutical Manufacturing: -65.72 %"
+  ],
+  "milestone": {
+    "Kangsheng Kangjian Pharmaceutical Co., Ltd. year-over-year employee change rate": "28.63 %",
+    "Industry of Kangsheng Kangjian Pharmaceutical Co., Ltd.": "Pharmaceutical Manufacturing",
+    "Minimum year-over-year employee change rate in Pharmaceutical Manufacturing": "-65.72 %",
+    "Difference between the two": 94.35
+  },
+  "answer": 94.35,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "reference": [
+    {
+      "company_operation_status": "aca07f14-a552-49a7-8fd9-cf23d30d5f81"
+    },
+    {
+      "national_industry_status": "9c71278e-97a3-4867-826d-e139f1dffc24"
+    },
+    {
+      "company_profile": "e655979a-a163-45f2-806f-97922582ee4d"
+    }
+  ],
+  "steps": [
+    "Extracted from company_operation_status.csv: Kangsheng Kangjian Pharmaceutical Co., Ltd. 2022 year-over-year employee change rate is 28.63 %",
+    "Extracted from company_profile.csv: the company's industry is Pharmaceutical Manufacturing",
+    "Extracted from national_industry_status.csv: for Pharmaceutical Manufacturing, the minimum year-over-year employee change rate is -65.72 %",
+    "Computed difference: 28.63 - (-65.72) = 94.35"
+  ],
+  "steps_num": 4
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium002_result.json b/assets/qa_raw/enterprise_industry_analysis/medium002_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..88b3b977123db2116e95339f561a6a408cacd983
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium002_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium002",
+  "question": "In 2022, what is the difference between the total number of employees of Kangsheng Kangjian Pharmaceutical Co., Ltd. and the industry average?",
+  "guidelines": "The answer must be an exact number and keep all significant decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Kangsheng Kangjian Pharmaceutical Co., Ltd. total employees in 2022: 2099.0",
+    "Industry of Kangsheng Kangjian Pharmaceutical Co., Ltd.: Pharmaceutical Manufacturing",
+    "Average total employees in Pharmaceutical Manufacturing: 3057.60986547085"
+  ],
+  "milestone": {
+    "Kangsheng Kangjian Pharmaceutical Co., Ltd. total employees in 2022": 2099.0,
+    "Industry of Kangsheng Kangjian Pharmaceutical Co., Ltd.": "Pharmaceutical Manufacturing",
+    "Average total employees in Pharmaceutical Manufacturing": 3057.60986547085,
+    "Difference (company - industry average)": -958.60986547085
+  },
+  "answer": -958.60986547085,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "reference": [
+    {
+      "company_operation_status": "aca07f14-a552-49a7-8fd9-cf23d30d5f81"
+    },
+    {
+      "national_industry_status": "9c71278e-97a3-4867-826d-e139f1dffc24"
+    },
+    {
+      "company_profile": "e655979a-a163-45f2-806f-97922582ee4d"
+    }
+  ],
+  "steps": [
+    "Extracted from company_operation_status.csv: Kangsheng Kangjian Pharmaceutical Co., Ltd. total employees in 2022 is 2099.0",
+    "Extracted from company_profile.csv: the company's industry is Pharmaceutical Manufacturing",
+    "Extracted from national_industry_status.csv: average total employees in Pharmaceutical Manufacturing is 3057.60986547085",
+    "Computed difference (company - industry average): 2099.0 - 3057.60986547085 = -958.60986547085"
+  ],
+  "steps_num": 4
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium003_result.json b/assets/qa_raw/enterprise_industry_analysis/medium003_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..953ceaf1879452f1e1f74c9db4965ba42e98d9f3
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium003_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium003",
+  "question": "In 2022, what is the difference between the total number of employees of Ling You Se Ye Da Zi Yuan Co., Ltd. and the industry maximum?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Ling You Se Ye Da Zi Yuan Co., Ltd. (2022 dataset value): total employees = 2367.0",
+    "The company belongs to Metal Smelting and Rolling Processing Industry",
+    "Maximum total employees in Metal Smelting and Rolling Processing Industry = 67377"
+  ],
+  "milestone": {
+    "Ling You Se Ye Da Zi Yuan Co., Ltd. total employees in 2022": 2367.0,
+    "Industry of Ling You Se Ye Da Zi Yuan Co., Ltd.": "Metal Smelting and Rolling Processing Industry",
+    "Maximum total employees in Metal Smelting and Rolling Processing Industry": 67377,
+    "Difference (company - industry maximum)": -65010.0
+  },
+  "answer": -65010.0,
+  "steps": [
+    "Extracted from company_operation_status.csv: Ling You Se Ye Da Zi Yuan Co., Ltd. total employees in 2022 = 2367.0",
+    "Extracted from company_profile.csv: the company belongs to Metal Smelting and Rolling Processing Industry",
+    "Extracted from national_industry_status.csv: industry maximum total employees = 67377",
+    "Calculated by requirement: difference (company - industry maximum) = 2367.0 - 67377 = -65010.0, output with one decimal place"
+  ],
+  "steps_num": 4,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "reference": [
+    {
+      "company_operation_status": "81e70393-8c21-444e-b99c-35cd45ea16b3"
+    },
+    {
+      "national_industry_status": "e12e3329-0ab1-4bf1-b19d-0a5c9610b5ce"
+    },
+    {
+      "company_profile": "be102125-d7e2-475a-8217-540be0ab9101"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium004_result.json b/assets/qa_raw/enterprise_industry_analysis/medium004_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..77695063eb1a3f6dde92e043dabe8ca91a3188a4
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium004_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium004",
+  "question": "In 2022, what is the difference between the year-over-year net profit change rate of Ling You Se Ye Da Zi Yuan Co., Ltd. and the minimum value of this indicator in its industry?",
+  "guidelines": "The answer must be a number with two decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Ling You Se Ye Da Zi Yuan Co., Ltd. (2022 dataset value): year-over-year net profit change rate = -38.88 %",
+    "The company belongs to Metal Smelting and Rolling Processing Industry",
+    "Minimum year-over-year net profit change rate in Metal Smelting and Rolling Processing Industry = -2149.3 %"
+  ],
+  "milestone": {
+    "Ling You Se Ye Da Zi Yuan Co., Ltd. year-over-year net profit change rate in 2022": "-38.88 %",
+    "Industry of Ling You Se Ye Da Zi Yuan Co., Ltd.": "Metal Smelting and Rolling Processing Industry",
+    "Minimum year-over-year net profit change rate in Metal Smelting and Rolling Processing Industry": "-2149.3 %",
+    "Difference between the two": 2110.42
+  },
+  "answer": 2110.42,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Ling You Se Ye Da Zi Yuan Co., Ltd. year-over-year net profit change rate in 2022 = -38.88 %",
+    "Extracted from company_profile.csv: the company belongs to Metal Smelting and Rolling Processing Industry",
+    "Extracted from national_industry_status.csv: minimum year-over-year net profit change rate in the industry = -2149.3 %",
+    "Calculated difference: -38.88 - (-2149.3) = 2110.42"
+  ],
+  "steps_num": 4,
+  "reference": [
+    {
+      "company_operation_status": "81e70393-8c21-444e-b99c-35cd45ea16b3"
+    },
+    {
+      "national_industry_status": "e12e3329-0ab1-4bf1-b19d-0a5c9610b5ce"
+    },
+    {
+      "company_profile": "be102125-d7e2-475a-8217-540be0ab9101"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium005_result.json b/assets/qa_raw/enterprise_industry_analysis/medium005_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2bddabe46f7fe1ee721dfa0c4afd4df756b742cf
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium005_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium005",
+  "question": "In 2022, which is higher: the year-over-year employee change rate of Yong Feng Xin Chuang Ke Ji Co., Ltd. or the maximum value of this indicator in its industry?",
+  "guidelines": "The answer must be either \"industry\" or the company name. Output only one word or the company name, without any explanation, analysis, or descriptive text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Yong Feng Xin Chuang Ke Ji Co., Ltd. (2022 dataset value): year-over-year employee change rate = 12.61 %",
+    "The company belongs to Information Transmission, Software and Information Technology Services",
+    "Maximum year-over-year employee change rate in Information Transmission, Software and Information Technology Services = 416.95 %"
+  ],
+  "milestone": {
+    "Yong Feng Xin Chuang Ke Ji Co., Ltd. year-over-year employee change rate in 2022": "12.61 %",
+    "Industry of Yong Feng Xin Chuang Ke Ji Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Maximum year-over-year employee change rate in Information Transmission, Software and Information Technology Services": "416.95 %",
+    "Comparison result (whether the industry is higher)": "Yes"
+  },
+  "answer": "industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Yong Feng Xin Chuang Ke Ji Co., Ltd. year-over-year employee change rate in 2022 = 12.61 %",
+    "Extracted from company_profile.csv: the company belongs to Information Transmission, Software and Information Technology Services",
+    "Extracted from national_industry_status.csv: industry maximum year-over-year employee change rate = 416.95 %",
+    "Compared 12.61 and 416.95; since 416.95 > 12.61, the industry is higher, so output \"industry\""
+  ],
+  "steps_num": 4,
+  "reference": [
+    {
+      "company_operation_status": "978a253c-f9e9-447b-8c57-ad0c50c7f58c"
+    },
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "2c82af76-a0ee-4655-baf5-b1206c9ac103"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium006_result.json b/assets/qa_raw/enterprise_industry_analysis/medium006_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..23a2105f25dbd13ff19ec1332623d3b7f60a6e4e
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium006_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium006",
+  "question": "In 2022, is the year-over-year net profit change rate of Yong Feng Xin Chuang Ke Ji Co., Ltd. higher than the median of this indicator in its industry?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Yong Feng Xin Chuang Ke Ji Co., Ltd. (2022 dataset value): year-over-year net profit change rate = 40.82 %",
+    "The company belongs to Information Transmission, Software and Information Technology Services",
+    "Median year-over-year net profit change rate in Information Transmission, Software and Information Technology Services = -15.96 %"
+  ],
+  "milestone": {
+    "Yong Feng Xin Chuang Ke Ji Co., Ltd. year-over-year net profit change rate in 2022": "40.82 %",
+    "Industry of Yong Feng Xin Chuang Ke Ji Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Median year-over-year net profit change rate in Information Transmission, Software and Information Technology Services": "-15.96 %",
+    "Comparison result (whether the company is higher than the industry median)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Yong Feng Xin Chuang Ke Ji Co., Ltd. year-over-year net profit change rate in 2022 = 40.82 %",
+    "Extracted from company_profile.csv: the company belongs to Information Transmission, Software and Information Technology Services",
+    "Extracted from national_industry_status.csv: industry median year-over-year net profit change rate = -15.96 %",
+    "Compared 40.82 and -15.96; since 40.82 > -15.96, the judgment is \"Yes\""
+  ],
+  "steps_num": 4,
+  "reference": [
+    {
+      "company_operation_status": "978a253c-f9e9-447b-8c57-ad0c50c7f58c"
+    },
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "2c82af76-a0ee-4655-baf5-b1206c9ac103"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium007_result.json b/assets/qa_raw/enterprise_industry_analysis/medium007_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..25148a8e70eefdfee81d29d3c358c732478a76ed
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium007_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "medium007",
+  "question": "In 2022, what is the difference between the operating profit amount of Lian Ji Chuang Ji Ji Chuang Co., Ltd. and the total operating profit amount of the same industry in its province?",
+  "guidelines": "The answer must be a number with two decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Lian Ji Chuang Ji Ji Chuang Co., Ltd. (2022 dataset value): operating profit amount = -41697747.51 CNY",
+    "The company is located in Anhui Province and belongs to Water Conservancy, Environment and Public Facilities Management",
+    "Total operating profit amount in Anhui Province for Water Conservancy, Environment and Public Facilities Management = 1697060777.41 CNY"
+  ],
+  "milestone": {
+    "Lian Ji Chuang Ji Ji Chuang Co., Ltd. operating profit amount in 2022 (CNY)": -41697747.51,
+    "Province of Lian Ji Chuang Ji Ji Chuang Co., Ltd.": "Anhui Province",
+    "Industry of Lian Ji Chuang Ji Ji Chuang Co., Ltd.": "Water Conservancy, Environment and Public Facilities Management",
+    "Total operating profit amount in Anhui Province for Water Conservancy, Environment and Public Facilities Management (CNY)": 1697060777.41,
+    "Difference (company - same-province same-industry total)": -1738758524.92
+  },
+  "answer": -1738758524.92,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Lian Ji Chuang Ji Ji Chuang Co., Ltd. operating profit amount in 2022 = -41697747.51 CNY",
+    "Extracted from company_profile.csv: the company is in Anhui Province and belongs to Water Conservancy, Environment and Public Facilities Management",
+    "Extracted from regional_industry_status.csv: total operating profit amount in Anhui Province for this industry = 1697060777.41 CNY",
+    "Calculated difference (company - same-province same-industry total): -41697747.51 - 1697060777.41 = -1738758524.92"
+  ],
+  "steps_num": 4,
+  "reference": [
+    {
+      "company_operation_status": "a81bbdaf-4973-42a6-92c1-d9aeced396da"
+    },
+    {
+      "regional_industry_status": "9827d220-2c11-4ff3-ac01-b163706f67b3"
+    },
+    {
+      "company_profile": "ba97b51b-d1a5-4e35-b2c9-d975b493d133"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium008_result.json b/assets/qa_raw/enterprise_industry_analysis/medium008_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..744ffa26e45a8c5be0a1b5bf6c6df6f93d92b7e2
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium008_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "medium008",
+  "question": "In 2022, compared with the average level of the same industry in its province, which is higher: the number of R&D personnel of Lianji Chuangji Machine Tool Company or the industry average?",
+  "guidelines": "The answer must be either the company name or the word \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Lianji Chuangji Machine Tool Company (2022 data in this dataset) has 290.0 R&D personnel",
+    "Lianji Chuangji Machine Tool Company is located in Anhui Province and belongs to the Water Conservancy, Environment and Public Facilities Management industry",
+    "The average number of R&D personnel in Anhui Province for the Water Conservancy, Environment and Public Facilities Management industry is 132.111111111111"
+  ],
+  "milestone": {
+    "Number of R&D personnel of Lianji Chuangji Machine Tool Company in 2022": 290.0,
+    "Province of Lianji Chuangji Machine Tool Company": "Anhui Province",
+    "Industry of Lianji Chuangji Machine Tool Company": "Water Conservancy, Environment and Public Facilities Management",
+    "Average number of R&D personnel in Anhui Province for the Water Conservancy, Environment and Public Facilities Management industry": 132.111111111111,
+    "Comparison result (whether the company is higher)": "Yes"
+  },
+  "answer": "Lianji Chuangji Machine Tool Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extract from company_operation_status.csv that the number of R&D personnel of Lianji Chuangji Machine Tool Company in 2022 is 290.0.",
+    "Extract from company_profile.csv that the company is located in Anhui Province and belongs to the Water Conservancy, Environment and Public Facilities Management industry.",
+    "Extract from regional_industry_status.csv that the average number of R&D personnel in Anhui Province for this industry is 132.111111111111.",
+    "Compare 290.0 with 132.111111111111; since 290.0 > 132.111111111111, the company is higher, so output \"Lianji Chuangji Machine Tool Company\"."
+  ],
+  "steps_num": 4,
+  "reference": [
+    {
+      "company_operation_status": "a81bbdaf-4973-42a6-92c1-d9aeced396da"
+    },
+    {
+      "regional_industry_status": "9827d220-2c11-4ff3-ac01-b163706f67b3"
+    },
+    {
+      "company_profile": "ba97b51b-d1a5-4e35-b2c9-d975b493d133"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium009_result.json b/assets/qa_raw/enterprise_industry_analysis/medium009_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c029b006c41c5f72dcfa39b6cc1372e4d10fb44c
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium009_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "medium009",
+  "question": "In 2022, which is higher: the operating profit amount of Run Hui Shu Zhi Xi Tong Co., Ltd. or the total operating profit amount of the same industry in its province?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Run Hui Shu Zhi Xi Tong Co., Ltd. (2022 dataset value): operating profit amount = 9041715.19 CNY",
+    "The company is located in Jilin Province and belongs to Scientific Research and Technical Services",
+    "Total operating profit amount in Jilin Province for Scientific Research and Technical Services = 0 CNY"
+  ],
+  "milestone": {
+    "Run Hui Shu Zhi Xi Tong Co., Ltd. operating profit amount in 2022 (CNY)": 9041715.19,
+    "Province of Run Hui Shu Zhi Xi Tong Co., Ltd.": "Jilin Province",
+    "Industry of Run Hui Shu Zhi Xi Tong Co., Ltd.": "Scientific Research and Technical Services",
+    "Total operating profit amount in Jilin Province for Scientific Research and Technical Services (CNY)": 0,
+    "Comparison result (whether the company is higher)": "Yes"
+  },
+  "answer": "Run Hui Shu Zhi Xi Tong Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Run Hui Shu Zhi Xi Tong Co., Ltd. operating profit amount in 2022 = 9041715.19 CNY",
+    "Extracted from company_profile.csv: the company is in Jilin Province and belongs to Scientific Research and Technical Services",
+    "Extracted from regional_industry_status.csv: total operating profit amount in Jilin Province for this industry = 0 CNY",
+    "Compared 9041715.19 and 0; since 9041715.19 > 0, the company is higher, so output \"Run Hui Shu Zhi Xi Tong Co., Ltd.\""
+  ],
+  "steps_num": 4,
+  "reference": [
+    {
+      "company_operation_status": "ec61653b-88b8-4a5e-8e7d-ed9fda3dcfba"
+    },
+    {
+      "regional_industry_status": "a93c5a67-46ad-438a-be74-d49e8a6cbc1b"
+    },
+    {
+      "company_profile": "1e954ae5-70b5-48e4-90ee-a640834447cb"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium010_result.json b/assets/qa_raw/enterprise_industry_analysis/medium010_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..404c502c57bb313186302e6f6008be53c84b5f7f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium010_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "medium010",
+  "question": "In 2022, is the operating revenue amount of Run Hui Shu Zhi Xi Tong Co., Ltd. higher than the total operating revenue amount of the corresponding industry in its province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Run Hui Shu Zhi Xi Tong Co., Ltd. (2022 dataset value): operating revenue amount = 157705748.68 CNY",
+    "The company is located in Jilin Province and belongs to Scientific Research and Technical Services",
+    "Total operating revenue amount in Jilin Province for Scientific Research and Technical Services = 0 CNY"
+  ],
+  "milestone": {
+    "Run Hui Shu Zhi Xi Tong Co., Ltd. operating revenue amount in 2022 (CNY)": 157705748.68,
+    "Province of Run Hui Shu Zhi Xi Tong Co., Ltd.": "Jilin Province",
+    "Industry of Run Hui Shu Zhi Xi Tong Co., Ltd.": "Scientific Research and Technical Services",
+    "Total operating revenue amount in Jilin Province for Scientific Research and Technical Services (CNY)": 0,
+    "Comparison result (whether the company is higher than the provincial same-industry total)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Run Hui Shu Zhi Xi Tong Co., Ltd. operating revenue amount in 2022 = 157705748.68 CNY",
+    "Extracted from company_profile.csv: the company is in Jilin Province and belongs to Scientific Research and Technical Services",
+    "Extracted from regional_industry_status.csv: total operating revenue amount in Jilin Province for this industry = 0 CNY",
+    "Compared 157705748.68 and 0; since 157705748.68 > 0, the judgment is \"Yes\""
+  ],
+  "steps_num": 4,
+  "reference": [
+    {
+      "company_operation_status": "ec61653b-88b8-4a5e-8e7d-ed9fda3dcfba"
+    },
+    {
+      "regional_industry_status": "a93c5a67-46ad-438a-be74-d49e8a6cbc1b"
+    },
+    {
+      "company_profile": "1e954ae5-70b5-48e4-90ee-a640834447cb"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium011_result.json b/assets/qa_raw/enterprise_industry_analysis/medium011_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e0a99add111ad03440a3134173b24ba4372c428a
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium011_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "medium011",
+  "question": "In 2022, which is higher: the cumulative number of PCT invention patent applications of Zhong Ji Da Chang Tong Ye Co., Ltd. or the minimum value of the same indicator in the same industry in its province?",
+  "guidelines": "The answer must be \"equal\", the company name, or \"industry\". Output only one word or the company name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhong Ji Da Chang Tong Ye Co., Ltd. (2022 dataset value): cumulative PCT invention patent applications = 1",
+    "The company is located in Jiangsu Province and belongs to Agriculture, Forestry, Animal Husbandry and Fishery",
+    "Minimum cumulative PCT invention patent applications in Jiangsu Province for Agriculture, Forestry, Animal Husbandry and Fishery = 1"
+  ],
+  "milestone": {
+    "Zhong Ji Da Chang Tong Ye Co., Ltd. cumulative PCT invention patent applications in 2022": 1,
+    "Province of Zhong Ji Da Chang Tong Ye Co., Ltd.": "Jiangsu Province",
+    "Industry of Zhong Ji Da Chang Tong Ye Co., Ltd.": "Agriculture, Forestry, Animal Husbandry and Fishery",
+    "Minimum cumulative PCT invention patent applications in Jiangsu Province for Agriculture, Forestry, Animal Husbandry and Fishery": 1,
+    "Comparison result": "Equal"
+  },
+  "answer": "equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Zhong Ji Da Chang Tong Ye Co., Ltd. cumulative PCT invention patent applications in 2022 = 1",
+    "Extracted from company_profile.csv: the company is in Jiangsu Province and belongs to Agriculture, Forestry, Animal Husbandry and Fishery",
+    "Extracted from regional_industry_status.csv: minimum cumulative PCT invention patent applications in the provincial same industry = 1",
+    "Compared 1 and 1; they are equal, so output \"equal\""
+  ],
+  "steps_num": 4,
+  "reference": [
+    {
+      "company_operation_status": "1a43f9c2-bb1c-4f31-91c1-7fd599da75e2"
+    },
+    {
+      "regional_industry_status": "183e9fca-3979-4971-8a7f-a97cf04f2cea"
+    },
+    {
+      "company_profile": "b6494485-93ab-4705-ab72-0aa551d2f1d6"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium012_result.json b/assets/qa_raw/enterprise_industry_analysis/medium012_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..307f375db561d8b862ead9d71d8f8fab3c1fb18d
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium012_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "medium012",
+  "question": "In 2022, is the debt-to-asset ratio of Zhong Ji Da Chang Tong Ye Co., Ltd. higher than the minimum debt-to-asset ratio of the corresponding industry in its province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Zhong Ji Da Chang Tong Ye Co., Ltd. (2022 dataset value): debt-to-asset ratio = 35.4 %",
+    "The company is located in Jiangsu Province and belongs to Agriculture, Forestry, Animal Husbandry and Fishery",
+    "Minimum debt-to-asset ratio in Jiangsu Province for Agriculture, Forestry, Animal Husbandry and Fishery = 23.15 %"
+  ],
+  "milestone": {
+    "Zhong Ji Da Chang Tong Ye Co., Ltd. debt-to-asset ratio in 2022": "35.4 %",
+    "Province of Zhong Ji Da Chang Tong Ye Co., Ltd.": "Jiangsu Province",
+    "Industry of Zhong Ji Da Chang Tong Ye Co., Ltd.": "Agriculture, Forestry, Animal Husbandry and Fishery",
+    "Minimum debt-to-asset ratio in Jiangsu Province for Agriculture, Forestry, Animal Husbandry and Fishery": "23.15 %",
+    "Comparison result (whether the company is higher than the provincial same-industry minimum)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: Zhong Ji Da Chang Tong Ye Co., Ltd. debt-to-asset ratio in 2022 = 35.4 %",
+    "Extracted from company_profile.csv: the company is in Jiangsu Province and belongs to Agriculture, Forestry, Animal Husbandry and Fishery",
+    "Extracted from regional_industry_status.csv: minimum debt-to-asset ratio in the provincial same industry = 23.15 %",
+    "Compared 35.4 and 23.15; since 35.4 > 23.15, the judgment is \"Yes\""
+  ],
+  "steps_num": 4,
+  "reference": [
+    {
+      "company_operation_status": "1a43f9c2-bb1c-4f31-91c1-7fd599da75e2"
+    },
+    {
+      "regional_industry_status": "183e9fca-3979-4971-8a7f-a97cf04f2cea"
+    },
+    {
+      "company_profile": "b6494485-93ab-4705-ab72-0aa551d2f1d6"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium013_result.json b/assets/qa_raw/enterprise_industry_analysis/medium013_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..04b9f0747320cc66530f43742a5237862fbb99ca
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium013_result.json
@@ -0,0 +1,42 @@
+{
+  "id": "medium013",
+  "question": "In 2022, is the market capitalization of Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. lower than the operating revenue of Long He Zhi Jin Zhi Ye Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. (2022 dataset value): market capitalization = 310.0 hundred million CNY",
+    "Long He Zhi Jin Zhi Ye Co., Ltd. (2022 dataset value): operating revenue = 1702394443.0 CNY"
+  ],
+  "milestone": {
+    "Market capitalization of Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. in 2022 (hundred million CNY)": 310.0,
+    "Converted market capitalization of Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. in 2022 (CNY)": 31000000000,
+    "Operating revenue of Long He Zhi Jin Zhi Ye Co., Ltd. in 2022 (CNY)": 1702394443.0,
+    "Comparison result (whether the market capitalization of Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. is lower than the operating revenue of Long He Zhi Jin Zhi Ye Co., Ltd.)": "No"
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_operation_status.csv: market capitalization of Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. in 2022 = 310.0 hundred million CNY",
+    "Converted 310.0 hundred million CNY to CNY: 310.0 * 100000000 = 31000000000 CNY",
+    "Extracted from company_operation_status.csv: operating revenue of Long He Zhi Jin Zhi Ye Co., Ltd. in 2022 = 1702394443.0 CNY",
+    "Compared 31000000000 and 1702394443.0; since 31000000000 > 1702394443.0, the judgment is \"No\""
+  ],
+  "steps_num": 4,
+  "reference": [
+    {
+      "company_operation_status": "42462fed-6dbc-400e-b9da-c46efc7fcf28"
+    },
+    {
+      "company_operation_status": "e42fabec-4192-4d35-8382-c3f2c2575de0"
+    },
+    {
+      "company_profile": "41fc40f8-35ba-4a45-9c72-56aa081d8caa"
+    },
+    {
+      "company_profile": "26b455a0-8084-4ea6-a2d9-b08d9305f26b"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium014_result.json b/assets/qa_raw/enterprise_industry_analysis/medium014_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5bd3f0b89a06430fd540565998863b17822e835c
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium014_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium014",
+  "question": "Comparing the number of SSE-listed state-owned enterprise institutes in the industry of Huijin Jinrui Wealth Management Co., Ltd. with the number of HKEX-listed sino-foreign joint ventures in the industry of Zhongke Zhiyun Data Services Co., Ltd., which value is larger?",
+  "guidelines": "The answer must be a company name, \"industry\", or \"Same\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Huijin Jinrui Wealth Management Co., Ltd.: Financial Industry",
+    "Financial Industry, number of state-owned enterprise institutes listed on SSE: 1",
+    "Industry of Zhongke Zhiyun Data Services Co., Ltd.: Information Transmission, Software and Information Technology Services",
+    "Information Transmission, Software and Information Technology Services, number of HKEX-listed sino-foreign joint ventures: 1"
+  ],
+  "milestone": {
+    "Industry of Huijin Jinrui Wealth Management Co., Ltd.": "Financial Industry",
+    "Number of SSE-listed state-owned enterprise institutes in Financial Industry": 1,
+    "Industry of Zhongke Zhiyun Data Services Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Number of HKEX-listed sino-foreign joint ventures in Information Transmission, Software and Information Technology Services": 1,
+    "Comparison result (answer output)": "Same"
+  },
+  "answer": "Same",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Huijin Jinrui Wealth Management Co., Ltd. is Financial Industry.",
+    "Extracted from national_industry_status.csv that the number of SSE-listed state-owned enterprise institutes in Financial Industry is 1.",
+    "Extracted from company_profile.csv that the industry of Zhongke Zhiyun Data Services Co., Ltd. is Information Transmission, Software and Information Technology Services.",
+    "Extracted from national_industry_status.csv that the number of HKEX-listed sino-foreign joint ventures in Information Transmission, Software and Information Technology Services is 1.",
+    "Compared the two counts and output the result corresponding to the predefined answer field: Same."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "5dc0d9db-af52-48e2-8eef-98a66f67632f"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    },
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "d6ab2642-2691-4f63-a197-6a4c29026c0d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium015_result.json b/assets/qa_raw/enterprise_industry_analysis/medium015_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d8b1b86a3bf8dc39dc6333bdd3df7cd39ab6645
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium015_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium015",
+  "question": "Is the minimum R&D personnel ratio in the industry of Huijin Jinrui Wealth Management Co., Ltd. lower than the minimum R&D personnel ratio in the industry of Zhongke Zhiyun Data Services Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Huijin Jinrui Wealth Management Co., Ltd.: Financial Industry",
+    "Minimum R&D personnel ratio in Financial Industry: 0.64 %",
+    "Industry of Zhongke Zhiyun Data Services Co., Ltd.: Information Transmission, Software and Information Technology Services",
+    "Minimum R&D personnel ratio in Information Transmission, Software and Information Technology Services: 0.92 %"
+  ],
+  "milestone": {
+    "Industry of Huijin Jinrui Wealth Management Co., Ltd.": "Financial Industry",
+    "Minimum R&D personnel ratio in Financial Industry": "0.64 %",
+    "Industry of Zhongke Zhiyun Data Services Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Minimum R&D personnel ratio in Information Transmission, Software and Information Technology Services": "0.92 %",
+    "Comparison result (whether the Financial Industry minimum is lower than that of Information Services)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Huijin Jinrui Wealth Management Co., Ltd. is Financial Industry.",
+    "Extracted from national_industry_status.csv that the minimum R&D personnel ratio in Financial Industry is 0.64 %.",
+    "Extracted from company_profile.csv that the industry of Zhongke Zhiyun Data Services Co., Ltd. is Information Transmission, Software and Information Technology Services.",
+    "Extracted from national_industry_status.csv that the minimum R&D personnel ratio in Information Transmission, Software and Information Technology Services is 0.92 %.",
+    "Judged whether 0.64 % is lower than 0.92 %; the conclusion is \"Yes\"."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "5dc0d9db-af52-48e2-8eef-98a66f67632f"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    },
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "d6ab2642-2691-4f63-a197-6a4c29026c0d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium016_result.json b/assets/qa_raw/enterprise_industry_analysis/medium016_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..6dc4ca40b81d398c764e61338416162a0bc020a0
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium016_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium016",
+  "question": "What is the difference between the median annual number of Chinese invention patent applications in the industry of Changqiao Jinchuang Technology Co., Ltd. and the corresponding metric for the industry in Zhejiang Province where Wuli Huida Chain Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Changqiao Jinchuang Technology Co., Ltd.: Consumer Electronics and Electrical Industry",
+    "Median annual Chinese invention patent applications in Consumer Electronics and Electrical Industry: 18",
+    "Province of Wuli Huida Chain Co., Ltd.: Zhejiang Province",
+    "Median annual Chinese invention patent applications in Zhejiang Wholesale and Retail Trade: 2"
+  ],
+  "milestone": {
+    "Industry of Changqiao Jinchuang Technology Co., Ltd.": "Consumer Electronics and Electrical Industry",
+    "Median annual Chinese invention patent applications in Consumer Electronics and Electrical Industry": 18,
+    "Province of Wuli Huida Chain Co., Ltd.": "Zhejiang Province",
+    "Median annual Chinese invention patent applications in Zhejiang Wholesale and Retail Trade": 2,
+    "Difference (Consumer Electronics and Electrical Industry - Zhejiang Wholesale and Retail Trade)": 16.0
+  },
+  "answer": 16.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Changqiao Jinchuang Technology Co., Ltd. = Consumer Electronics and Electrical Industry",
+    "Extracted from national_industry_status.csv: median annual Chinese invention patent applications in Consumer Electronics and Electrical Industry = 18",
+    "Extracted from company_profile.csv: province of Wuli Huida Chain Co., Ltd. = Zhejiang Province",
+    "Extracted from regional_industry_status.csv: median annual Chinese invention patent applications in Zhejiang Wholesale and Retail Trade = 2",
+    "Calculated the difference: 18 - 2 = 16.0"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "51401b53-3d2a-4dc5-bda7-8a5fc02dd831"
+    },
+    {
+      "company_profile": "58eb73a2-c225-4c83-9324-b515fec3a76f"
+    },
+    {
+      "regional_industry_status": "121a55c3-2823-410f-b764-12733c5e8754"
+    },
+    {
+      "company_profile": "24d73db3-b3be-4f9d-87e8-a8f7bcd7b0f9"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium017_result.json b/assets/qa_raw/enterprise_industry_analysis/medium017_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2489f889ffb6d78b5f1802fd083d9242b9ce7608
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium017_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium017",
+  "question": "What is the difference between the minimum R&D personnel ratio in the industry of Changqiao Jinchuang Technology Co., Ltd. and the corresponding metric for the industry in Zhejiang Province where Wuli Huida Chain Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to two decimal places. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Changqiao Jinchuang Technology Co., Ltd.: Consumer Electronics and Electrical Industry",
+    "Minimum R&D personnel ratio in Consumer Electronics and Electrical Industry: 1.87%",
+    "Province of Wuli Huida Chain Co., Ltd.: Zhejiang Province",
+    "Minimum R&D personnel ratio in Zhejiang Wholesale and Retail Trade: 0.83%"
+  ],
+  "milestone": {
+    "Industry of Changqiao Jinchuang Technology Co., Ltd.": "Consumer Electronics and Electrical Industry",
+    "Minimum R&D personnel ratio in Consumer Electronics and Electrical Industry": "1.87%",
+    "Province of Wuli Huida Chain Co., Ltd.": "Zhejiang Province",
+    "Minimum R&D personnel ratio in Zhejiang Wholesale and Retail Trade": "0.83%",
+    "Difference (Consumer Electronics and Electrical Industry - Zhejiang Wholesale and Retail Trade)": 1.04
+  },
+  "answer": 1.04,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Changqiao Jinchuang Technology Co., Ltd. = Consumer Electronics and Electrical Industry",
+    "Extracted from national_industry_status.csv: minimum R&D personnel ratio in Consumer Electronics and Electrical Industry = 1.87%",
+    "Extracted from company_profile.csv: province of Wuli Huida Chain Co., Ltd. = Zhejiang Province",
+    "Extracted from regional_industry_status.csv: minimum R&D personnel ratio in Zhejiang Wholesale and Retail Trade = 0.83%",
+    "Calculated the difference: 1.87 - 0.83 = 1.04"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "51401b53-3d2a-4dc5-bda7-8a5fc02dd831"
+    },
+    {
+      "company_profile": "58eb73a2-c225-4c83-9324-b515fec3a76f"
+    },
+    {
+      "regional_industry_status": "121a55c3-2823-410f-b764-12733c5e8754"
+    },
+    {
+      "company_profile": "24d73db3-b3be-4f9d-87e8-a8f7bcd7b0f9"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium018_result.json b/assets/qa_raw/enterprise_industry_analysis/medium018_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d5819907b10d7d6b4116415808c7fabc55bb563f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium018_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium018",
+  "question": "Which is greater: the number of SSE-listed local state-owned enterprises in the corresponding industry of the province where Zhong Ke Shu Ruan Software Co., Ltd. is located, or the number of SZSE-listed foreign-funded enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Zhong Ke Shu Ruan Software Co., Ltd.: Guangdong Province",
+    "Number of SSE-listed local state-owned enterprises in Guangdong Information Transmission, Software and IT Services: 1",
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.: Real Estate",
+    "Number of SZSE-listed foreign-funded enterprises in Real Estate: 2"
+  ],
+  "milestone": {
+    "Province of Zhong Ke Shu Ruan Software Co., Ltd.": "Guangdong Province",
+    "SSE-listed local state-owned enterprise count in Guangdong Information Transmission, Software and IT Services": 1,
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "SZSE-listed foreign-funded enterprise count in Real Estate": 2,
+    "Comparison result (which value is greater)": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."
+  },
+  "answer": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Zhong Ke Shu Ruan Software Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: SSE-listed local state-owned enterprise count in Guangdong Information Transmission, Software and IT Services = 1",
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: SZSE-listed foreign-funded enterprise count in Real Estate = 2",
+    "Compared 1 and 2; since 2 is greater, output \"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "5eb24927-5b7a-4561-96df-b86398387e2b"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "a0b01b7e-ed2c-4c4a-b7d7-15062fab3ef4"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium019_result.json b/assets/qa_raw/enterprise_industry_analysis/medium019_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..1c44ad35a1ef535944af6e0f2deb14a30593a56f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium019_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium019",
+  "question": "Which is greater: the number of HKEX-listed foreign-funded enterprises in the corresponding industry of the province where Zhong Ke Shu Ruan Software Co., Ltd. is located, or the number of HKEX-listed enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Zhong Ke Shu Ruan Software Co., Ltd.: Guangdong Province",
+    "Number of HKEX-listed foreign-funded enterprises in Guangdong Information Transmission, Software and IT Services: 5",
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.: Real Estate",
+    "Number of HKEX-listed enterprises in Real Estate: 185"
+  ],
+  "milestone": {
+    "Province of Zhong Ke Shu Ruan Software Co., Ltd.": "Guangdong Province",
+    "HKEX-listed foreign-funded enterprise count in Guangdong Information Transmission, Software and IT Services": 5,
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "HKEX-listed enterprise count in Real Estate": 185,
+    "Comparison result (which value is greater)": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."
+  },
+  "answer": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Zhong Ke Shu Ruan Software Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: HKEX-listed foreign-funded enterprise count in Guangdong Information Transmission, Software and IT Services = 5",
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: HKEX-listed enterprise count in Real Estate = 185",
+    "Compared 5 and 185; since 185 is greater, output \"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "5eb24927-5b7a-4561-96df-b86398387e2b"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "a0b01b7e-ed2c-4c4a-b7d7-15062fab3ef4"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium020_result.json b/assets/qa_raw/enterprise_industry_analysis/medium020_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..cdc2f52dce43e57e54f4cf5255f299c6880ed946
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium020_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium020",
+  "question": "Which is greater: the number of SZSE-listed central state-owned enterprises in the industry of Bi Yuan Zhi Ze Urban Development Co., Ltd., or the number of SSE-listed local state-owned enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Bi Yuan Zhi Ze Urban Development Co., Ltd.: Real Estate",
+    "Number of SZSE-listed central state-owned enterprises in Real Estate: 7",
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.: Real Estate",
+    "Number of SSE-listed local state-owned enterprises in Real Estate: 38"
+  ],
+  "milestone": {
+    "Industry of Bi Yuan Zhi Ze Urban Development Co., Ltd.": "Real Estate",
+    "SZSE-listed central state-owned enterprise count in Real Estate": 7,
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "SSE-listed local state-owned enterprise count in Real Estate": 38,
+    "Comparison result (which is greater)": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."
+  },
+  "answer": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Bi Yuan Zhi Ze Urban Development Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: SZSE-listed central state-owned enterprise count in Real Estate = 7",
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: SSE-listed local state-owned enterprise count in Real Estate = 38",
+    "Compared 7 and 38; since 38 is greater, output \"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "1497d311-a090-4fdb-9af8-30fee577d417"
+    },
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "a0b01b7e-ed2c-4c4a-b7d7-15062fab3ef4"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium021_result.json b/assets/qa_raw/enterprise_industry_analysis/medium021_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5fe37aeef77a0f6ab7d027da4e703a0f5f6037e2
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium021_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium021",
+  "question": "What is the difference between the median operating profit amount of the industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd. and that of the industry of Tong Tong Ze Hong Securities Co., Ltd.?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd.: Real Estate",
+    "Median operating profit amount in Real Estate: 130368786 yuan",
+    "Industry of Tong Tong Ze Hong Securities Co., Ltd.: Financial Industry",
+    "Median operating profit amount in Financial Industry: 1010930425 yuan"
+  ],
+  "milestone": {
+    "Industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd.": "Real Estate",
+    "Median operating profit amount in Real Estate (yuan)": 130368786,
+    "Industry of Tong Tong Ze Hong Securities Co., Ltd.": "Financial Industry",
+    "Median operating profit amount in Financial Industry (yuan)": 1010930425,
+    "Difference (Real Estate - Financial Industry)": -880561639.0
+  },
+  "answer": -880561639.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: median operating profit amount in Real Estate = 130368786",
+    "Extracted from company_profile.csv: industry of Tong Tong Ze Hong Securities Co., Ltd. = Financial Industry",
+    "Extracted from national_industry_status.csv: median operating profit amount in Financial Industry = 1010930425",
+    "Calculated the difference (Real Estate - Financial Industry): 130368786 - 1010930425 = -880561639.0"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "3f2c07ea-c142-43aa-ab56-ee76d8aaa4e5"
+    },
+    {
+      "national_industry_status": "5dc0d9db-af52-48e2-8eef-98a66f67632f"
+    },
+    {
+      "company_profile": "63debe9f-0257-4192-a552-fe35fb82a435"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium022_result.json b/assets/qa_raw/enterprise_industry_analysis/medium022_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..adde6ecbdcc4593d407fd157f76adb459c10d2fd
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium022_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium022",
+  "question": "Comparing the number of SZSE-listed central state-owned enterprises in the industry of Zhaoye Huachang Real Estate Development Co., Ltd. with the number of SZSE-listed enterprises in the industry of Tongtong Zehong Securities Co., Ltd., which is larger?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhaoye Huachang Real Estate Development Co., Ltd.: Real Estate",
+    "Real Estate, number of SZSE-listed central state-owned enterprises: 7",
+    "Industry of Tongtong Zehong Securities Co., Ltd.: Financial Industry",
+    "Financial Industry, number of SZSE-listed enterprises: 38"
+  ],
+  "answer": "Tongtong Zehong Securities Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Zhaoye Huachang Real Estate Development Co., Ltd.": "Real Estate",
+    "Number of SZSE-listed central state-owned enterprises in Real Estate": 7,
+    "Industry of Tongtong Zehong Securities Co., Ltd.": "Financial Industry",
+    "Number of SZSE-listed enterprises in Financial Industry": 38,
+    "Conclusion (larger value)": "Tongtong Zehong Securities Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Zhaoye Huachang Real Estate Development Co., Ltd. is Real Estate.",
+    "Extracted from national_industry_status.csv that the number of SZSE-listed central state-owned enterprises in Real Estate is 7.",
+    "Extracted from company_profile.csv that the industry of Tongtong Zehong Securities Co., Ltd. is Financial Industry.",
+    "Extracted from national_industry_status.csv that the number of SZSE-listed enterprises in Financial Industry is 38.",
+    "Compared 7 and 38; 38 is larger, so output \"Tongtong Zehong Securities Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "3f2c07ea-c142-43aa-ab56-ee76d8aaa4e5"
+    },
+    {
+      "national_industry_status": "5dc0d9db-af52-48e2-8eef-98a66f67632f"
+    },
+    {
+      "company_profile": "63debe9f-0257-4192-a552-fe35fb82a435"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium023_result.json b/assets/qa_raw/enterprise_industry_analysis/medium023_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4daa9bc5a2f7d28122fc386a60254ccd3a2eb2bb
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium023_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium023",
+  "question": "Comparing the number of SZSE-listed local state-owned enterprises in the industry of Aijian Yikang Fuzhongxin Co., Ltd. with the number of SZSE-listed sino-foreign joint ventures in the industry of Zhongke Zhiyun Data Services Co., Ltd., which value is larger?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Aijian Yikang Fuzhongxin Co., Ltd.: Health and Social Work",
+    "Health and Social Work, number of SZSE-listed local state-owned enterprises: 1",
+    "Industry of Zhongke Zhiyun Data Services Co., Ltd.: Information Transmission, Software and Information Technology Services",
+    "Information Transmission, Software and Information Technology Services, number of SZSE-listed sino-foreign joint ventures: 2"
+  ],
+  "answer": "Zhongke Zhiyun Data Services Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Aijian Yikang Fuzhongxin Co., Ltd.": "Health and Social Work",
+    "Number of SZSE-listed local state-owned enterprises in Health and Social Work": 1,
+    "Industry of Zhongke Zhiyun Data Services Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Number of SZSE-listed sino-foreign joint ventures in Information Transmission, Software and Information Technology Services": 2,
+    "Conclusion (larger value)": "Zhongke Zhiyun Data Services Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Aijian Yikang Fuzhongxin Co., Ltd. is Health and Social Work.",
+    "Extracted from national_industry_status.csv that the number of SZSE-listed local state-owned enterprises in Health and Social Work is 1.",
+    "Extracted from company_profile.csv that the industry of Zhongke Zhiyun Data Services Co., Ltd. is Information Transmission, Software and Information Technology Services.",
+    "Extracted from national_industry_status.csv that the number of SZSE-listed sino-foreign joint ventures in Information Transmission, Software and Information Technology Services is 2.",
+    "Compared 1 and 2; 2 is larger, so output \"Zhongke Zhiyun Data Services Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "a08b718a-c663-4252-a4f4-56c3034bae0c"
+    },
+    {
+      "company_profile": "10890d3a-b03b-417b-b39e-f5070604b51e"
+    },
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "d6ab2642-2691-4f63-a197-6a4c29026c0d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium024_result.json b/assets/qa_raw/enterprise_industry_analysis/medium024_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b288708bef8c7ad1b45290670e0922ee5a966a4c
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium024_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium024",
+  "question": "What is the difference between the minimum value of provincial enterprise technology innovation awards in the industry of Aijian Yikang Fuzhongxin Co., Ltd. and that in the industry of Zhongke Zhiyun Data Services Co., Ltd.?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Aijian Yikang Fuzhongxin Co., Ltd.: Health and Social Work",
+    "Minimum value of provincial enterprise technology innovation awards in Health and Social Work: 0",
+    "Industry of Zhongke Zhiyun Data Services Co., Ltd.: Information Transmission, Software and Information Technology Services",
+    "Minimum value of provincial enterprise technology innovation awards in Information Transmission, Software and Information Technology Services: 0"
+  ],
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Aijian Yikang Fuzhongxin Co., Ltd.": "Health and Social Work",
+    "Minimum value of provincial enterprise technology innovation awards in Health and Social Work": 0,
+    "Industry of Zhongke Zhiyun Data Services Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Minimum value of provincial enterprise technology innovation awards in Information Transmission, Software and Information Technology Services": 0,
+    "Difference": 0.0
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Aijian Yikang Fuzhongxin Co., Ltd. is Health and Social Work.",
+    "Extracted from national_industry_status.csv that the minimum value of provincial enterprise technology innovation awards in Health and Social Work is 0.",
+    "Extracted from company_profile.csv that the industry of Zhongke Zhiyun Data Services Co., Ltd. is Information Transmission, Software and Information Technology Services.",
+    "Extracted from national_industry_status.csv that the minimum value of provincial enterprise technology innovation awards in Information Transmission, Software and Information Technology Services is 0.",
+    "Calculated the difference: 0 - 0 = 0.0."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "a08b718a-c663-4252-a4f4-56c3034bae0c"
+    },
+    {
+      "company_profile": "10890d3a-b03b-417b-b39e-f5070604b51e"
+    },
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "d6ab2642-2691-4f63-a197-6a4c29026c0d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium025_result.json b/assets/qa_raw/enterprise_industry_analysis/medium025_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f26d78d051ef1ad716d8362f0ffdcd9895baf5e
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium025_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium025",
+  "question": "Is the average number of provincial or ministerial natural science awards in the industry of Biyuan Shenghua Construction Development Co., Ltd. lower than that in the industry of Baoxin Huihui Network Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Biyuan Shenghua Construction Development Co., Ltd.: Real Estate",
+    "Average number of provincial or ministerial natural science awards in Real Estate: 0",
+    "Industry of Baoxin Huihui Network Co., Ltd.: Information Transmission, Software and Information Technology Services",
+    "Average number of provincial or ministerial natural science awards in Information Transmission, Software and Information Technology Services: 1"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Biyuan Shenghua Construction Development Co., Ltd.": "Real Estate",
+    "Average number of provincial or ministerial natural science awards in Real Estate": 0,
+    "Industry of Baoxin Huihui Network Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Average number of provincial or ministerial natural science awards in Information Transmission, Software and Information Technology Services": 1,
+    "Whether lower": "Yes"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Biyuan Shenghua Construction Development Co., Ltd. is Real Estate.",
+    "Extracted from national_industry_status.csv that the average number of provincial or ministerial natural science awards in Real Estate is 0.",
+    "Extracted from company_profile.csv that the industry of Baoxin Huihui Network Co., Ltd. is Information Transmission, Software and Information Technology Services.",
+    "Extracted from national_industry_status.csv that the average number of provincial or ministerial natural science awards in Information Transmission, Software and Information Technology Services is 1.",
+    "Judged whether 0 is lower than 1; the conclusion is \"Yes\"."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "a0b01b7e-ed2c-4c4a-b7d7-15062fab3ef4"
+    },
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "fe635f7d-bee1-4e2e-9ed6-3923871590d0"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium026_result.json b/assets/qa_raw/enterprise_industry_analysis/medium026_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..60eb8c2f4a48afea4e2028e99b4e960728bad7b3
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium026_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium026",
+  "question": "Is the median cumulative citation count of all patents in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. lower than the same metric in the industry of Bao Xin Hui Hui Wang Luo Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.: Real Estate",
+    "Median cumulative citation count of all patents in Real Estate: 63",
+    "Industry of Bao Xin Hui Hui Wang Luo Co., Ltd.: Information Transmission, Software and IT Services",
+    "Median cumulative citation count of all patents in Information Transmission, Software and IT Services: 340.5"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "Median cumulative citation count of all patents in Real Estate": 63,
+    "Industry of Bao Xin Hui Hui Wang Luo Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Median cumulative citation count of all patents in Information Transmission, Software and IT Services": 340.5,
+    "Whether lower": "Yes"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: median cumulative citation count of all patents in Real Estate = 63",
+    "Extracted from company_profile.csv: industry of Bao Xin Hui Hui Wang Luo Co., Ltd. = Information Transmission, Software and IT Services",
+    "Extracted from national_industry_status.csv: median cumulative citation count of all patents in Information Transmission, Software and IT Services = 340.5",
+    "Compared 63 and 340.5; since 63 is lower, the judgment is \"Yes\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "a0b01b7e-ed2c-4c4a-b7d7-15062fab3ef4"
+    },
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "fe635f7d-bee1-4e2e-9ed6-3923871590d0"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium027_result.json b/assets/qa_raw/enterprise_industry_analysis/medium027_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa25055b32e3376d6bd8380b3f8a48b42746f580
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium027_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium027",
+  "question": "Is the number of SSE-listed enterprises in the industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. equal to the number of SSE-listed state-owned institute enterprises in the industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd.: Health and Social Work",
+    "Number of SSE-listed enterprises in Health and Social Work: 2",
+    "Industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd.: Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Number of SSE-listed state-owned institute enterprises in that industry: 1"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd.": "Health and Social Work",
+    "Number of SSE-listed enterprises in Health and Social Work": 2,
+    "Industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd.": "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Number of SSE-listed state-owned institute enterprises in that industry": 1,
+    "Whether equal": "No"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. = Health and Social Work",
+    "Extracted from national_industry_status.csv: number of SSE-listed enterprises in Health and Social Work = 2",
+    "Extracted from company_profile.csv: industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd. = Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Extracted from national_industry_status.csv: number of SSE-listed state-owned institute enterprises in that industry = 1",
+    "Compared 2 and 1; since they are not equal, the judgment is \"No\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "a08b718a-c663-4252-a4f4-56c3034bae0c"
+    },
+    {
+      "company_profile": "10890d3a-b03b-417b-b39e-f5070604b51e"
+    },
+    {
+      "national_industry_status": "4e4122cb-0cfc-47c7-8dde-7296312f31a7"
+    },
+    {
+      "company_profile": "b5c4bcce-8ecd-4a6e-b6ca-e5b6eb41fb8e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium028_result.json b/assets/qa_raw/enterprise_industry_analysis/medium028_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..988dbbb5a2a1dcc54718076203bcc5c4ec0d8da5
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium028_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium028",
+  "question": "Are the maximum values of the State Technological Invention Award metric the same between the industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. and the industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd.: Health and Social Work",
+    "Maximum State Technological Invention Award value in Health and Social Work: 0",
+    "Industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd.: Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Maximum State Technological Invention Award value in that industry: 0"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd.": "Health and Social Work",
+    "Maximum State Technological Invention Award value in Health and Social Work": 0,
+    "Industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd.": "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Maximum State Technological Invention Award value in that industry": 0,
+    "Whether identical": "Yes"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. = Health and Social Work",
+    "Extracted from national_industry_status.csv: maximum State Technological Invention Award value in Health and Social Work = 0",
+    "Extracted from company_profile.csv: industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd. = Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Extracted from national_industry_status.csv: maximum State Technological Invention Award value in that industry = 0",
+    "Compared 0 and 0; since they are the same, the judgment is \"Yes\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "a08b718a-c663-4252-a4f4-56c3034bae0c"
+    },
+    {
+      "company_profile": "10890d3a-b03b-417b-b39e-f5070604b51e"
+    },
+    {
+      "national_industry_status": "4e4122cb-0cfc-47c7-8dde-7296312f31a7"
+    },
+    {
+      "company_profile": "b5c4bcce-8ecd-4a6e-b6ca-e5b6eb41fb8e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium029_result.json b/assets/qa_raw/enterprise_industry_analysis/medium029_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d8114f753663d10d6e7c1f3038a67d8a07185b52
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium029_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium029",
+  "question": "Which is greater: the number of SSE-listed central state-owned enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd., or the number of SSE-listed enterprises in the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.: Real Estate",
+    "Number of SSE-listed central state-owned enterprises in Real Estate: 3",
+    "Industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.: Education",
+    "Number of SSE-listed enterprises in Education: 5"
+  ],
+  "answer": "Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "Number of SSE-listed central state-owned enterprises in Real Estate": 3,
+    "Industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.": "Education",
+    "Number of SSE-listed enterprises in Education": 5,
+    "Comparison result (greater)": "Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: number of SSE-listed central state-owned enterprises in Real Estate = 3",
+    "Extracted from company_profile.csv: industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. = Education",
+    "Extracted from national_industry_status.csv: number of SSE-listed enterprises in Education = 5",
+    "Compared 3 and 5; since 5 is greater, output \"Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "a0b01b7e-ed2c-4c4a-b7d7-15062fab3ef4"
+    },
+    {
+      "national_industry_status": "c6fd582c-fa1a-4b4f-8d85-c334991e56da"
+    },
+    {
+      "company_profile": "d5008c9a-7492-4ce4-b428-22e6861c68a7"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium030_result.json b/assets/qa_raw/enterprise_industry_analysis/medium030_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..09cf8be2c233931b1dcc16c71a5bcc8a0f93ad00
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium030_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium030",
+  "question": "Which is lower: the minimum total liabilities value in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd., or the same metric in the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.: Real Estate",
+    "Minimum total liabilities in Real Estate: -16582853780.5 yuan",
+    "Industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.: Education",
+    "Minimum total liabilities in Education: 30298132.02 yuan"
+  ],
+  "answer": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "Minimum total liabilities in Real Estate (yuan)": -16582853780.5,
+    "Industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.": "Education",
+    "Minimum total liabilities in Education (yuan)": 30298132.02,
+    "Comparison result (lower)": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: minimum total liabilities in Real Estate = -16582853780.5 yuan",
+    "Extracted from company_profile.csv: industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. = Education",
+    "Extracted from national_industry_status.csv: minimum total liabilities in Education = 30298132.02 yuan",
+    "Compared -16582853780.5 and 30298132.02; since the former is lower, output \"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "a0b01b7e-ed2c-4c4a-b7d7-15062fab3ef4"
+    },
+    {
+      "national_industry_status": "c6fd582c-fa1a-4b4f-8d85-c334991e56da"
+    },
+    {
+      "company_profile": "d5008c9a-7492-4ce4-b428-22e6861c68a7"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium031_result.json b/assets/qa_raw/enterprise_industry_analysis/medium031_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..46bb18dd88600c1289099a977bcff134b6ebf542
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium031_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium031",
+  "question": "Which is greater: the number of SSE-listed state-owned institute enterprises in the industry of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd., or the number of SZSE-listed private enterprises in the industry of Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.: Financial Industry",
+    "Number of SSE-listed state-owned institute enterprises in Financial Industry: 1",
+    "Industry of Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.: Real Estate",
+    "Number of SZSE-listed private enterprises in Real Estate: 24"
+  ],
+  "answer": "Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.": "Financial Industry",
+    "Number of SSE-listed state-owned institute enterprises in Financial Industry": 1,
+    "Industry of Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.": "Real Estate",
+    "Number of SZSE-listed private enterprises in Real Estate": 24,
+    "Comparison result (greater)": "Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. = Financial Industry",
+    "Extracted from national_industry_status.csv: number of SSE-listed state-owned institute enterprises in Financial Industry = 1",
+    "Extracted from company_profile.csv: industry of Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: number of SZSE-listed private enterprises in Real Estate = 24",
+    "Compared 1 and 24; since 24 is greater, output \"Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "5dc0d9db-af52-48e2-8eef-98a66f67632f"
+    },
+    {
+      "company_profile": "62612789-d0a3-4b19-bbee-abe8a89ab1bd"
+    },
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "1497d311-a090-4fdb-9af8-30fee577d417"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium032_result.json b/assets/qa_raw/enterprise_industry_analysis/medium032_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a87adaf1c916e4da407b09715622b2e6da564409
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium032_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium032",
+  "question": "Comparing the number of SZSE-listed private enterprises in the industry of Jinzhi Hongsheng Asset Management Co., Ltd. with the number of SSE-listed private enterprises in the industry of Biyuan Zhize Urban Development Co., Ltd., which is larger?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Jinzhi Hongsheng Asset Management Co., Ltd.: Financial Industry",
+    "Financial Industry, number of SZSE-listed private enterprises: 14",
+    "Industry of Biyuan Zhize Urban Development Co., Ltd.: Real Estate",
+    "Real Estate, number of SSE-listed private enterprises: 12"
+  ],
+  "answer": "Jinzhi Hongsheng Asset Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Jinzhi Hongsheng Asset Management Co., Ltd.": "Financial Industry",
+    "Number of SZSE-listed private enterprises in Financial Industry": 14,
+    "Industry of Biyuan Zhize Urban Development Co., Ltd.": "Real Estate",
+    "Number of SSE-listed private enterprises in Real Estate": 12,
+    "Conclusion (larger value)": "Jinzhi Hongsheng Asset Management Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Jinzhi Hongsheng Asset Management Co., Ltd. is Financial Industry.",
+    "Extracted from national_industry_status.csv that the number of SZSE-listed private enterprises in Financial Industry is 14.",
+    "Extracted from company_profile.csv that the industry of Biyuan Zhize Urban Development Co., Ltd. is Real Estate.",
+    "Extracted from national_industry_status.csv that the number of SSE-listed private enterprises in Real Estate is 12.",
+    "Compared 14 and 12; 14 is larger, so output \"Jinzhi Hongsheng Asset Management Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "5dc0d9db-af52-48e2-8eef-98a66f67632f"
+    },
+    {
+      "company_profile": "62612789-d0a3-4b19-bbee-abe8a89ab1bd"
+    },
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "1497d311-a090-4fdb-9af8-30fee577d417"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium033_result.json b/assets/qa_raw/enterprise_industry_analysis/medium033_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c3fe34a09827841de81a2ec034ac9986f9c05294
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium033_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium033",
+  "question": "Comparing the number of enterprises in Health and Social Work in the industry of Jianfan Ningze Elderly Care Services Co., Ltd. with the number of SSE-listed central state-owned enterprises in the industry of Zhongche Yuanze Shipbuilding Co., Ltd., which value is larger?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Jianfan Ningze Elderly Care Services Co., Ltd.: Health and Social Work",
+    "Health and Social Work, number of enterprises in Health and Social Work: 29",
+    "Industry of Zhongche Yuanze Shipbuilding Co., Ltd.: Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing, number of SSE-listed central state-owned enterprises: 25"
+  ],
+  "answer": "Jianfan Ningze Elderly Care Services Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Jianfan Ningze Elderly Care Services Co., Ltd.": "Health and Social Work",
+    "Number of enterprises in Health and Social Work": 29,
+    "Industry of Zhongche Yuanze Shipbuilding Co., Ltd.": "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Number of SSE-listed central state-owned enterprises in Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing": 25,
+    "Conclusion (larger value)": "Jianfan Ningze Elderly Care Services Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Jianfan Ningze Elderly Care Services Co., Ltd. is Health and Social Work.",
+    "Extracted from national_industry_status.csv that the number of enterprises in Health and Social Work is 29.",
+    "Extracted from company_profile.csv that the industry of Zhongche Yuanze Shipbuilding Co., Ltd. is Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing.",
+    "Extracted from national_industry_status.csv that the number of SSE-listed central state-owned enterprises in this industry is 25.",
+    "Compared 29 and 25; 29 is larger, so output \"Jianfan Ningze Elderly Care Services Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "a08b718a-c663-4252-a4f4-56c3034bae0c"
+    },
+    {
+      "company_profile": "ca56f5f4-4ea0-433d-9eb2-cf2c630fc69d"
+    },
+    {
+      "national_industry_status": "4e4122cb-0cfc-47c7-8dde-7296312f31a7"
+    },
+    {
+      "company_profile": "b5c4bcce-8ecd-4a6e-b6ca-e5b6eb41fb8e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium034_result.json b/assets/qa_raw/enterprise_industry_analysis/medium034_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..6919c2582bc4b8c8a6ff481b193fed5206655442
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium034_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium034",
+  "question": "Is the minimum market capitalization in the industry of Jianfan Ningze Elderly Care Services Co., Ltd. lower than the minimum market capitalization in the industry of Zhongche Yuanze Shipbuilding Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or notes. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Jianfan Ningze Elderly Care Services Co., Ltd.: Health and Social Work",
+    "Minimum market capitalization in Health and Social Work: 2.04 hundred million yuan",
+    "Industry of Zhongche Yuanze Shipbuilding Co., Ltd.: Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Minimum market capitalization in Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing: 6.2 hundred million yuan"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Jianfan Ningze Elderly Care Services Co., Ltd.": "Health and Social Work",
+    "Minimum market capitalization in Health and Social Work (hundred million yuan)": 2.04,
+    "Industry of Zhongche Yuanze Shipbuilding Co., Ltd.": "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Minimum market capitalization in Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing (hundred million yuan)": 6.2,
+    "Whether lower": "Yes"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Jianfan Ningze Elderly Care Services Co., Ltd. is Health and Social Work.",
+    "Extracted from national_industry_status.csv that the minimum market capitalization in Health and Social Work is 2.04 hundred million yuan.",
+    "Extracted from company_profile.csv that the industry of Zhongche Yuanze Shipbuilding Co., Ltd. is Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing.",
+    "Extracted from national_industry_status.csv that the minimum market capitalization in this industry is 6.2 hundred million yuan.",
+    "Judged whether 2.04 is lower than 6.2; the conclusion is \"Yes\"."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "a08b718a-c663-4252-a4f4-56c3034bae0c"
+    },
+    {
+      "company_profile": "ca56f5f4-4ea0-433d-9eb2-cf2c630fc69d"
+    },
+    {
+      "national_industry_status": "4e4122cb-0cfc-47c7-8dde-7296312f31a7"
+    },
+    {
+      "company_profile": "b5c4bcce-8ecd-4a6e-b6ca-e5b6eb41fb8e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium035_result.json b/assets/qa_raw/enterprise_industry_analysis/medium035_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..6eec33a5cf684d1bf378b5d3b310b554dcd2922d
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium035_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium035",
+  "question": "Comparing the number of HKEX-listed sino-foreign joint ventures in the industry of Yihai Changjin Business Co., Ltd. with the number of SSE-listed local state-owned enterprises in the corresponding industry of Shanghai where Jinzhi Hongsheng Asset Management Co., Ltd. is located, which is larger?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Yihai Changjin Business Co., Ltd.: Leasing and Business Services",
+    "Leasing and Business Services, number of HKEX-listed sino-foreign joint ventures: 1",
+    "Province/municipality where Jinzhi Hongsheng Asset Management Co., Ltd. is located: Shanghai Municipality",
+    "Shanghai Municipality, Financial Industry, number of SSE-listed local state-owned enterprises: 8"
+  ],
+  "answer": "Jinzhi Hongsheng Asset Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Yihai Changjin Business Co., Ltd.": "Leasing and Business Services",
+    "Number of HKEX-listed sino-foreign joint ventures in Leasing and Business Services": 1,
+    "Province/municipality where Jinzhi Hongsheng Asset Management Co., Ltd. is located": "Shanghai Municipality",
+    "Number of SSE-listed local state-owned enterprises in Shanghai Financial Industry": 8,
+    "Conclusion (larger value)": "Jinzhi Hongsheng Asset Management Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Yihai Changjin Business Co., Ltd. is Leasing and Business Services.",
+    "Extracted from national_industry_status.csv that the number of HKEX-listed sino-foreign joint ventures in Leasing and Business Services is 1.",
+    "Extracted from company_profile.csv that Jinzhi Hongsheng Asset Management Co., Ltd. is located in Shanghai Municipality.",
+    "Extracted from regional_industry_status.csv that the number of SSE-listed local state-owned enterprises in Shanghai Financial Industry is 8.",
+    "Compared 1 and 8; 8 is larger, so output \"Jinzhi Hongsheng Asset Management Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "e4578b09-ad56-4edf-adc8-9cd6b54149b7"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    },
+    {
+      "regional_industry_status": "6d7fb68b-9d0a-4cfb-bab2-abac83115eed"
+    },
+    {
+      "company_profile": "62612789-d0a3-4b19-bbee-abe8a89ab1bd"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium036_result.json b/assets/qa_raw/enterprise_industry_analysis/medium036_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc52f1ecae2e7a9392f421ef033dfee56b951ecf
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium036_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium036",
+  "question": "Is the mean annual number of Chinese invention patent applications in the industry of Yi Hai Chang Jin Shang Wu Co., Ltd. greater than the corresponding industry metric in Shanghai, where Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. is located?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.: Leasing and Business Services",
+    "Mean annual Chinese invention patent applications in Leasing and Business Services: 9.27586206896552",
+    "Province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.: Shanghai Municipality",
+    "Mean annual Chinese invention patent applications in Shanghai Financial Industry: 35.2380952380952"
+  ],
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Mean annual Chinese invention patent applications in Leasing and Business Services": 9.27586206896552,
+    "Province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.": "Shanghai Municipality",
+    "Mean annual Chinese invention patent applications in Shanghai Financial Industry": 35.2380952380952,
+    "Whether greater": "No"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Yi Hai Chang Jin Shang Wu Co., Ltd. = Leasing and Business Services",
+    "Extracted from national_industry_status.csv: mean annual Chinese invention patent applications in Leasing and Business Services = 9.27586206896552",
+    "Extracted from company_profile.csv: province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: mean annual Chinese invention patent applications in Shanghai Financial Industry = 35.2380952380952",
+    "Compared 9.27586206896552 and 35.2380952380952; since the former is smaller, the answer is \"No\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "e4578b09-ad56-4edf-adc8-9cd6b54149b7"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    },
+    {
+      "regional_industry_status": "6d7fb68b-9d0a-4cfb-bab2-abac83115eed"
+    },
+    {
+      "company_profile": "62612789-d0a3-4b19-bbee-abe8a89ab1bd"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium037_result.json b/assets/qa_raw/enterprise_industry_analysis/medium037_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5803cbca9f6650d17fa02d426de0461a8ebfaa5e
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium037_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium037",
+  "question": "Which is greater: the number of SZSE-listed local state-owned enterprises in the industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd., or the number of HKEX-listed enterprises in the corresponding industry of Guangdong Province where Gao Yin Ze Tong Pi Fa Co., Ltd. is located?",
+  "guidelines": "The answer must be either \"Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. industry local state-owned enterprise SZSE-listed count\" or \"HKEX-listed enterprise count\". Output only the answer text without explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.: Information Transmission, Software and IT Services",
+    "SZSE-listed local state-owned enterprise count in that industry: 26",
+    "Province of Gao Yin Ze Tong Pi Fa Co., Ltd.: Guangdong Province",
+    "HKEX-listed enterprise count in Guangdong Wholesale and Retail: 18"
+  ],
+  "answer": "Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. industry local state-owned enterprise SZSE-listed count",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.": "Information Transmission, Software and IT Services",
+    "SZSE-listed local state-owned enterprise count in that industry": 26,
+    "Province of Gao Yin Ze Tong Pi Fa Co., Ltd.": "Guangdong Province",
+    "HKEX-listed enterprise count in Guangdong Wholesale and Retail": 18,
+    "Comparison result (greater)": "Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. industry local state-owned enterprise SZSE-listed count"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. = Information Transmission, Software and IT Services",
+    "Extracted from national_industry_status.csv: SZSE-listed local state-owned enterprise count in that industry = 26",
+    "Extracted from company_profile.csv: province of Gao Yin Ze Tong Pi Fa Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: HKEX-listed enterprise count in Guangdong Wholesale and Retail = 18",
+    "Compared 26 and 18; since 26 is greater, output the first allowed option"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "d6ab2642-2691-4f63-a197-6a4c29026c0d"
+    },
+    {
+      "regional_industry_status": "f9807af7-8cc0-4d23-9032-70752be13a89"
+    },
+    {
+      "company_profile": "a9b797ac-f74d-4a35-acfa-01b494fe2b3f"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium038_result.json b/assets/qa_raw/enterprise_industry_analysis/medium038_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd571bb05d4c869af1594bd8a6f08af0873062d3
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium038_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium038",
+  "question": "What is the difference between the maximum capitalized R&D expenditure of the industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. and the corresponding industry metric in Guangdong Province where Gao Yin Ze Tong Pi Fa Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to two decimal places. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.: Information Transmission, Software and IT Services",
+    "Maximum capitalized R&D expenditure in that industry: 4055287084 yuan",
+    "Province of Gao Yin Ze Tong Pi Fa Co., Ltd.: Guangdong Province",
+    "Maximum capitalized R&D expenditure in Guangdong Wholesale and Retail: 29209910.11 yuan"
+  ],
+  "answer": 4026077173.89,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Maximum capitalized R&D expenditure in that industry (yuan)": 4055287084,
+    "Province of Gao Yin Ze Tong Pi Fa Co., Ltd.": "Guangdong Province",
+    "Maximum capitalized R&D expenditure in Guangdong Wholesale and Retail (yuan)": 29209910.11,
+    "Difference": 4026077173.89
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. = Information Transmission, Software and IT Services",
+    "Extracted from national_industry_status.csv: maximum capitalized R&D expenditure in that industry = 4055287084",
+    "Extracted from company_profile.csv: province of Gao Yin Ze Tong Pi Fa Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: maximum capitalized R&D expenditure in Guangdong Wholesale and Retail = 29209910.11",
+    "Calculated difference (national industry metric - provincial industry metric): 4055287084 - 29209910.11 = 4026077173.89"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "d6ab2642-2691-4f63-a197-6a4c29026c0d"
+    },
+    {
+      "regional_industry_status": "f9807af7-8cc0-4d23-9032-70752be13a89"
+    },
+    {
+      "company_profile": "a9b797ac-f74d-4a35-acfa-01b494fe2b3f"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium039_result.json b/assets/qa_raw/enterprise_industry_analysis/medium039_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..9f5f42016da092db5c30eac9ef03536947dca4b9
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium039_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium039",
+  "question": "Is there any difference between the minimum State Technological Invention Award value in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. and the corresponding industry metric in Shanghai where Lang Ji Hui Ruan Technology Co., Ltd. is located?",
+  "guidelines": "The answer must be \"No difference\" or the other entity mentioned in the question. Output only the entity text without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.: Real Estate",
+    "Minimum State Technological Invention Award value in Real Estate: 0",
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.: Shanghai Municipality",
+    "Minimum State Technological Invention Award value in Shanghai Information Transmission, Software and IT Services: 0"
+  ],
+  "answer": "No difference",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "Minimum State Technological Invention Award value in Real Estate": 0,
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.": "Shanghai Municipality",
+    "Minimum State Technological Invention Award value in Shanghai Information Transmission, Software and IT Services": 0,
+    "Whether different": "No difference"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: minimum State Technological Invention Award value in Real Estate = 0",
+    "Extracted from company_profile.csv: province of Lang Ji Hui Ruan Technology Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: minimum State Technological Invention Award value in Shanghai Information Transmission, Software and IT Services = 0",
+    "Both values are 0, so output \"No difference\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "a0b01b7e-ed2c-4c4a-b7d7-15062fab3ef4"
+    },
+    {
+      "regional_industry_status": "be71aa1d-8666-44e9-8f26-ffd92dc87001"
+    },
+    {
+      "company_profile": "e9d63626-4106-4f18-8022-b5396454e690"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium040_result.json b/assets/qa_raw/enterprise_industry_analysis/medium040_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d8415fbe9370fba3f738c6b90a84fc55b999eec
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium040_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium040",
+  "question": "Which is larger: the number of HKEX-listed foreign-funded enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd., or the number of SSE-listed enterprises in the corresponding industry of Shanghai where Lang Ji Hui Ruan Technology Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.: Real Estate",
+    "HKEX-listed foreign-funded enterprise count in Real Estate: 34",
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.: Shanghai Municipality",
+    "SSE-listed enterprise count in Shanghai Information Transmission, Software and IT Services: 27"
+  ],
+  "answer": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.": "Real Estate",
+    "HKEX-listed foreign-funded enterprise count in Real Estate": 34,
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.": "Shanghai Municipality",
+    "SSE-listed enterprise count in Shanghai Information Transmission, Software and IT Services": 27,
+    "Comparison result (larger)": "Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: HKEX-listed foreign-funded enterprise count in Real Estate = 34",
+    "Extracted from company_profile.csv: province of Lang Ji Hui Ruan Technology Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: SSE-listed enterprise count in Shanghai Information Transmission, Software and IT Services = 27",
+    "Compared 34 and 27; since 34 is larger, output \"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "a0b01b7e-ed2c-4c4a-b7d7-15062fab3ef4"
+    },
+    {
+      "regional_industry_status": "be71aa1d-8666-44e9-8f26-ffd92dc87001"
+    },
+    {
+      "company_profile": "e9d63626-4106-4f18-8022-b5396454e690"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium041_result.json b/assets/qa_raw/enterprise_industry_analysis/medium041_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d7cf25e5f3b10e30fc0e82496da764a81fbd8aed
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium041_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium041",
+  "question": "Which is larger: the number of SSE-listed local state-owned enterprises in the industry of Hua Xin Yuan Shi New Materials Co., Ltd., or the number of BSE-listed enterprises in the corresponding industry of Guangdong where Zhong Ke Ke Shu Software Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hua Xin Yuan Shi New Materials Co., Ltd.: Non-metallic Mineral Products",
+    "SSE-listed local state-owned enterprise count in Non-metallic Mineral Products: 13",
+    "Province of Zhong Ke Ke Shu Software Co., Ltd.: Guangdong Province",
+    "BSE-listed enterprise count in Guangdong Information Transmission, Software and IT Services: 1"
+  ],
+  "answer": "Hua Xin Yuan Shi New Materials Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Hua Xin Yuan Shi New Materials Co., Ltd.": "Non-metallic Mineral Products",
+    "SSE-listed local state-owned enterprise count in Non-metallic Mineral Products": 13,
+    "Province of Zhong Ke Ke Shu Software Co., Ltd.": "Guangdong Province",
+    "BSE-listed enterprise count in Guangdong Information Transmission, Software and IT Services": 1,
+    "Comparison result (larger)": "Hua Xin Yuan Shi New Materials Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv: industry of Hua Xin Yuan Shi New Materials Co., Ltd. = Non-metallic Mineral Products",
+    "Extracted from national_industry_status.csv: SSE-listed local state-owned enterprise count in Non-metallic Mineral Products = 13",
+    "Extracted from company_profile.csv: province of Zhong Ke Ke Shu Software Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: BSE-listed enterprise count in Guangdong Information Transmission, Software and IT Services = 1",
+    "Compared 13 and 1; since 13 is larger, output \"Hua Xin Yuan Shi New Materials Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "be3c0164-12b0-4453-8c7a-e198721d914b"
+    },
+    {
+      "company_profile": "f150e113-74af-4929-b33d-7b30a892e86d"
+    },
+    {
+      "regional_industry_status": "5eb24927-5b7a-4561-96df-b86398387e2b"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium042_result.json b/assets/qa_raw/enterprise_industry_analysis/medium042_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..35a9401cc6d4611cf3fc80cef8c510132678cce3
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium042_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium042",
+  "question": "Comparing the minimum cumulative citation count of core patents in the industry of Huijin Jinrui Wealth Management Co., Ltd. with the same indicator in the corresponding industry of the province where the company is located, which value is larger?",
+  "guidelines": "The answer must be either \"minimum cumulative citation count of core patents in the industry of Huijin Jinrui Wealth Management Co., Ltd.\" or \"the same indicator in the corresponding industry of the province where the company is located\". Output only the province or region name without any explanation, analysis, or descriptive text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Huijin Jinrui Wealth Management Co., Ltd.: Financial Industry",
+    "Minimum cumulative citation count of core patents in Financial Industry: 0",
+    "Province where Huijin Jinrui Wealth Management Co., Ltd. is located: Guangdong Province",
+    "Minimum cumulative citation count of core patents in Guangdong Financial Industry: 0"
+  ],
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Huijin Jinrui Wealth Management Co., Ltd.": "Financial Industry",
+    "Minimum cumulative citation count of core patents in Financial Industry": 0,
+    "Province where Huijin Jinrui Wealth Management Co., Ltd. is located": "Guangdong Province",
+    "Minimum cumulative citation count of core patents in Guangdong Financial Industry": 0,
+    "Comparison conclusion": "Equal"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Huijin Jinrui Wealth Management Co., Ltd. is Financial Industry.",
+    "Extracted from national_industry_status.csv that the minimum cumulative citation count of core patents in Financial Industry is 0.",
+    "Extracted from company_profile.csv that the province where Huijin Jinrui Wealth Management Co., Ltd. is located is Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the minimum cumulative citation count of core patents in Guangdong Financial Industry is 0.",
+    "Both values are 0, so neither side is larger; output \"Equal\"."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "5dc0d9db-af52-48e2-8eef-98a66f67632f"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    },
+    {
+      "regional_industry_status": "69c8be8c-6e7d-4ef2-b1ba-f721802f73cb"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium043_result.json b/assets/qa_raw/enterprise_industry_analysis/medium043_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce82cb525e11b72fb7cb5af14d8f5494b291576c
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium043_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium043",
+  "question": "What is the difference between the number of SZSE-listed local state-owned enterprises in the industry of Huijin Jinrui Wealth Management Co., Ltd. and the number of SZSE-listed enterprises in the corresponding industry of the province where the company is located?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number without units, commas, or any explanatory text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Huijin Jinrui Wealth Management Co., Ltd.: Financial Industry",
+    "Financial Industry, number of SZSE-listed local state-owned enterprises: 18",
+    "Province where Huijin Jinrui Wealth Management Co., Ltd. is located: Guangdong Province",
+    "Guangdong Financial Industry, number of SZSE-listed enterprises: 8"
+  ],
+  "answer": 10.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Huijin Jinrui Wealth Management Co., Ltd.": "Financial Industry",
+    "Number of SZSE-listed local state-owned enterprises in Financial Industry": 18,
+    "Province where Huijin Jinrui Wealth Management Co., Ltd. is located": "Guangdong Province",
+    "Number of SZSE-listed enterprises in Guangdong Financial Industry": 8,
+    "Difference": 10.0
+  },
+  "steps": [
+    "Extracted from company_profile.csv that the industry of Huijin Jinrui Wealth Management Co., Ltd. is Financial Industry.",
+    "Extracted from national_industry_status.csv that the number of SZSE-listed local state-owned enterprises in Financial Industry is 18.",
+    "Extracted from company_profile.csv that the province where Huijin Jinrui Wealth Management Co., Ltd. is located is Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the number of SZSE-listed enterprises in Guangdong Financial Industry is 8.",
+    "Calculated the difference (national-industry local-SOE SZSE count minus provincial-industry SZSE count): 18 - 8 = 10.0."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "national_industry_status": "5dc0d9db-af52-48e2-8eef-98a66f67632f"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    },
+    {
+      "regional_industry_status": "69c8be8c-6e7d-4ef2-b1ba-f721802f73cb"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium044_result.json b/assets/qa_raw/enterprise_industry_analysis/medium044_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c3142a523e37f64855184950dc7c22bdc2ee6a9a
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium044_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium044",
+  "question": "Which is higher: the number of SSE-listed enterprises in the corresponding industry of the province where Biyuan Shenghua Construction Development Co., Ltd. is located, or the number of HKEX-listed local state-owned enterprises in the industry of Huaying Taisheng Wealth Management Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province where Biyuan Shenghua Construction Development Co., Ltd. is located: Guangdong Province",
+    "Guangdong Real Estate, number of SSE-listed enterprises: 7",
+    "Industry of Huaying Taisheng Wealth Management Co., Ltd.: Financial Industry",
+    "Financial Industry, number of HKEX-listed local state-owned enterprises: 40"
+  ],
+  "answer": "Huaying Taisheng Wealth Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province where Biyuan Shenghua Construction Development Co., Ltd. is located": "Guangdong Province",
+    "Number of SSE-listed enterprises in Guangdong Real Estate": 7,
+    "Industry of Huaying Taisheng Wealth Management Co., Ltd.": "Financial Industry",
+    "Number of HKEX-listed local state-owned enterprises in Financial Industry": 40,
+    "Comparison conclusion (higher value)": "Huaying Taisheng Wealth Management Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv that Biyuan Shenghua Construction Development Co., Ltd. is located in Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the number of SSE-listed enterprises in Guangdong Real Estate is 7.",
+    "Extracted from company_profile.csv that the industry of Huaying Taisheng Wealth Management Co., Ltd. is Financial Industry.",
+    "Extracted from national_industry_status.csv that the number of HKEX-listed local state-owned enterprises in Financial Industry is 40.",
+    "Compared 7 and 40; 40 is higher, so output \"Huaying Taisheng Wealth Management Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "1e971ba1-a0ba-4221-908d-a308281f06d3"
+    },
+    {
+      "company_profile": "a0b01b7e-ed2c-4c4a-b7d7-15062fab3ef4"
+    },
+    {
+      "national_industry_status": "5dc0d9db-af52-48e2-8eef-98a66f67632f"
+    },
+    {
+      "company_profile": "b3ffe2d5-c5b8-4499-b7e7-c6955e84e897"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium045_result.json b/assets/qa_raw/enterprise_industry_analysis/medium045_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8c87c693944e6405cd77a70cc4f78e17c50627b
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium045_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium045",
+  "question": "Comparing the number of HKEX-listed enterprises in the province where Huatu Wenjiao Online Education Co., Ltd. is located with the number of SZSE-listed enterprises in the industry of Yihai Changjin Business Co., Ltd., which value is larger?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province where Huatu Wenjiao Online Education Co., Ltd. is located: Shanghai Municipality",
+    "Shanghai Education, number of HKEX-listed enterprises: 3",
+    "Industry of Yihai Changjin Business Co., Ltd.: Leasing and Business Services",
+    "Leasing and Business Services, number of SZSE-listed enterprises: 44"
+  ],
+  "answer": "Yihai Changjin Business Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province where Huatu Wenjiao Online Education Co., Ltd. is located": "Shanghai Municipality",
+    "Number of HKEX-listed enterprises in Shanghai Education": 3,
+    "Industry of Yihai Changjin Business Co., Ltd.": "Leasing and Business Services",
+    "Number of SZSE-listed enterprises in Leasing and Business Services": 44,
+    "Comparison conclusion (larger value)": "Yihai Changjin Business Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv that Huatu Wenjiao Online Education Co., Ltd. is located in Shanghai Municipality.",
+    "Extracted from regional_industry_status.csv that the number of HKEX-listed enterprises in Shanghai Education is 3.",
+    "Extracted from company_profile.csv that the industry of Yihai Changjin Business Co., Ltd. is Leasing and Business Services.",
+    "Extracted from national_industry_status.csv that the number of SZSE-listed enterprises in Leasing and Business Services is 44.",
+    "Compared 3 and 44; 44 is larger, so output \"Yihai Changjin Business Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "fdbbb94b-1ec1-433a-951c-dbf795fc477d"
+    },
+    {
+      "company_profile": "d5008c9a-7492-4ce4-b428-22e6861c68a7"
+    },
+    {
+      "national_industry_status": "e4578b09-ad56-4edf-adc8-9cd6b54149b7"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium046_result.json b/assets/qa_raw/enterprise_industry_analysis/medium046_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf0709810391855f6133f69a46784741234a772d
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium046_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium046",
+  "question": "Which is higher: the median operating profit of the corresponding industry in Shanghai, where Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. is located, or the median operating profit of the industry of Yi Hai Chang Jin Shang Wu Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.: Shanghai Municipality",
+    "Shanghai Municipality Education Operating profit (median)：129544365.28 Yuan",
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.: Leasing and Business Services",
+    "Median operating profit in Leasing and Business Services: 11445832 Yuan"
+  ],
+  "answer": "Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Education Operating profit (median)（Yuan）": 129544365.28,
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Median operating profit in Leasing and Business Services (Yuan)": 11445832,
+    "Comparison result (higher)": "Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: median operating profit in Shanghai Education = 129544365.28 Yuan",
+    "Extracted from company_profile.csv: industry of Yi Hai Chang Jin Shang Wu Co., Ltd. = Leasing and Business Services",
+    "Extracted from national_industry_status.csv: median operating profit in Leasing and Business Services = 11445832 Yuan",
+    "Compared the two values; since 129544365.28 is higher, output \"Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "fdbbb94b-1ec1-433a-951c-dbf795fc477d"
+    },
+    {
+      "company_profile": "d5008c9a-7492-4ce4-b428-22e6861c68a7"
+    },
+    {
+      "national_industry_status": "e4578b09-ad56-4edf-adc8-9cd6b54149b7"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium047_result.json b/assets/qa_raw/enterprise_industry_analysis/medium047_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..6a54f079b06f3bc2bcc7ff06cc542cf944a1b805
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium047_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium047",
+  "question": "Which is higher: the maximum cumulative citations of all patents in the corresponding industry in Shanghai, where Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. is located, or the same metric in the industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.: Shanghai Municipality",
+    "Shanghai Municipality Financial Industry Cumulative citations of all patents (maximum)：11161 items",
+    "Industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.: Real Estate",
+    "Maximum cumulative citations of all patents in Real Estate: 6946"
+  ],
+  "answer": "Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Financial Industry Cumulative citations of all patents (maximum)": 11161,
+    "Industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.": "Real Estate",
+    "Maximum cumulative citations of all patents in Real Estate": 6946,
+    "Comparison result (higher)": "Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: maximum cumulative citations of all patents in Shanghai Financial Industry = 11161",
+    "Extracted from company_profile.csv: industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: maximum cumulative citations of all patents in Real Estate = 6946",
+    "Compared 11161 and 6946; since 11161 is higher, output \"Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "6d7fb68b-9d0a-4cfb-bab2-abac83115eed"
+    },
+    {
+      "company_profile": "62612789-d0a3-4b19-bbee-abe8a89ab1bd"
+    },
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "3f2c07ea-c142-43aa-ab56-ee76d8aaa4e5"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium048_result.json b/assets/qa_raw/enterprise_industry_analysis/medium048_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..6026a8ed2b128ca5c03f4b20e061d18c26796f91
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium048_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium048",
+  "question": "Which is larger: the number of HKEX-listed central state-owned enterprises in the corresponding industry in Shanghai where Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. is located, or the number of SSE-listed local state-owned enterprises in the industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.: Shanghai Municipality",
+    "Shanghai Municipality Financial Industry Central state-owned enterprise_Number of HKEX-listed enterprises：3",
+    "Industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.: Real Estate",
+    "SSE-listed local state-owned enterprise count in Real Estate: 38"
+  ],
+  "answer": "Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Financial Industry Central state-owned enterprise_Number of HKEX-listed enterprises": 3,
+    "Industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.": "Real Estate",
+    "SSE-listed local state-owned enterprise count in Real Estate": 38,
+    "Comparison result (larger)": "Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: HKEX-listed central state-owned enterprise count in Shanghai Financial Industry = 3",
+    "Extracted from company_profile.csv: industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. = Real Estate",
+    "Extracted from national_industry_status.csv: SSE-listed local state-owned enterprise count in Real Estate = 38",
+    "Compared 3 and 38; since 38 is larger, output \"Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "6d7fb68b-9d0a-4cfb-bab2-abac83115eed"
+    },
+    {
+      "company_profile": "62612789-d0a3-4b19-bbee-abe8a89ab1bd"
+    },
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "3f2c07ea-c142-43aa-ab56-ee76d8aaa4e5"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium049_result.json b/assets/qa_raw/enterprise_industry_analysis/medium049_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0402be8cfda43d56d49bd491511ce309f9f86c62
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium049_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium049",
+  "question": "What is the difference between the total number of enterprises in the corresponding industry of Tianjin, where Bi Yuan Chan Jin Bu Dong Chan Co., Ltd. is located, and the number of SZSE-listed state-owned institute enterprises in the industry of Zhang Qiao Jin Chuang Technology Co., Ltd.?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number, without units, commas, or any text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Bi Yuan Chan Jin Bu Dong Chan Co., Ltd.: Tianjin Municipality",
+    "Total enterprise count in Tianjin Real Estate: 7",
+    "Industry of Zhang Qiao Jin Chuang Technology Co., Ltd.: Consumer Electronics and Electrical Industry",
+    "SZSE-listed state-owned institute enterprise count in Consumer Electronics and Electrical Industry: 1"
+  ],
+  "answer": 6.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Bi Yuan Chan Jin Bu Dong Chan Co., Ltd.": "Tianjin Municipality",
+    "Total enterprise count in Tianjin Real Estate": 7,
+    "Industry of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Consumer Electronics and Electrical Industry",
+    "SZSE-listed state-owned institute enterprise count in Consumer Electronics and Electrical Industry": 1,
+    "difference": 6.0
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Bi Yuan Chan Jin Bu Dong Chan Co., Ltd. = Tianjin Municipality",
+    "Extracted from regional_industry_status.csv: total enterprise count in Tianjin Real Estate = 7",
+    "Extracted from company_profile.csv: industry of Zhang Qiao Jin Chuang Technology Co., Ltd. = Consumer Electronics and Electrical Industry",
+    "Extracted from national_industry_status.csv: SZSE-listed state-owned institute enterprise count in Consumer Electronics and Electrical Industry = 1",
+    "Calculated difference: 7 - 1 = 6.0"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "f570645f-cb9b-4537-bbc4-39cfb7471a8c"
+    },
+    {
+      "company_profile": "7569b54a-60ea-4334-be2d-085b96a7730b"
+    },
+    {
+      "national_industry_status": "51401b53-3d2a-4dc5-bda7-8a5fc02dd831"
+    },
+    {
+      "company_profile": "58eb73a2-c225-4c83-9324-b515fec3a76f"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium050_result.json b/assets/qa_raw/enterprise_industry_analysis/medium050_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..87663a30758d87ef2f39dafd431bbabd7b07f285
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium050_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium050",
+  "question": "What is the difference between the number of SZSE-listed local state-owned enterprises in the corresponding industry of Tianjin, where Bi Yuan Chan Jin Bu Dong Chan Co., Ltd. is located, and the number of HKEX-listed private enterprises in the industry of Zhang Qiao Jin Chuang Technology Co., Ltd.?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number, without units, commas, or any text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Bi Yuan Chan Jin Bu Dong Chan Co., Ltd.: Tianjin Municipality",
+    "SZSE-listed local state-owned enterprise count in Tianjin Real Estate: 2",
+    "Industry of Zhang Qiao Jin Chuang Technology Co., Ltd.: Consumer Electronics and Electrical Industry",
+    "HKEX-listed private enterprise count in Consumer Electronics and Electrical Industry: 45"
+  ],
+  "answer": -43.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Bi Yuan Chan Jin Bu Dong Chan Co., Ltd.": "Tianjin Municipality",
+    "SZSE-listed local state-owned enterprise count in Tianjin Real Estate": 2,
+    "Industry of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Consumer Electronics and Electrical Industry",
+    "HKEX-listed private enterprise count in Consumer Electronics and Electrical Industry": 45,
+    "difference": -43.0
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Bi Yuan Chan Jin Bu Dong Chan Co., Ltd. = Tianjin Municipality",
+    "Extracted from regional_industry_status.csv: SZSE-listed local state-owned enterprise count in Tianjin Real Estate = 2",
+    "Extracted from company_profile.csv: industry of Zhang Qiao Jin Chuang Technology Co., Ltd. = Consumer Electronics and Electrical Industry",
+    "Extracted from national_industry_status.csv: HKEX-listed private enterprise count in Consumer Electronics and Electrical Industry = 45",
+    "Calculated difference: 2 - 45 = -43.0"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "f570645f-cb9b-4537-bbc4-39cfb7471a8c"
+    },
+    {
+      "company_profile": "7569b54a-60ea-4334-be2d-085b96a7730b"
+    },
+    {
+      "national_industry_status": "51401b53-3d2a-4dc5-bda7-8a5fc02dd831"
+    },
+    {
+      "company_profile": "58eb73a2-c225-4c83-9324-b515fec3a76f"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium051_result.json b/assets/qa_raw/enterprise_industry_analysis/medium051_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b3245aaaa45a67a87506f15096a1a2d0f596ec77
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium051_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium051",
+  "question": "What is the gap between the number of SZSE-listed foreign-funded enterprises in the corresponding industry of Guangdong, where Gao Yin Ze Tong Pi Fa Co., Ltd. is located, and the number of SSE-listed enterprises in the industry of Lang Ji Hui Ruan Technology Co., Ltd.?",
+  "guidelines": "The answer must be a number with one decimal place. Output only the number, without units, commas, or any text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Gao Yin Ze Tong Pi Fa Co., Ltd.: Guangdong Province",
+    "SZSE-listed foreign-funded enterprise count in Guangdong Wholesale and Retail: 2",
+    "Industry of Lang Ji Hui Ruan Technology Co., Ltd.: Information Transmission, Software and IT Services",
+    "SSE-listed enterprise count in Information Transmission, Software and IT Services: 141"
+  ],
+  "answer": -139.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Gao Yin Ze Tong Pi Fa Co., Ltd.": "Guangdong Province",
+    "SZSE-listed foreign-funded enterprise count in Guangdong Wholesale and Retail": 2,
+    "Industry of Lang Ji Hui Ruan Technology Co., Ltd.": "Information Transmission, Software and IT Services",
+    "SSE-listed enterprise count in Information Transmission, Software and IT Services": 141,
+    "difference": -139.0
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Gao Yin Ze Tong Pi Fa Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: SZSE-listed foreign-funded enterprise count in Guangdong Wholesale and Retail = 2",
+    "Extracted from company_profile.csv: industry of Lang Ji Hui Ruan Technology Co., Ltd. = Information Transmission, Software and IT Services",
+    "Extracted from national_industry_status.csv: SSE-listed enterprise count in Information Transmission, Software and IT Services = 141",
+    "Calculated difference: 2 - 141 = -139.0"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "f9807af7-8cc0-4d23-9032-70752be13a89"
+    },
+    {
+      "company_profile": "a9b797ac-f74d-4a35-acfa-01b494fe2b3f"
+    },
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "e9d63626-4106-4f18-8022-b5396454e690"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium052_result.json b/assets/qa_raw/enterprise_industry_analysis/medium052_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..374f6c8351a6dbadf725e4e29661981ebcfd1305
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium052_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium052",
+  "question": "Comparing the minimum value of State Natural Science Awards in the corresponding industry of the province where Gaoyin Zetong Wholesale Co., Ltd. is located with the minimum value of State Natural Science Awards in the industry of Langji Huiruan Technology Co., Ltd., which value is larger?",
+  "guidelines": "The answer must be a company name, \"industry\", or \"Equal\". Output only the name or \"Equal\" without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province where Gaoyin Zetong Wholesale Co., Ltd. is located: Guangdong Province",
+    "Minimum value of State Natural Science Awards in Guangdong wholesale and retail industry: 0",
+    "Industry of Langji Huiruan Technology Co., Ltd.: Information Transmission, Software and Information Technology Services",
+    "Minimum value of State Natural Science Awards in Information Transmission, Software and Information Technology Services: 0"
+  ],
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province where Gaoyin Zetong Wholesale Co., Ltd. is located": "Guangdong Province",
+    "Minimum value of State Natural Science Awards in Guangdong wholesale and retail industry": 0,
+    "Industry of Langji Huiruan Technology Co., Ltd.": "Information Transmission, Software and Information Technology Services",
+    "Minimum value of State Natural Science Awards in Information Transmission, Software and Information Technology Services": 0,
+    "Comparison conclusion": "Equal"
+  },
+  "steps": [
+    "Extracted from company_profile.csv that Gaoyin Zetong Wholesale Co., Ltd. is located in Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the minimum value of State Natural Science Awards in Guangdong wholesale and retail industry is 0.",
+    "Extracted from company_profile.csv that the industry of Langji Huiruan Technology Co., Ltd. is Information Transmission, Software and Information Technology Services.",
+    "Extracted from national_industry_status.csv that the minimum value of State Natural Science Awards in this industry is 0.",
+    "Both values are 0, so neither side is larger; output \"Equal\"."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "f9807af7-8cc0-4d23-9032-70752be13a89"
+    },
+    {
+      "company_profile": "a9b797ac-f74d-4a35-acfa-01b494fe2b3f"
+    },
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "e9d63626-4106-4f18-8022-b5396454e690"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium053_result.json b/assets/qa_raw/enterprise_industry_analysis/medium053_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ff62ea68d438120708aeadbba567cf8d96ae4fb
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium053_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium053",
+  "question": "What is the difference between the number of HKEX-listed private enterprises in the corresponding industry of the province where Zhangqiao Jinchuang Technology Co., Ltd. is located and the number of HKEX-listed enterprises in the corresponding industry of the province where Zhaoye Huachang Real Estate Development Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province where Zhangqiao Jinchuang Technology Co., Ltd. is located: Guangdong Province",
+    "Guangdong consumer electronics and electrical industry, number of HKEX-listed private enterprises: 27",
+    "Province where Zhaoye Huachang Real Estate Development Co., Ltd. is located: Guangdong Province",
+    "Guangdong real estate industry, number of HKEX-listed enterprises: 38"
+  ],
+  "answer": -11.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province where Zhangqiao Jinchuang Technology Co., Ltd. is located": "Guangdong Province",
+    "Number of HKEX-listed private enterprises in Guangdong consumer electronics and electrical industry": 27,
+    "Province where Zhaoye Huachang Real Estate Development Co., Ltd. is located": "Guangdong Province",
+    "Number of HKEX-listed enterprises in Guangdong real estate industry": 38,
+    "Difference": -11.0
+  },
+  "steps": [
+    "Extracted from company_profile.csv that Zhangqiao Jinchuang Technology Co., Ltd. is located in Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the number of HKEX-listed private enterprises in Guangdong consumer electronics and electrical industry is 27.",
+    "Extracted from company_profile.csv that Zhaoye Huachang Real Estate Development Co., Ltd. is located in Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the number of HKEX-listed enterprises in Guangdong real estate industry is 38.",
+    "Calculated the difference (former minus latter): 27 - 38 = -11.0."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "c17be31b-e2ca-440d-ac53-1d0061273700"
+    },
+    {
+      "company_profile": "58eb73a2-c225-4c83-9324-b515fec3a76f"
+    },
+    {
+      "regional_industry_status": "1e971ba1-a0ba-4221-908d-a308281f06d3"
+    },
+    {
+      "company_profile": "3f2c07ea-c142-43aa-ab56-ee76d8aaa4e5"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium054_result.json b/assets/qa_raw/enterprise_industry_analysis/medium054_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..acf3c616d47c10112fdb7294533558f48efd2bc6
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium054_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium054",
+  "question": "Which is higher: the minimum R&D expenditure ratio in the corresponding industry of the province where Zhangqiao Jinchuang Technology Co., Ltd. is located, or the minimum R&D expenditure ratio in the corresponding industry of the province where Zhaoye Huachang Real Estate Development Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province where Zhangqiao Jinchuang Technology Co., Ltd. is located: Guangdong Province",
+    "Minimum R&D expenditure ratio in Guangdong consumer electronics and electrical industry: 1.3 %",
+    "Province where Zhaoye Huachang Real Estate Development Co., Ltd. is located: Guangdong Province",
+    "Minimum R&D expenditure ratio in Guangdong real estate industry: 0 %"
+  ],
+  "answer": "Zhangqiao Jinchuang Technology Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province where Zhangqiao Jinchuang Technology Co., Ltd. is located": "Guangdong Province",
+    "Minimum R&D expenditure ratio in Guangdong consumer electronics and electrical industry": 1.3,
+    "Province where Zhaoye Huachang Real Estate Development Co., Ltd. is located": "Guangdong Province",
+    "Minimum R&D expenditure ratio in Guangdong real estate industry": 0,
+    "Comparison conclusion (higher value)": "Zhangqiao Jinchuang Technology Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv that Zhangqiao Jinchuang Technology Co., Ltd. is located in Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the minimum R&D expenditure ratio in Guangdong consumer electronics and electrical industry is 1.3 %.",
+    "Extracted from company_profile.csv that Zhaoye Huachang Real Estate Development Co., Ltd. is located in Guangdong Province.",
+    "Extracted from regional_industry_status.csv that the minimum R&D expenditure ratio in Guangdong real estate industry is 0 %.",
+    "Compared 1.3 % and 0 %; 1.3 % is higher, so output \"Zhangqiao Jinchuang Technology Co., Ltd.\"."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "c17be31b-e2ca-440d-ac53-1d0061273700"
+    },
+    {
+      "company_profile": "58eb73a2-c225-4c83-9324-b515fec3a76f"
+    },
+    {
+      "regional_industry_status": "1e971ba1-a0ba-4221-908d-a308281f06d3"
+    },
+    {
+      "company_profile": "3f2c07ea-c142-43aa-ab56-ee76d8aaa4e5"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium055_result.json b/assets/qa_raw/enterprise_industry_analysis/medium055_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae5e7463c60c0428d70185c72dbfe696554fefc8
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium055_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium055",
+  "question": "Between the median annual number of PCT invention patent applications for the corresponding industry in the province where Jinzhi Hongsheng Asset Management Company is located and that in the province where Zhonghai Gongchangjin Architectural Design Company is located, which is higher?",
+  "guidelines": "The answer must be either a company name or the word \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Jinzhi Hongsheng Asset Management Company: Shanghai",
+    "Median annual number of PCT invention patent applications in Shanghai for the financial industry: 2",
+    "Province of Zhonghai Gongchangjin Architectural Design Company: Guangdong Province",
+    "Median annual number of PCT invention patent applications in Guangdong Province for the construction industry: 0"
+  ],
+  "answer": "Jinzhi Hongsheng Asset Management Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Jinzhi Hongsheng Asset Management Company": "Shanghai",
+    "Median annual number of PCT invention patent applications in Shanghai for the financial industry": 2,
+    "Province of Zhonghai Gongchangjin Architectural Design Company": "Guangdong Province",
+    "Median annual number of PCT invention patent applications in Guangdong Province for the construction industry": 0,
+    "Comparison result (higher one)": "Jinzhi Hongsheng Asset Management Company"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Jinzhi Hongsheng Asset Management Company is located in Shanghai.",
+    "Extract from regional_industry_status.csv that the median annual number of PCT invention patent applications in Shanghai for the financial industry is 2.",
+    "Extract from company_profile.csv that Zhonghai Gongchangjin Architectural Design Company is located in Guangdong Province.",
+    "Extract from regional_industry_status.csv that the median annual number of PCT invention patent applications in Guangdong Province for the construction industry is 0.",
+    "Compare 2 and 0; 2 is higher, so output \"Jinzhi Hongsheng Asset Management Company\"."
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "6d7fb68b-9d0a-4cfb-bab2-abac83115eed"
+    },
+    {
+      "company_profile": "62612789-d0a3-4b19-bbee-abe8a89ab1bd"
+    },
+    {
+      "regional_industry_status": "3b977bc5-a654-409d-a92b-b463231c766a"
+    },
+    {
+      "company_profile": "f3608099-34f7-49a3-b2f5-0781705eeafc"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium056_result.json b/assets/qa_raw/enterprise_industry_analysis/medium056_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3bb469be2f48284a5db4e1bba20bc1a4e57d5e60
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium056_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium056",
+  "question": "Is the total liabilities (total) for the industry corresponding to the province where Jin Zhi Hong Sheng Zi Chan Management Co., Ltd. is located higher than the total liabilities (total) for the industry corresponding to the province where Zhong Hai Gong Chang Jin Jian Zhu She Ji Co., Ltd. is located?",
+  "guidelines": "The answer must be \"Yes\" or \"No\", output only one word, without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Jin Zhi Hong Sheng Zi Chan Management Co., Ltd.: Shanghai Municipality",
+    "Shanghai Municipality Financial Services Total liabilities (total): 44405835055827.7 Yuan",
+    "Province of Zhong Hai Gong Chang Jin Jian Zhu She Ji Co., Ltd.: Guangdong Province",
+    "Guangdong Province Construction Total liabilities (total): 247639719094.03 Yuan"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Jin Zhi Hong Sheng Zi Chan Management Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Financial Services Total liabilities (total) (Yuan)": 44405835055827.7,
+    "Province of Zhong Hai Gong Chang Jin Jian Zhu She Ji Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Construction Total liabilities (total) (Yuan)": 247639719094.03,
+    "Whether higher than": "Yes"
+  },
+  "steps": [
+    "Extract from company_profile.csv that the province of Jin Zhi Hong Sheng Zi Chan Management Co., Ltd. is Shanghai Municipality",
+    "Extract from regional_industry_status.csv that Shanghai Municipality Financial Services total liabilities (total) is 44405835055827.7 Yuan",
+    "Extract from company_profile.csv that the province of Zhong Hai Gong Chang Jin Jian Zhu She Ji Co., Ltd. is Guangdong Province",
+    "Extract from regional_industry_status.csv that Guangdong Province Construction total liabilities (total) is 247639719094.03 Yuan",
+    "Determine whether the former is higher than the latter; conclusion is \"Yes\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "6d7fb68b-9d0a-4cfb-bab2-abac83115eed"
+    },
+    {
+      "company_profile": "62612789-d0a3-4b19-bbee-abe8a89ab1bd"
+    },
+    {
+      "regional_industry_status": "3b977bc5-a654-409d-a92b-b463231c766a"
+    },
+    {
+      "company_profile": "f3608099-34f7-49a3-b2f5-0781705eeafc"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium057_result.json b/assets/qa_raw/enterprise_industry_analysis/medium057_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5d655a1ded61a77db25453adb56db7264bcea70c
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium057_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium057",
+  "question": "Is the maximum State Science and Technology Progress Award value in the corresponding industry of Shanghai, where Lang Ji Hui Ruan Technology Co., Ltd. is located, the same as the corresponding value in Beijing, where Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. is located?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.: Shanghai Municipality",
+    "Shanghai Municipality Information Transmission, Software and IT Services State Science and Technology Progress Award (maximum): 0",
+    "Province of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd.: Beijing Municipality",
+    "Beijing Municipality Health and Social Work State Science and Technology Progress Award (maximum)：0 units"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Information Transmission, Software and IT Services State Science and Technology Progress Award (maximum)": 0,
+    "Province of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd.": "Beijing Municipality",
+    "Beijing Municipality Health and Social Work State Science and Technology Progress Award (maximum)": 0,
+    "Whether identical": "Yes"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Lang Ji Hui Ruan Technology Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: maximum State Science and Technology Progress Award in Shanghai Information Transmission, Software and IT Services = 0",
+    "Extracted from company_profile.csv: province of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. = Beijing Municipality",
+    "Extracted from regional_industry_status.csv: maximum State Science and Technology Progress Award in Beijing Health and Social Work = 0",
+    "Both values are 0, so the answer is \"Yes\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "be71aa1d-8666-44e9-8f26-ffd92dc87001"
+    },
+    {
+      "company_profile": "e9d63626-4106-4f18-8022-b5396454e690"
+    },
+    {
+      "regional_industry_status": "20e7aa24-290d-47e5-bc90-5e8b8e8ab3ef"
+    },
+    {
+      "company_profile": "10890d3a-b03b-417b-b39e-f5070604b51e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium058_result.json b/assets/qa_raw/enterprise_industry_analysis/medium058_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..39fe80370a38f1cfaaf2c55bff0ac0822e5b3f67
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium058_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium058",
+  "question": "By how much is the mean capitalized R&D expenditure in the corresponding industry of Shanghai, where Lang Ji Hui Ruan Technology Co., Ltd. is located, higher than the corresponding value in Beijing, where Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. is located?",
+  "guidelines": "The answer must be an exact number, preserving all significant decimal places. Output only the number without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.: Shanghai Municipality",
+    "Shanghai Municipality Information Transmission, Software and IT Services Capitalized R&D expenditure (mean): 85678136.92375 Yuan",
+    "Province of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd.: Beijing Municipality",
+    "Beijing Municipality Health and Social Work Capitalized R&D expenditure (mean)：0 Yuan"
+  ],
+  "answer": 85678136.92375,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Information Transmission, Software and IT Services Capitalized R&D expenditure (mean) (Yuan)": 85678136.92375,
+    "Province of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd.": "Beijing Municipality",
+    "Beijing Municipality Health and Social Work Capitalized R&D expenditure (mean)（Yuan）": 0,
+    "Amount higher": 85678136.92375
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Lang Ji Hui Ruan Technology Co., Ltd. = Shanghai Municipality",
+    "Extracted from regional_industry_status.csv: mean capitalized R&D expenditure in Shanghai Information Transmission, Software and IT Services = 85678136.92375 Yuan",
+    "Extracted from company_profile.csv: province of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. = Beijing Municipality",
+    "Extracted from regional_industry_status.csv: mean capitalized R&D expenditure in Beijing Health and Social Work = 0 Yuan",
+    "Calculated the amount higher: 85678136.92375 - 0 = 85678136.92375"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "be71aa1d-8666-44e9-8f26-ffd92dc87001"
+    },
+    {
+      "company_profile": "e9d63626-4106-4f18-8022-b5396454e690"
+    },
+    {
+      "regional_industry_status": "20e7aa24-290d-47e5-bc90-5e8b8e8ab3ef"
+    },
+    {
+      "company_profile": "10890d3a-b03b-417b-b39e-f5070604b51e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium059_result.json b/assets/qa_raw/enterprise_industry_analysis/medium059_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f276571bae7edae9cdfef97e0997cfdfa784594
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium059_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium059",
+  "question": "Which is larger: the HKEX-listed private enterprise count in the corresponding industry of Guangdong, where Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. is located, or the total HKEX-listed enterprise count in the corresponding industry of Guangdong, where Zhao Ye Ze Jin Di Chan Holdings Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.: Guangdong Province",
+    "Guangdong Province Real Estate private enterprise HKEX-listed count: 26",
+    "Province of Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.: Guangdong Province",
+    "Total HKEX-listed enterprise count in Guangdong Real Estate: 38"
+  ],
+  "answer": "Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Real Estate private enterprise HKEX-listed count": 26,
+    "Province of Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Real Estate HKEX-listed enterprise count": 38,
+    "Comparison result (larger)": "Zhao Ye Ze Jin Di Chan Holdings Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: HKEX-listed private enterprise count in Guangdong Real Estate = 26",
+    "Extracted from company_profile.csv: province of Zhao Ye Ze Jin Di Chan Holdings Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: total HKEX-listed enterprise count in Guangdong Real Estate = 38",
+    "Compared 26 and 38; since 38 is larger, output \"Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "1e971ba1-a0ba-4221-908d-a308281f06d3"
+    },
+    {
+      "company_profile": "3f2c07ea-c142-43aa-ab56-ee76d8aaa4e5"
+    },
+    {
+      "regional_industry_status": "1e971ba1-a0ba-4221-908d-a308281f06d3"
+    },
+    {
+      "company_profile": "df986c37-1dd4-4c6e-a419-04cc2fd2a9ca"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium060_result.json b/assets/qa_raw/enterprise_industry_analysis/medium060_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3bd5e165cbaa5daca9d926fd1b3cdbd27db5bb1f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium060_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium060",
+  "question": "Are the minimum values of the Provincial or Ministerial Science and Technology Progress Award metric the same between the corresponding industries in Guangdong for Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. and Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.: Guangdong Province",
+    "Minimum Provincial or Ministerial Science and Technology Progress Award value in Guangdong Real Estate: 0",
+    "Province of Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.: Guangdong Province",
+    "Minimum Provincial or Ministerial Science and Technology Progress Award value in Guangdong Real Estate (comparison side): 0"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.": "Guangdong Province",
+    "Minimum Provincial or Ministerial Science and Technology Progress Award value in Guangdong Real Estate": 0,
+    "Province of Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.": "Guangdong Province",
+    "Minimum Provincial or Ministerial Science and Technology Progress Award value in Guangdong Real Estate (comparison side)": 0,
+    "Whether identical": "Yes"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: minimum Provincial or Ministerial Science and Technology Progress Award value in Guangdong Real Estate = 0",
+    "Extracted from company_profile.csv: province of Zhao Ye Ze Jin Di Chan Holdings Co., Ltd. = Guangdong Province",
+    "Extracted from regional_industry_status.csv: minimum Provincial or Ministerial Science and Technology Progress Award value in Guangdong Real Estate = 0",
+    "Both values are 0, so the answer is \"Yes\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "1e971ba1-a0ba-4221-908d-a308281f06d3"
+    },
+    {
+      "company_profile": "3f2c07ea-c142-43aa-ab56-ee76d8aaa4e5"
+    },
+    {
+      "regional_industry_status": "1e971ba1-a0ba-4221-908d-a308281f06d3"
+    },
+    {
+      "company_profile": "df986c37-1dd4-4c6e-a419-04cc2fd2a9ca"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium061_result.json b/assets/qa_raw/enterprise_industry_analysis/medium061_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4624f1111b83ccfd67a07a9df05fd5b52895fd00
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium061_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium061",
+  "question": "Are the total State Natural Science Award values the same between the corresponding industry in Hong Kong, where Rui Xing Jian Kang Zhi Yao Co., Ltd. is located, and the corresponding industry in Zhejiang, where Wu Li Hui Da Chain Co., Ltd. is located?",
+  "guidelines": "The answer must be \"Yes\" or \"No\". Output only one word without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Rui Xing Jian Kang Zhi Yao Co., Ltd.: Hong Kong SAR",
+    "Total State Natural Science Award value in Hong Kong Pharmaceutical Manufacturing: 0",
+    "Province of Wu Li Hui Da Chain Co., Ltd.: Zhejiang Province",
+    "Total State Natural Science Award value in Zhejiang Wholesale and Retail: 0"
+  ],
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Rui Xing Jian Kang Zhi Yao Co., Ltd.": "Hong Kong SAR",
+    "Total State Natural Science Award value in Hong Kong Pharmaceutical Manufacturing": 0,
+    "Province of Wu Li Hui Da Chain Co., Ltd.": "Zhejiang Province",
+    "Total State Natural Science Award value in Zhejiang Wholesale and Retail": 0,
+    "Whether identical": "Yes"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: province of Rui Xing Jian Kang Zhi Yao Co., Ltd. = Hong Kong SAR",
+    "Extracted from regional_industry_status.csv: total State Natural Science Award value in Hong Kong Pharmaceutical Manufacturing = 0",
+    "Extracted from company_profile.csv: province of Wu Li Hui Da Chain Co., Ltd. = Zhejiang Province",
+    "Extracted from regional_industry_status.csv: total State Natural Science Award value in Zhejiang Wholesale and Retail = 0",
+    "Both values are 0, so the answer is \"Yes\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "regional_industry_status": "481cddd5-7d34-4acc-ae6b-1a8e7de7444b"
+    },
+    {
+      "company_profile": "4c3a2a9c-278d-458c-8242-de58a40d50a1"
+    },
+    {
+      "regional_industry_status": "121a55c3-2823-410f-b764-12733c5e8754"
+    },
+    {
+      "company_profile": "24d73db3-b3be-4f9d-87e8-a8f7bcd7b0f9"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium062_result.json b/assets/qa_raw/enterprise_industry_analysis/medium062_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7da67948e68caf239601b0dc78f7af372573ecb4
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium062_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium062",
+  "question": "Rui Xing Jian Kang Zhi Yao Co., Ltd.industry in its province R&D headcount YoY change (minimum) and Wu Li Hui Da Chain Co., Ltd.industry in its province R&D headcount YoY change (minimum)compared with difference how much？",
+  "guidelines": "The answer must units, .Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "evidence": [
+    "Rui Xing Jian Kang Zhi Yao Co., Ltd.province：Hong Kong SAR",
+    "Hong Kong SAR-：R&D headcount YoY change (minimum) -33.3 %",
+    "Wu Li Hui Da Chain Co., Ltd.province：Zhejiang Province",
+    "Zhejiang Province-Retail：R&D headcount YoY change (minimum) -14.11 %"
+  ],
+  "steps": [
+    "Extracted from company_profile.csv: Rui Xing Jian Kang Zhi Yao Co., Ltd.province = Hong Kong SAR, industry=",
+    "regional_industry_status.csv in extractHong Kong SAR- R&D headcount YoY change (minimum)= -33.3 %",
+    "Extracted from company_profile.csv: Wu Li Hui Da Chain Co., Ltd.province = Zhejiang Province, industry=Retail",
+    "regional_industry_status.csv in extractZhejiang Province-Retail R&D headcount YoY change (minimum)= -14.11 %",
+    "Calculate difference: -33.3 - (-14.11) = -19.19"
+  ],
+  "steps_num": 5,
+  "answer": -19.19,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Rui Xing Jian Kang Zhi Yao Co., Ltd.province": "Hong Kong SAR",
+    "Hong Kong SAR  R&D headcount YoY change (minimum)": -33.3,
+    "Wu Li Hui Da Chain Co., Ltd.province": "Zhejiang Province",
+    "Zhejiang Province Retail R&D headcount YoY change (minimum)": -14.11,
+    "difference": -19.19
+  },
+  "reference": [
+    {
+      "regional_industry_status": "481cddd5-7d34-4acc-ae6b-1a8e7de7444b"
+    },
+    {
+      "company_profile": "4c3a2a9c-278d-458c-8242-de58a40d50a1"
+    },
+    {
+      "regional_industry_status": "121a55c3-2823-410f-b764-12733c5e8754"
+    },
+    {
+      "company_profile": "24d73db3-b3be-4f9d-87e8-a8f7bcd7b0f9"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium063_result.json b/assets/qa_raw/enterprise_industry_analysis/medium063_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0974a233def418ccc762de7deb52833b1af52617
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium063_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium063",
+  "question": "What is the difference between the number of central state-owned enterprises listed on the Shenzhen Stock Exchange in the corresponding industry of the province where Baoxin Huihui Network Company is located and the number of central state-owned enterprises in the metal smelting and rolling industry listed on the Hong Kong Stock Exchange?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Baoxin Huihui Network Company: Beijing Municipality",
+    "Beijing Municipality – Information Transmission, Software and IT Services: number of central state-owned enterprises listed on the Shenzhen Stock Exchange = 4",
+    "Metal smelting and rolling processing industry: number of central state-owned enterprises listed on the Hong Kong Stock Exchange = 6"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Baoxin Huihui Network Company is located in Beijing Municipality and its corresponding industry is Information Transmission, Software and IT Services.",
+    "Extract from regional_industry_status.csv that the number of central state-owned enterprises in Beijing Municipality – Information Transmission, Software and IT Services that are listed on the Shenzhen Stock Exchange is 4.",
+    "Extract from national_industry_status.csv that the number of central state-owned enterprises in the metal smelting and rolling processing industry that are listed on the Hong Kong Stock Exchange is 6.",
+    "Calculate the difference: 4 - 6 = -2.0."
+  ],
+  "steps_num": 4,
+  "answer": -2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Baoxin Huihui Network Company": "Beijing Municipality",
+    "Number of central state-owned enterprises listed on the Shenzhen Stock Exchange in Beijing Municipality – Information Transmission, Software and IT Services": 4,
+    "Number of central state-owned enterprises listed on the Hong Kong Stock Exchange in the metal smelting and rolling processing industry": 6,
+    "Difference": -2.0
+  },
+  "reference": [
+    {
+      "regional_industry_status": "e7d170ae-b276-465b-b737-971c81e11168"
+    },
+    {
+      "company_profile": "fe635f7d-bee1-4e2e-9ed6-3923871590d0"
+    },
+    {
+      "national_industry_status": "e12e3329-0ab1-4bf1-b19d-0a5c9610b5ce"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium064_result.json b/assets/qa_raw/enterprise_industry_analysis/medium064_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a4210713e4bccf69a3fbb5654d1c703953b9a831
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium064_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium064",
+  "question": "Between the median R&D expenditure ratio of the corresponding industry in the province where Baoxin Huihui Network Company is located and that of the metal smelting and rolling processing industry, which is higher?",
+  "guidelines": "The answer must be either a company name or the word \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Baoxin Huihui Network Company: Beijing Municipality",
+    "Beijing Municipality – Information Transmission, Software and IT Services: median R&D expenditure ratio 12.16 %",
+    "Metal smelting and rolling processing industry: median R&D expenditure ratio 3.03 %"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Baoxin Huihui Network Company is located in Beijing Municipality and its corresponding industry is Information Transmission, Software and IT Services.",
+    "Extract from regional_industry_status.csv that the median R&D expenditure ratio for Information Transmission, Software and IT Services in Beijing Municipality is 12.16 %.",
+    "Extract from national_industry_status.csv that the median R&D expenditure ratio for the metal smelting and rolling processing industry is 3.03 %.",
+    "Compare the two medians: 12.16 % > 3.03 %, so the corresponding industry in the province where Baoxin Huihui Network Company is located is higher."
+  ],
+  "steps_num": 4,
+  "answer": "Baoxin Huihui Network Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Baoxin Huihui Network Company": "Beijing Municipality",
+    "Median R&D expenditure ratio for Information Transmission, Software and IT Services in Beijing Municipality": 12.16,
+    "Median R&D expenditure ratio for the metal smelting and rolling processing industry": 3.03,
+    "Comparison result (higher one)": "Baoxin Huihui Network Company"
+  },
+  "reference": [
+    {
+      "regional_industry_status": "e7d170ae-b276-465b-b737-971c81e11168"
+    },
+    {
+      "company_profile": "fe635f7d-bee1-4e2e-9ed6-3923871590d0"
+    },
+    {
+      "national_industry_status": "e12e3329-0ab1-4bf1-b19d-0a5c9610b5ce"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium065_result.json b/assets/qa_raw/enterprise_industry_analysis/medium065_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..8ddf907f1c56d726948d8a9fe7b82d26a0006180
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium065_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium065",
+  "question": "Between the number of Shanghai Stock Exchange-listed private enterprises in the corresponding industry of the province where Beikong Zejing Water Company is located and the number of Beijing Stock Exchange-listed enterprises in the Information Transmission, Software and IT Services industry, which is larger?",
+  "guidelines": "The answer must be either \"the number of SSE-listed private enterprises in the corresponding industry of the province where Beikong Zejing Water Company is located\" or \"the number of BSE-listed enterprises in the Information Transmission, Software and IT Services industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Beikong Zejing Water Company: Guangdong Province",
+    "Guangdong Province – Water Conservancy, Environment and Public Facilities Management: number of SSE-listed private enterprises = 1",
+    "Information Transmission, Software and IT Services: number of BSE-listed enterprises = 17"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Beikong Zejing Water Company is located in Guangdong Province and its corresponding industry is Water Conservancy, Environment and Public Facilities Management.",
+    "Extract from regional_industry_status.csv that the number of SSE-listed private enterprises in Guangdong Province – Water Conservancy, Environment and Public Facilities Management is 1.",
+    "Extract from national_industry_status.csv that the number of BSE-listed enterprises in the Information Transmission, Software and IT Services industry is 17.",
+    "Compare the counts: 17 > 1, so the number of BSE-listed enterprises in the Information Transmission, Software and IT Services industry is larger."
+  ],
+  "steps_num": 4,
+  "answer": "the number of BSE-listed enterprises in the Information Transmission, Software and IT Services industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Beikong Zejing Water Company": "Guangdong Province",
+    "Number of SSE-listed private enterprises in Guangdong Province – Water Conservancy, Environment and Public Facilities Management": 1,
+    "Number of BSE-listed enterprises in the Information Transmission, Software and IT Services industry": 17,
+    "Comparison result (larger one)": "the number of BSE-listed enterprises in the Information Transmission, Software and IT Services industry"
+  },
+  "reference": [
+    {
+      "regional_industry_status": "0e750378-8b90-47a3-8828-1b5f5ea324da"
+    },
+    {
+      "company_profile": "b32ab5eb-bfc5-4320-8527-8c31fd09ac36"
+    },
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium066_result.json b/assets/qa_raw/enterprise_industry_analysis/medium066_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..122ae8ceab9204d71f0dff9a3494e6793824ec22
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium066_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium066",
+  "question": "Which is larger: the total assets of the corresponding industry in the province where Beikong Zejing Water Company is located, or the total assets of the Information Transmission, Software and IT Services industry?",
+  "guidelines": "The answer must be either \"the total assets of the corresponding industry in the province where Beikong Zejing Water Company is located\" or \"the Information Transmission, Software and IT Services industry\". Output only the name, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Beikong Zejing Water Company: Guangdong Province",
+    "Guangdong Province – Water Conservancy, Environment and Public Facilities Management: total assets = 194502203663.44 Yuan",
+    "Information Transmission, Software and IT Services: total assets = 18848109950318 Yuan"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Beikong Zejing Water Company is located in Guangdong Province and its corresponding industry is Water Conservancy, Environment and Public Facilities Management.",
+    "Extract from regional_industry_status.csv that the total assets of Water Conservancy, Environment and Public Facilities Management in Guangdong Province are 194502203663.44 Yuan.",
+    "Extract from national_industry_status.csv that the total assets of the Information Transmission, Software and IT Services industry are 18848109950318 Yuan.",
+    "Compare the total assets: 18848109950318 > 194502203663.44, so the Information Transmission, Software and IT Services industry is larger."
+  ],
+  "steps_num": 4,
+  "answer": "the Information Transmission, Software and IT Services industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Beikong Zejing Water Company": "Guangdong Province",
+    "Total assets of Water Conservancy, Environment and Public Facilities Management in Guangdong Province (Yuan)": 194502203663.44,
+    "Total assets of the Information Transmission, Software and IT Services industry (Yuan)": 18848109950318,
+    "Comparison result (larger one)": "the Information Transmission, Software and IT Services industry"
+  },
+  "reference": [
+    {
+      "regional_industry_status": "0e750378-8b90-47a3-8828-1b5f5ea324da"
+    },
+    {
+      "company_profile": "b32ab5eb-bfc5-4320-8527-8c31fd09ac36"
+    },
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium067_result.json b/assets/qa_raw/enterprise_industry_analysis/medium067_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d83d032627122bd06844325ffb371d4f686a99ed
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium067_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium067",
+  "question": "Between the number of HKEX-listed foreign-funded enterprises in the corresponding industry of the province where Aijian Yikang Fuzhongxin Company is located and the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry, which value is larger?",
+  "guidelines": "The answer must be either \"the number of HKEX-listed foreign-funded enterprises in the corresponding industry of the province where Aijian Yikang Fuzhongxin Company is located\" or \"the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Aijian Yikang Fuzhongxin Company: Beijing Municipality",
+    "Beijing Municipality – Health and Social Work: number of HKEX-listed foreign-funded enterprises = 2",
+    "Pharmaceutical Manufacturing: number of SSE-listed central state-owned enterprises = 7"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Aijian Yikang Fuzhongxin Company is located in Beijing Municipality and its corresponding industry is Health and Social Work.",
+    "Extract from regional_industry_status.csv that the number of HKEX-listed foreign-funded enterprises in Beijing Municipality – Health and Social Work is 2.",
+    "Extract from national_industry_status.csv that the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry is 7.",
+    "Compare the counts: 7 > 2, so the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry is larger."
+  ],
+  "steps_num": 4,
+  "answer": "the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Aijian Yikang Fuzhongxin Company": "Beijing Municipality",
+    "Number of HKEX-listed foreign-funded enterprises in Beijing Municipality – Health and Social Work": 2,
+    "Number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry": 7,
+    "Comparison result (larger one)": "the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry"
+  },
+  "reference": [
+    {
+      "regional_industry_status": "20e7aa24-290d-47e5-bc90-5e8b8e8ab3ef"
+    },
+    {
+      "company_profile": "10890d3a-b03b-417b-b39e-f5070604b51e"
+    },
+    {
+      "national_industry_status": "9c71278e-97a3-4867-826d-e139f1dffc24"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium068_result.json b/assets/qa_raw/enterprise_industry_analysis/medium068_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0d9c3b832f88d010a403d0f8279b0005fc83bfe7
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium068_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium068",
+  "question": "What is the difference between the maximum annual number of China patent applications for the corresponding industry in the province where Aijian Yikang Fuzhongxin Company is located and the maximum annual number of China patent applications for the pharmaceutical manufacturing industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Aijian Yikang Fuzhongxin Company: Beijing Municipality",
+    "Beijing Municipality - Health and Social Work: maximum annual number of China patent applications = 0",
+    "Pharmaceutical manufacturing industry: maximum annual number of China patent applications = 329"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Aijian Yikang Fuzhongxin Company is located in Beijing Municipality and its corresponding industry is Health and Social Work.",
+    "Extract from regional_industry_status.csv that the maximum annual number of China patent applications in Beijing Municipality - Health and Social Work is 0.",
+    "Extract from national_industry_status.csv that the maximum annual number of China patent applications in the pharmaceutical manufacturing industry is 329.",
+    "Calculate the difference: 0 - 329 = -329.0."
+  ],
+  "steps_num": 4,
+  "answer": -329.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Aijian Yikang Fuzhongxin Company": "Beijing Municipality",
+    "Maximum annual number of China patent applications in Beijing Municipality - Health and Social Work": 0,
+    "Maximum annual number of China patent applications in the pharmaceutical manufacturing industry": 329,
+    "Difference": -329.0
+  },
+  "reference": [
+    {
+      "regional_industry_status": "20e7aa24-290d-47e5-bc90-5e8b8e8ab3ef"
+    },
+    {
+      "company_profile": "10890d3a-b03b-417b-b39e-f5070604b51e"
+    },
+    {
+      "national_industry_status": "9c71278e-97a3-4867-826d-e139f1dffc24"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium069_result.json b/assets/qa_raw/enterprise_industry_analysis/medium069_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b457a465280bff03239df2c2d569b854a9ed30ba
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium069_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium069",
+  "question": "Between the maximum capitalized R&D expenditure of the corresponding industry in the province where Zhongche Yuanze Shipbuilding Company is located and the maximum capitalized R&D expenditure of the Electricity, Heat, Gas and Water Production and Supply industry, which is higher?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Zhongche Yuanze Shipbuilding Company: Beijing Municipality",
+    "Beijing Municipality - Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing: maximum capitalized R&D expenditure = 353771542.85 Yuan",
+    "Electricity, Heat, Gas and Water Production and Supply industry: maximum capitalized R&D expenditure = 1982740987.15 Yuan"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Zhongche Yuanze Shipbuilding Company is located in Beijing Municipality and its corresponding industry is Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing.",
+    "Extract from regional_industry_status.csv that the maximum capitalized R&D expenditure in Beijing Municipality - Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing is 353771542.85 Yuan.",
+    "Extract from national_industry_status.csv that the maximum capitalized R&D expenditure in the Electricity, Heat, Gas and Water Production and Supply industry is 1982740987.15 Yuan.",
+    "Compare the two maximum values: 1982740987.15 Yuan > 353771542.85 Yuan, so the industry is higher."
+  ],
+  "steps_num": 4,
+  "answer": "industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Zhongche Yuanze Shipbuilding Company": "Beijing Municipality",
+    "Maximum capitalized R&D expenditure in Beijing Municipality - Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing (Yuan)": 353771542.85,
+    "Maximum capitalized R&D expenditure in the Electricity, Heat, Gas and Water Production and Supply industry (Yuan)": 1982740987.15,
+    "Comparison result (higher one)": "industry"
+  },
+  "reference": [
+    {
+      "regional_industry_status": "7df0da2e-6bbb-41f2-aeb7-90620f4964a8"
+    },
+    {
+      "company_profile": "b5c4bcce-8ecd-4a6e-b6ca-e5b6eb41fb8e"
+    },
+    {
+      "national_industry_status": "5aa7f8db-d709-4228-8ae0-6a216eb839bc"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium070_result.json b/assets/qa_raw/enterprise_industry_analysis/medium070_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c3b297230832879ede2d147220d75ffeefa7e65f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium070_result.json
@@ -0,0 +1,38 @@
+{
+  "id": "medium070",
+  "question": "What is the difference between the number of local state-owned enterprises listed on the Shanghai Stock Exchange in the corresponding industry in Beijing Municipality and the number of private enterprises listed on the Shanghai Stock Exchange in the Electricity, Heat, Gas and Water Production and Supply industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Beijing Municipality - Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing: number of SSE-listed local state-owned enterprises = 1",
+    "Electricity, Heat, Gas and Water Production and Supply industry: number of SSE-listed private enterprises = 19"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Zhongche Yuanze Shipbuilding Company is located in Beijing Municipality and its corresponding industry is Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing.",
+    "Extract from regional_industry_status.csv that the number of SSE-listed local state-owned enterprises in Beijing Municipality - Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing is 1.",
+    "Extract from national_industry_status.csv that the number of SSE-listed private enterprises in the Electricity, Heat, Gas and Water Production and Supply industry is 19.",
+    "Calculate the difference: 1 - 19 = -18.0."
+  ],
+  "steps_num": 4,
+  "answer": -18.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Number of SSE-listed local state-owned enterprises in Beijing Municipality - Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing": 1,
+    "Number of SSE-listed private enterprises in the Electricity, Heat, Gas and Water Production and Supply industry": 19,
+    "Difference": -18.0
+  },
+  "reference": [
+    {
+      "regional_industry_status": "7df0da2e-6bbb-41f2-aeb7-90620f4964a8"
+    },
+    {
+      "company_profile": "b5c4bcce-8ecd-4a6e-b6ca-e5b6eb41fb8e"
+    },
+    {
+      "national_industry_status": "5aa7f8db-d709-4228-8ae0-6a216eb839bc"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium071_result.json b/assets/qa_raw/enterprise_industry_analysis/medium071_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac3be267dd097d831077443d89bbfcce20a3e123
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium071_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium071",
+  "question": "Which is larger: the number of SSE-listed enterprises in the province where Biyuan Chanjin Real Estate Company is located, or the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry?",
+  "guidelines": "The answer must be either \"the number of SSE-listed enterprises in the province where Biyuan Chanjin Real Estate Company is located\" or \"the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Biyuan Chanjin Real Estate Company: Tianjin Municipality",
+    "Tianjin Municipality - Real Estate: number of SSE-listed enterprises = 3",
+    "Transportation, Storage and Postal Services: number of SZSE-listed local state-owned enterprises = 15"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Biyuan Chanjin Real Estate Company is located in Tianjin Municipality and its corresponding industry is Real Estate.",
+    "Extract from regional_industry_status.csv that the number of SSE-listed enterprises in Tianjin Municipality - Real Estate is 3.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry is 15.",
+    "Compare the counts: 15 > 3, so the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry is larger."
+  ],
+  "steps_num": 4,
+  "answer": "the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Biyuan Chanjin Real Estate Company": "Tianjin Municipality",
+    "Number of SSE-listed enterprises in Tianjin Municipality - Real Estate": 3,
+    "Number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry": 15,
+    "Comparison result (larger one)": "the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry"
+  },
+  "reference": [
+    {
+      "regional_industry_status": "f570645f-cb9b-4537-bbc4-39cfb7471a8c"
+    },
+    {
+      "company_profile": "7569b54a-60ea-4334-be2d-085b96a7730b"
+    },
+    {
+      "national_industry_status": "c5e98e13-e208-406f-aa43-9e029e2c4de6"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium072_result.json b/assets/qa_raw/enterprise_industry_analysis/medium072_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc034f48582584301700ea32a822d895c7210704
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium072_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium072",
+  "question": "Which value is larger: the total number of real estate enterprises in the province where Biyuan Chanjin Real Estate Company is located, or the number of SZSE-listed central state-owned enterprises in the Transportation, Storage and Postal Services industry?",
+  "guidelines": "The answer must be either \"the total number of real estate enterprises in the province where Biyuan Chanjin Real Estate Company is located\" or \"the number of SZSE-listed central state-owned enterprises in the Transportation, Storage and Postal Services industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Biyuan Chanjin Real Estate Company: Tianjin Municipality",
+    "Tianjin Municipality - Real Estate: total number of enterprises = 7",
+    "Transportation, Storage and Postal Services: number of SZSE-listed central state-owned enterprises = 5"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Biyuan Chanjin Real Estate Company is located in Tianjin Municipality and its corresponding industry is Real Estate.",
+    "Extract from regional_industry_status.csv that the total number of enterprises in Tianjin Municipality - Real Estate is 7.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed central state-owned enterprises in the Transportation, Storage and Postal Services industry is 5.",
+    "Compare the values: 7 > 5, so the total number of real estate enterprises in the province where Biyuan Chanjin Real Estate Company is located is larger."
+  ],
+  "steps_num": 4,
+  "answer": "the total number of real estate enterprises in the province where Biyuan Chanjin Real Estate Company is located",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Biyuan Chanjin Real Estate Company": "Tianjin Municipality",
+    "Total number of enterprises in Tianjin Municipality - Real Estate": 7,
+    "Number of SZSE-listed central state-owned enterprises in the Transportation, Storage and Postal Services industry": 5,
+    "Comparison result (larger one)": "the total number of real estate enterprises in the province where Biyuan Chanjin Real Estate Company is located"
+  },
+  "reference": [
+    {
+      "regional_industry_status": "f570645f-cb9b-4537-bbc4-39cfb7471a8c"
+    },
+    {
+      "company_profile": "7569b54a-60ea-4334-be2d-085b96a7730b"
+    },
+    {
+      "national_industry_status": "c5e98e13-e208-406f-aa43-9e029e2c4de6"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium073_result.json b/assets/qa_raw/enterprise_industry_analysis/medium073_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f3739901fc9d40a5441ccdf161b2f4bf40ff4ee9
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium073_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium073",
+  "question": "What is the difference between the number of SZSE-listed foreign-funded enterprises in the industry where Zhongke Keshu Software Company operates and the number of BSE-listed private enterprises in the construction industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhongke Keshu Software Company: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services: number of SZSE-listed foreign-funded enterprises = 4",
+    "Construction industry: number of BSE-listed private enterprises = 2"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Zhongke Keshu Software Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed foreign-funded enterprises in Information Transmission, Software and IT Services is 4.",
+    "Extract from national_industry_status.csv that the number of BSE-listed private enterprises in the construction industry is 2.",
+    "Calculate the difference: 4 - 2 = 2.0."
+  ],
+  "steps_num": 4,
+  "answer": 2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Zhongke Keshu Software Company": "Information Transmission, Software and IT Services",
+    "Number of SZSE-listed foreign-funded enterprises in Information Transmission, Software and IT Services": 4,
+    "Number of BSE-listed private enterprises in the construction industry": 2,
+    "Difference": 2.0
+  },
+  "reference": [
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "national_industry_status": "a3ea618e-b6c2-4ae4-a46d-7d4fafa9aff9"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium074_result.json b/assets/qa_raw/enterprise_industry_analysis/medium074_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a7b144bd77d948548dd6724a30c0bd420de2cbb4
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium074_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium074",
+  "question": "What is the difference between the number of HKEX-listed local state-owned enterprises in the industry where Zhongke Keshu Software Company operates and the number of HKEX-listed foreign-funded enterprises in the construction industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhongke Keshu Software Company: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services: number of HKEX-listed local state-owned enterprises = 9",
+    "Construction industry: number of HKEX-listed foreign-funded enterprises = 6"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Zhongke Keshu Software Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed local state-owned enterprises in Information Transmission, Software and IT Services is 9.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed foreign-funded enterprises in the construction industry is 6.",
+    "Calculate the difference: 9 - 6 = 3.0."
+  ],
+  "steps_num": 4,
+  "answer": 3.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Zhongke Keshu Software Company": "Information Transmission, Software and IT Services",
+    "Number of HKEX-listed local state-owned enterprises in Information Transmission, Software and IT Services": 9,
+    "Number of HKEX-listed foreign-funded enterprises in the construction industry": 6,
+    "Difference": 3.0
+  },
+  "reference": [
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "national_industry_status": "a3ea618e-b6c2-4ae4-a46d-7d4fafa9aff9"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium075_result.json b/assets/qa_raw/enterprise_industry_analysis/medium075_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3bc735ba9443ab148a189a417f7ef5e3dccdf787
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium075_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium075",
+  "question": "What is the difference between the number of SSE-listed foreign-funded enterprises in the industry where Hengli Kezhi Software Company operates and the number of SSE-listed private enterprises in the conglomerates industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hengli Kezhi Software Company: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services: number of SSE-listed foreign-funded enterprises = 7",
+    "Conglomerates: number of SSE-listed private enterprises = 5"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Hengli Kezhi Software Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the number of SSE-listed foreign-funded enterprises in Information Transmission, Software and IT Services is 7.",
+    "Extract from national_industry_status.csv that the number of SSE-listed private enterprises in the conglomerates industry is 5.",
+    "Calculate the difference: 7 - 5 = 2.0."
+  ],
+  "steps_num": 4,
+  "answer": 2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Hengli Kezhi Software Company": "Information Transmission, Software and IT Services",
+    "Number of SSE-listed foreign-funded enterprises in Information Transmission, Software and IT Services": 7,
+    "Number of SSE-listed private enterprises in the conglomerates industry": 5,
+    "Difference": 2.0
+  },
+  "reference": [
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "6c1bc3d3-0763-4686-bb01-a61e146eee63"
+    },
+    {
+      "national_industry_status": "fee7bf00-36e3-4f1a-a296-f5612ad8cedd"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium076_result.json b/assets/qa_raw/enterprise_industry_analysis/medium076_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..85cebaa42ebcc8950a7fa43fcebc0f4862b76c67
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium076_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium076",
+  "question": "Between the median number of industry standards participated in drafting in the industry where Hengli Kezhi Software Company operates and that of the conglomerates industry, which value is larger?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hengli Kezhi Software Company: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services: median number of industry standards participated in drafting = 2",
+    "Conglomerates: median number of industry standards participated in drafting = 0"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Hengli Kezhi Software Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the median number of industry standards participated in drafting for Information Transmission, Software and IT Services is 2.",
+    "Extract from national_industry_status.csv that the median number of industry standards participated in drafting for the conglomerates industry is 0.",
+    "Compare the two medians: 2 > 0, so the industry where Hengli Kezhi Software Company operates is larger."
+  ],
+  "steps_num": 4,
+  "answer": "Hengli Kezhi Software Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Hengli Kezhi Software Company": "Information Transmission, Software and IT Services",
+    "Median number of industry standards participated in drafting for Information Transmission, Software and IT Services": 2,
+    "Median number of industry standards participated in drafting for the conglomerates industry": 0,
+    "Comparison result (larger one)": "Hengli Kezhi Software Company"
+  },
+  "reference": [
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "6c1bc3d3-0763-4686-bb01-a61e146eee63"
+    },
+    {
+      "national_industry_status": "fee7bf00-36e3-4f1a-a296-f5612ad8cedd"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium077_result.json b/assets/qa_raw/enterprise_industry_analysis/medium077_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..595e009c4639b1962591d95d2ca0d8f84d6700ea
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium077_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium077",
+  "question": "Which is greater: the number of BSE-listed enterprises in the industry of Lang Ji Hui Ruan Technology Co., Ltd., or the number of enterprises in chemical raw materials and chemical products manufacturing?",
+  "guidelines": "The answer must be either \"Number of BSE-listed enterprises in the industry of Lang Ji Hui Ruan Technology Co., Ltd.\" or \"Number of enterprises in chemical raw materials and chemical products manufacturing\"; output only the designated answer text, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Lang Ji Hui Ruan Technology Co., Ltd. industry: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services: Number of BSE-listed enterprises 17",
+    "Chemical Raw Materials and Chemical Products Manufacturing: Number of enterprises 364"
+  ],
+  "steps": [
+    "From company_profile.csv, extract that Lang Ji Hui Ruan Technology Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "From national_industry_status.csv, extract that the number of BSE-listed enterprises for Information Transmission, Software and IT Services is 17",
+    "From national_industry_status.csv, extract that the number of enterprises in chemical raw materials and chemical products manufacturing is 364",
+    "Compare counts: 364 > 17; therefore the number of enterprises in chemical raw materials and chemical products manufacturing is greater"
+  ],
+  "steps_num": 4,
+  "answer": "Chemical Raw Materials and Chemical Products Manufacturing",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Lang Ji Hui Ruan Technology Co., Ltd. industry": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services Number of BSE-listed enterprises": 17,
+    "Chemical Raw Materials and Chemical Products Manufacturing Number of enterprises": 364,
+    "Comparison conclusion (greater)": "Chemical Raw Materials and Chemical Products Manufacturing"
+  },
+  "reference": [
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "e9d63626-4106-4f18-8022-b5396454e690"
+    },
+    {
+      "national_industry_status": "6162f9dd-dbef-46a3-a3fc-558b013d65bf"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium078_result.json b/assets/qa_raw/enterprise_industry_analysis/medium078_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..46c309f478bd21ea650853beff36be6d85654acd
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium078_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium078",
+  "question": "Between the mean change in R&D expenditure ratio for the industry where Langji Huiruan Technology Company operates and the same indicator for the Chemical Raw Materials and Chemical Products Manufacturing industry, which is higher?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Langji Huiruan Technology Company: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services: mean change in R&D expenditure ratio = 1.98752122241087 %",
+    "Chemical Raw Materials and Chemical Products Manufacturing: mean change in R&D expenditure ratio = -0.0206470588235294 %"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Langji Huiruan Technology Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the mean change in R&D expenditure ratio for Information Transmission, Software and IT Services is 1.98752122241087 %.",
+    "Extract from national_industry_status.csv that the mean change in R&D expenditure ratio for Chemical Raw Materials and Chemical Products Manufacturing is -0.0206470588235294 %.",
+    "Compare the means: 1.98752122241087 % > -0.0206470588235294 %, so the industry where Langji Huiruan Technology Company operates is higher."
+  ],
+  "steps_num": 4,
+  "answer": "Langji Huiruan Technology Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Langji Huiruan Technology Company": "Information Transmission, Software and IT Services",
+    "Mean change in R&D expenditure ratio for Information Transmission, Software and IT Services": 1.98752122241087,
+    "Mean change in R&D expenditure ratio for Chemical Raw Materials and Chemical Products Manufacturing": -0.0206470588235294,
+    "Comparison result (higher one)": "Langji Huiruan Technology Company"
+  },
+  "reference": [
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "e9d63626-4106-4f18-8022-b5396454e690"
+    },
+    {
+      "national_industry_status": "6162f9dd-dbef-46a3-a3fc-558b013d65bf"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium079_result.json b/assets/qa_raw/enterprise_industry_analysis/medium079_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..91ecfbaaad760e7778172ef42934e8b0f21de5c0
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium079_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium079",
+  "question": "What is the difference between the number of SZSE-listed central state-owned enterprises in the industry where Huijin Jinrui Wealth Management Company operates and the number of HKEX-listed central state-owned enterprises in the Electrical Machinery and Equipment Manufacturing industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Huijin Jinrui Wealth Management Company: Financial industry",
+    "Financial industry: number of SZSE-listed central state-owned enterprises = 5",
+    "Electrical Machinery and Equipment Manufacturing industry: number of HKEX-listed central state-owned enterprises = 3"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Huijin Jinrui Wealth Management Company belongs to the financial industry.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed central state-owned enterprises in the financial industry is 5.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed central state-owned enterprises in the Electrical Machinery and Equipment Manufacturing industry is 3.",
+    "Calculate the difference: 5 - 3 = 2.0."
+  ],
+  "steps_num": 4,
+  "answer": 2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Huijin Jinrui Wealth Management Company": "Financial industry",
+    "Number of SZSE-listed central state-owned enterprises in the financial industry": 5,
+    "Number of HKEX-listed central state-owned enterprises in the Electrical Machinery and Equipment Manufacturing industry": 3,
+    "Difference": 2.0
+  },
+  "reference": [
+    {
+      "national_industry_status": "5dc0d9db-af52-48e2-8eef-98a66f67632f"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    },
+    {
+      "national_industry_status": "15263bd4-fe7b-43f3-bd79-b697b63a42b7"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium080_result.json b/assets/qa_raw/enterprise_industry_analysis/medium080_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..9263fc847034079c2530ae5cf2600a05ac831a70
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium080_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium080",
+  "question": "Between the median net profit amount of Huijin Jinrui Wealth Management Company and that of the Electrical Machinery and Equipment Manufacturing industry, which is higher?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Huijin Jinrui Wealth Management Company: Financial industry",
+    "Financial industry: median net profit amount = 837989445.74 Yuan",
+    "Electrical Machinery and Equipment Manufacturing industry: median net profit amount = 117943061.895 Yuan"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Huijin Jinrui Wealth Management Company belongs to the financial industry.",
+    "Extract from national_industry_status.csv that the median net profit amount of the financial industry is 837989445.74 Yuan.",
+    "Extract from national_industry_status.csv that the median net profit amount of the Electrical Machinery and Equipment Manufacturing industry is 117943061.895 Yuan.",
+    "Compare the two medians: 837989445.74 Yuan > 117943061.895 Yuan, so Huijin Jinrui Wealth Management Company is higher."
+  ],
+  "steps_num": 4,
+  "answer": "Huijin Jinrui Wealth Management Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Huijin Jinrui Wealth Management Company": "Financial industry",
+    "Median net profit amount of the financial industry (Yuan)": 837989445.74,
+    "Median net profit amount of the Electrical Machinery and Equipment Manufacturing industry (Yuan)": 117943061.895,
+    "Comparison result (higher one)": "Huijin Jinrui Wealth Management Company"
+  },
+  "reference": [
+    {
+      "national_industry_status": "5dc0d9db-af52-48e2-8eef-98a66f67632f"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    },
+    {
+      "national_industry_status": "15263bd4-fe7b-43f3-bd79-b697b63a42b7"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium081_result.json b/assets/qa_raw/enterprise_industry_analysis/medium081_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b35c113cb5cddda9aab433f2af9e36530d334b57
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium081_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium081",
+  "question": "What is the difference between the number of SZSE-listed enterprises in the industry where Zhaoye Huachang Real Estate Development Company operates and the number of SZSE-listed private enterprises in the General Equipment Manufacturing industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhaoye Huachang Real Estate Development Company: Real Estate",
+    "Real Estate: number of SZSE-listed enterprises = 51",
+    "General Equipment Manufacturing: number of SZSE-listed private enterprises = 86"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Zhaoye Huachang Real Estate Development Company belongs to the Real Estate industry.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed enterprises in the Real Estate industry is 51.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed private enterprises in the General Equipment Manufacturing industry is 86.",
+    "Calculate the difference: 51 - 86 = -35.0."
+  ],
+  "steps_num": 4,
+  "answer": -35.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Zhaoye Huachang Real Estate Development Company": "Real Estate",
+    "Number of SZSE-listed enterprises in the Real Estate industry": 51,
+    "Number of SZSE-listed private enterprises in the General Equipment Manufacturing industry": 86,
+    "Difference": -35.0
+  },
+  "reference": [
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "3f2c07ea-c142-43aa-ab56-ee76d8aaa4e5"
+    },
+    {
+      "national_industry_status": "d6494fad-ee50-4718-bed4-9401fb74a494"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium082_result.json b/assets/qa_raw/enterprise_industry_analysis/medium082_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f77177ac88d49c12b633f31eb332e0eafd5c15bc
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium082_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium082",
+  "question": "Between the maximum operating revenue amount of the industry where Zhaoye Huachang Real Estate Development Company operates and that of the General Equipment Manufacturing industry, which is higher?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhaoye Huachang Real Estate Development Company: Real Estate",
+    "Real Estate: maximum operating revenue amount = 503838390573.74 Yuan",
+    "General Equipment Manufacturing: maximum operating revenue amount = 117623139663 Yuan"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Zhaoye Huachang Real Estate Development Company belongs to the Real Estate industry.",
+    "Extract from national_industry_status.csv that the maximum operating revenue amount of the Real Estate industry is 503838390573.74 Yuan.",
+    "Extract from national_industry_status.csv that the maximum operating revenue amount of the General Equipment Manufacturing industry is 117623139663 Yuan.",
+    "Compare the two maximum values: 503838390573.74 Yuan > 117623139663 Yuan, so Zhaoye Huachang Real Estate Development Company is higher."
+  ],
+  "steps_num": 4,
+  "answer": "Zhaoye Huachang Real Estate Development Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Zhaoye Huachang Real Estate Development Company": "Real Estate",
+    "Maximum operating revenue amount of the Real Estate industry (Yuan)": 503838390573.74,
+    "Maximum operating revenue amount of the General Equipment Manufacturing industry (Yuan)": 117623139663,
+    "Comparison result (higher one)": "Zhaoye Huachang Real Estate Development Company"
+  },
+  "reference": [
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "3f2c07ea-c142-43aa-ab56-ee76d8aaa4e5"
+    },
+    {
+      "national_industry_status": "d6494fad-ee50-4718-bed4-9401fb74a494"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium083_result.json b/assets/qa_raw/enterprise_industry_analysis/medium083_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..494302ed7c1b357b2be33800a339b7a834ea9b47
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium083_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium083",
+  "question": "What is the difference between the number of SZSE-listed enterprises in the industry where Yihai Changjin Business Company operates and the number of SZSE-listed central state-owned enterprises in the Communication Transmission Equipment industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Yihai Changjin Business Company: Leasing and Business Services",
+    "Leasing and Business Services: number of SZSE-listed enterprises = 44",
+    "Communication Transmission Equipment industry: number of SZSE-listed central state-owned enterprises = 3"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Yihai Changjin Business Company belongs to the Leasing and Business Services industry.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed enterprises in the Leasing and Business Services industry is 44.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed central state-owned enterprises in the Communication Transmission Equipment industry is 3.",
+    "Calculate the difference: 44 - 3 = 41.0."
+  ],
+  "steps_num": 4,
+  "answer": 41.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Yihai Changjin Business Company": "Leasing and Business Services",
+    "Number of SZSE-listed enterprises in the Leasing and Business Services industry": 44,
+    "Number of SZSE-listed central state-owned enterprises in the Communication Transmission Equipment industry": 3,
+    "Difference": 41.0
+  },
+  "reference": [
+    {
+      "national_industry_status": "e4578b09-ad56-4edf-adc8-9cd6b54149b7"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    },
+    {
+      "national_industry_status": "22fbfe9c-7c9b-4c5c-82a5-46baa3b08f2e"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium084_result.json b/assets/qa_raw/enterprise_industry_analysis/medium084_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5d929ec9a622ccfb25a89e52392106534f0a2e23
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium084_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium084",
+  "question": "Which is larger: the number of SSE-listed central state-owned enterprises in the industry where Yihai Changjin Business Company operates, or the total number of enterprises in the Communication Transmission Equipment industry?",
+  "guidelines": "The answer must be either \"the number of SSE-listed central state-owned enterprises in the industry where Yihai Changjin Business Company operates\" or \"the total number of enterprises in the Communication Transmission Equipment industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Yihai Changjin Business Company: Leasing and Business Services",
+    "Leasing and Business Services: number of SSE-listed central state-owned enterprises = 3",
+    "Communication Transmission Equipment industry: total number of enterprises = 120"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Yihai Changjin Business Company belongs to the Leasing and Business Services industry.",
+    "Extract from national_industry_status.csv that the number of SSE-listed central state-owned enterprises in the Leasing and Business Services industry is 3.",
+    "Extract from national_industry_status.csv that the total number of enterprises in the Communication Transmission Equipment industry is 120.",
+    "Compare the counts: 120 > 3, so the total number of enterprises in the Communication Transmission Equipment industry is larger."
+  ],
+  "steps_num": 4,
+  "answer": "the total number of enterprises in the Communication Transmission Equipment industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Yihai Changjin Business Company": "Leasing and Business Services",
+    "Number of SSE-listed central state-owned enterprises in the Leasing and Business Services industry": 3,
+    "Total number of enterprises in the Communication Transmission Equipment industry": 120,
+    "Comparison result (larger one)": "the total number of enterprises in the Communication Transmission Equipment industry"
+  },
+  "reference": [
+    {
+      "national_industry_status": "e4578b09-ad56-4edf-adc8-9cd6b54149b7"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    },
+    {
+      "national_industry_status": "22fbfe9c-7c9b-4c5c-82a5-46baa3b08f2e"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium085_result.json b/assets/qa_raw/enterprise_industry_analysis/medium085_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0380a66a0d19580d994bf7135e119a2fdbde726
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium085_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium085",
+  "question": "Which is larger: the number of HKEX-listed local state-owned enterprises in the industry where Zhongke Zhiyun Data Services Company operates, or the number of BSE-listed enterprises in the Transportation, Storage and Postal Services industry?",
+  "guidelines": "The answer must be either \"Zhongke Zhiyun Data Services Company\" or \"the number of BSE-listed enterprises in the Transportation, Storage and Postal Services industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhongke Zhiyun Data Services Company: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services: number of HKEX-listed local state-owned enterprises = 9",
+    "Transportation, Storage and Postal Services: number of BSE-listed enterprises = 2"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Zhongke Zhiyun Data Services Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed local state-owned enterprises in Information Transmission, Software and IT Services is 9.",
+    "Extract from national_industry_status.csv that the number of BSE-listed enterprises in the Transportation, Storage and Postal Services industry is 2.",
+    "Compare the counts: 9 > 2, so the number of HKEX-listed local state-owned enterprises in the industry where Zhongke Zhiyun Data Services Company operates is larger."
+  ],
+  "steps_num": 4,
+  "answer": "Zhongke Zhiyun Data Services Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Zhongke Zhiyun Data Services Company": "Information Transmission, Software and IT Services",
+    "Number of HKEX-listed local state-owned enterprises in Information Transmission, Software and IT Services": 9,
+    "Number of BSE-listed enterprises in the Transportation, Storage and Postal Services industry": 2,
+    "Comparison result (larger one)": "Zhongke Zhiyun Data Services Company"
+  },
+  "reference": [
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "d6ab2642-2691-4f63-a197-6a4c29026c0d"
+    },
+    {
+      "national_industry_status": "c5e98e13-e208-406f-aa43-9e029e2c4de6"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium086_result.json b/assets/qa_raw/enterprise_industry_analysis/medium086_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c3aa78ed1e102d80dcc46f58cd1087df26c07c4b
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium086_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium086",
+  "question": "Which is larger: the number of SSE-listed central state-owned enterprises in the industry where Zhongke Zhiyun Data Services Company operates, or the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry?",
+  "guidelines": "The answer must be either \"the number of SSE-listed central state-owned enterprises in the industry where Zhongke Zhiyun Data Services Company operates\" or \"the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhongke Zhiyun Data Services Company: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services: number of SSE-listed central state-owned enterprises = 12",
+    "Transportation, Storage and Postal Services: number of SSE-listed enterprises = 75"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Zhongke Zhiyun Data Services Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the number of SSE-listed central state-owned enterprises in Information Transmission, Software and IT Services is 12.",
+    "Extract from national_industry_status.csv that the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry is 75.",
+    "Compare the counts: 75 > 12, so the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry is larger."
+  ],
+  "steps_num": 4,
+  "answer": "the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Zhongke Zhiyun Data Services Company": "Information Transmission, Software and IT Services",
+    "Number of SSE-listed central state-owned enterprises in Information Transmission, Software and IT Services": 12,
+    "Number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry": 75,
+    "Comparison result (larger one)": "the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry"
+  },
+  "reference": [
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "d6ab2642-2691-4f63-a197-6a4c29026c0d"
+    },
+    {
+      "national_industry_status": "c5e98e13-e208-406f-aa43-9e029e2c4de6"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium087_result.json b/assets/qa_raw/enterprise_industry_analysis/medium087_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a798d44a26bd2361781ddb5f23001a3fcbb49d14
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium087_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium087",
+  "question": "Which is larger: the number of SZSE-listed central state-owned enterprises in the industry where Wuli Huida Chain Company operates, or the number of HKEX-listed foreign-funded enterprises in the Communication Transmission Equipment industry?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Wuli Huida Chain Company: Wholesale and Retail",
+    "Wholesale and Retail: number of SZSE-listed central state-owned enterprises = 8",
+    "Communication Transmission Equipment industry: number of HKEX-listed foreign-funded enterprises = 1"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Wuli Huida Chain Company belongs to the Wholesale and Retail industry.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed central state-owned enterprises in the Wholesale and Retail industry is 8.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed foreign-funded enterprises in the Communication Transmission Equipment industry is 1.",
+    "Compare the counts: 8 > 1, so the industry where Wuli Huida Chain Company operates is larger."
+  ],
+  "steps_num": 4,
+  "answer": "Wuli Huida Chain Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Wuli Huida Chain Company": "Wholesale and Retail",
+    "Number of SZSE-listed central state-owned enterprises in the Wholesale and Retail industry": 8,
+    "Number of HKEX-listed foreign-funded enterprises in the Communication Transmission Equipment industry": 1,
+    "Comparison result (larger one)": "Wuli Huida Chain Company"
+  },
+  "reference": [
+    {
+      "national_industry_status": "85944882-55f3-4542-bc32-331105184238"
+    },
+    {
+      "company_profile": "24d73db3-b3be-4f9d-87e8-a8f7bcd7b0f9"
+    },
+    {
+      "national_industry_status": "22fbfe9c-7c9b-4c5c-82a5-46baa3b08f2e"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium088_result.json b/assets/qa_raw/enterprise_industry_analysis/medium088_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3edc4d5f944acafc44b2dac3736e0572968d9579
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium088_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium088",
+  "question": "Which is higher: the total capitalized R&D expenditure of the industry where Wuli Huida Chain Company operates, or that of the Communication Transmission Equipment industry?",
+  "guidelines": "The answer must be either \"the total capitalized R&D expenditure of the industry where Wuli Huida Chain Company operates\" or \"the Communication Transmission Equipment industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Wuli Huida Chain Company: Wholesale and Retail",
+    "Wholesale and Retail: total capitalized R&D expenditure = 2537411708.13 Yuan",
+    "Communication Transmission Equipment industry: total capitalized R&D expenditure = 5279289114.7 Yuan"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Wuli Huida Chain Company belongs to the Wholesale and Retail industry.",
+    "Extract from national_industry_status.csv that the total capitalized R&D expenditure of the Wholesale and Retail industry is 2537411708.13 Yuan.",
+    "Extract from national_industry_status.csv that the total capitalized R&D expenditure of the Communication Transmission Equipment industry is 5279289114.7 Yuan.",
+    "Compare the totals: 5279289114.7 Yuan > 2537411708.13 Yuan, so the Communication Transmission Equipment industry is higher."
+  ],
+  "steps_num": 4,
+  "answer": "the Communication Transmission Equipment industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Wuli Huida Chain Company": "Wholesale and Retail",
+    "Total capitalized R&D expenditure of the Wholesale and Retail industry (Yuan)": 2537411708.13,
+    "Total capitalized R&D expenditure of the Communication Transmission Equipment industry (Yuan)": 5279289114.7,
+    "Comparison result (higher one)": "the Communication Transmission Equipment industry"
+  },
+  "reference": [
+    {
+      "national_industry_status": "85944882-55f3-4542-bc32-331105184238"
+    },
+    {
+      "company_profile": "24d73db3-b3be-4f9d-87e8-a8f7bcd7b0f9"
+    },
+    {
+      "national_industry_status": "22fbfe9c-7c9b-4c5c-82a5-46baa3b08f2e"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium089_result.json b/assets/qa_raw/enterprise_industry_analysis/medium089_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c4148536604772f431378602bdfa46fdd85370e
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium089_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium089",
+  "question": "What is the difference between the minimum cumulative citation count of core patents in the industry where Huaxin Yuanshi New Materials Company operates and the same metric in the Scientific Research and Technical Services industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Huaxin Yuanshi New Materials Company: Non-metallic Mineral Products",
+    "Non-metallic Mineral Products: minimum cumulative citation count of core patents = 0",
+    "Scientific Research and Technical Services: minimum cumulative citation count of core patents = 0"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Huaxin Yuanshi New Materials Company belongs to the Non-metallic Mineral Products industry.",
+    "Extract from national_industry_status.csv that the minimum cumulative citation count of core patents in the Non-metallic Mineral Products industry is 0.",
+    "Extract from national_industry_status.csv that the minimum cumulative citation count of core patents in the Scientific Research and Technical Services industry is 0.",
+    "Calculate the difference: 0 - 0 = 0.0."
+  ],
+  "steps_num": 4,
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Huaxin Yuanshi New Materials Company": "Non-metallic Mineral Products",
+    "Minimum cumulative citation count of core patents in the Non-metallic Mineral Products industry": 0,
+    "Minimum cumulative citation count of core patents in the Scientific Research and Technical Services industry": 0,
+    "Difference": 0.0
+  },
+  "reference": [
+    {
+      "national_industry_status": "be3c0164-12b0-4453-8c7a-e198721d914b"
+    },
+    {
+      "company_profile": "f150e113-74af-4929-b33d-7b30a892e86d"
+    },
+    {
+      "national_industry_status": "83fd9785-8bdc-4a18-ad0a-83f92b9aef7c"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium090_result.json b/assets/qa_raw/enterprise_industry_analysis/medium090_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..22c6496f52ee72536ee3b9340bb4436ea9a9a0bd
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium090_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium090",
+  "question": "Compared with the same indicator in the Scientific Research and Technical Services industry, what is the difference in the mean cumulative number of PCT invention patent applications for the industry where Huaxin Yuanshi New Materials Company operates?",
+  "guidelines": "The answer must be an exact number, preserving all meaningful decimal places. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Huaxin Yuanshi New Materials Company: Non-metallic Mineral Products",
+    "Non-metallic Mineral Products: mean cumulative number of PCT invention patent applications = 12.063829787234",
+    "Scientific Research and Technical Services: mean cumulative number of PCT invention patent applications = 34.0425531914894"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Huaxin Yuanshi New Materials Company belongs to the Non-metallic Mineral Products industry.",
+    "Extract from national_industry_status.csv that the mean cumulative number of PCT invention patent applications in the Non-metallic Mineral Products industry is 12.063829787234.",
+    "Extract from national_industry_status.csv that the mean cumulative number of PCT invention patent applications in the Scientific Research and Technical Services industry is 34.0425531914894.",
+    "Calculate the difference: 12.063829787234 - 34.0425531914894 = -21.9787234042554."
+  ],
+  "steps_num": 4,
+  "answer": -21.9787234042554,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Huaxin Yuanshi New Materials Company": "Non-metallic Mineral Products",
+    "Mean cumulative number of PCT invention patent applications in the Non-metallic Mineral Products industry": 12.063829787234,
+    "Mean cumulative number of PCT invention patent applications in the Scientific Research and Technical Services industry": 34.0425531914894,
+    "Difference": -21.9787234042554
+  },
+  "reference": [
+    {
+      "national_industry_status": "be3c0164-12b0-4453-8c7a-e198721d914b"
+    },
+    {
+      "company_profile": "f150e113-74af-4929-b33d-7b30a892e86d"
+    },
+    {
+      "national_industry_status": "83fd9785-8bdc-4a18-ad0a-83f92b9aef7c"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium091_result.json b/assets/qa_raw/enterprise_industry_analysis/medium091_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..31d43ec6d8ac85c32624efc67e502ecd5a455b8c
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium091_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium091",
+  "question": "What is the difference between the maximum number of international industry awards in the industry where Aijian Yikang Fuzhongxin Company operates and the maximum number of international industry awards in the Chemical Fiber Manufacturing industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Aijian Yikang Fuzhongxin Company: Health and Social Work",
+    "Health and Social Work: maximum number of international industry awards = 0",
+    "Chemical Fiber Manufacturing industry: maximum number of international industry awards = 0"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Aijian Yikang Fuzhongxin Company belongs to the Health and Social Work industry.",
+    "Extract from national_industry_status.csv that the maximum number of international industry awards in the Health and Social Work industry is 0.",
+    "Extract from national_industry_status.csv that the maximum number of international industry awards in the Chemical Fiber Manufacturing industry is 0.",
+    "Calculate the difference: 0 - 0 = 0.0."
+  ],
+  "steps_num": 4,
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Aijian Yikang Fuzhongxin Company": "Health and Social Work",
+    "Maximum number of international industry awards in the Health and Social Work industry": 0,
+    "Maximum number of international industry awards in the Chemical Fiber Manufacturing industry": 0,
+    "Difference": 0.0
+  },
+  "reference": [
+    {
+      "national_industry_status": "a08b718a-c663-4252-a4f4-56c3034bae0c"
+    },
+    {
+      "company_profile": "10890d3a-b03b-417b-b39e-f5070604b51e"
+    },
+    {
+      "national_industry_status": "e53c0704-dbd1-451f-82b2-2a0df9488daf"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium092_result.json b/assets/qa_raw/enterprise_industry_analysis/medium092_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4e1f7d34c1cd1b16708b898bd5eb759b1a855b5
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium092_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium092",
+  "question": "What is the difference between the median year-on-year change in operating profit for the industry where Aijian Yikang Fuzhongxin Company operates and that of the Chemical Fiber Manufacturing industry?",
+  "guidelines": "The answer must be a single number with three decimal places. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Aijian Yikang Fuzhongxin Company: Health and Social Work",
+    "Health and Social Work: median year-on-year change in operating profit = -15.11 %",
+    "Chemical Fiber Manufacturing industry: median year-on-year change in operating profit = -31.785 %"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Aijian Yikang Fuzhongxin Company belongs to the Health and Social Work industry.",
+    "Extract from national_industry_status.csv that the median year-on-year change in operating profit for the Health and Social Work industry is -15.11 %.",
+    "Extract from national_industry_status.csv that the median year-on-year change in operating profit for the Chemical Fiber Manufacturing industry is -31.785 %.",
+    "Calculate the difference: -15.11 - (-31.785) = 16.675."
+  ],
+  "steps_num": 4,
+  "answer": 16.675,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Aijian Yikang Fuzhongxin Company": "Health and Social Work",
+    "Median year-on-year change in operating profit for the Health and Social Work industry": -15.11,
+    "Median year-on-year change in operating profit for the Chemical Fiber Manufacturing industry": -31.785,
+    "Difference": 16.675
+  },
+  "reference": [
+    {
+      "national_industry_status": "a08b718a-c663-4252-a4f4-56c3034bae0c"
+    },
+    {
+      "company_profile": "10890d3a-b03b-417b-b39e-f5070604b51e"
+    },
+    {
+      "national_industry_status": "e53c0704-dbd1-451f-82b2-2a0df9488daf"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium093_result.json b/assets/qa_raw/enterprise_industry_analysis/medium093_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..6a21f9bb5982ee74e37eb045b2152251cc3734c7
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium093_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium093",
+  "question": "What is the difference between the number of SSE-listed private enterprises in the industry where Zhongke Keshu Software Company operates and the number of SZSE-listed enterprises in Other Manufacturing in China?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhongke Keshu Software Company: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services: number of SSE-listed private enterprises = 96",
+    "Other Manufacturing (China): number of SZSE-listed enterprises = 25"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Zhongke Keshu Software Company belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from national_industry_status.csv that the number of SSE-listed private enterprises in Information Transmission, Software and IT Services is 96.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed enterprises in Other Manufacturing is 25.",
+    "Calculate the difference: 96 - 25 = 71.0."
+  ],
+  "steps_num": 4,
+  "answer": 71.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Zhongke Keshu Software Company": "Information Transmission, Software and IT Services",
+    "Number of SSE-listed private enterprises in Information Transmission, Software and IT Services": 96,
+    "Number of SZSE-listed enterprises in Other Manufacturing": 25,
+    "Difference": 71.0
+  },
+  "reference": [
+    {
+      "national_industry_status": "52cd1497-bff9-43db-b3a1-001e4cac545c"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "national_industry_status": "15b9907e-9b45-4636-bb57-dee26f48b2d3"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium094_result.json b/assets/qa_raw/enterprise_industry_analysis/medium094_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..9be65f7878331eb12128497b2817bca771e3de4f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium094_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium094",
+  "question": "What is the difference between the number of SZSE-listed private enterprises in the industry where Yihai Changjin Business Company operates and that in China's Transportation, Storage and Postal Services industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Yihai Changjin Business Company: Leasing and Business Services",
+    "Leasing and Business Services: number of SZSE-listed private enterprises = 29",
+    "Transportation, Storage and Postal Services (China): number of SZSE-listed private enterprises = 15"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Yihai Changjin Business Company belongs to the Leasing and Business Services industry.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed private enterprises in the Leasing and Business Services industry is 29.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed private enterprises in the Transportation, Storage and Postal Services industry is 15.",
+    "Calculate the difference: 29 - 15 = 14.0."
+  ],
+  "steps_num": 4,
+  "answer": 14.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Yihai Changjin Business Company": "Leasing and Business Services",
+    "Number of SZSE-listed private enterprises in the Leasing and Business Services industry": 29,
+    "Number of SZSE-listed private enterprises in the Transportation, Storage and Postal Services industry": 15,
+    "Difference": 14.0
+  },
+  "reference": [
+    {
+      "national_industry_status": "e4578b09-ad56-4edf-adc8-9cd6b54149b7"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    },
+    {
+      "national_industry_status": "c5e98e13-e208-406f-aa43-9e029e2c4de6"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium095_result.json b/assets/qa_raw/enterprise_industry_analysis/medium095_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf3247530874b72cd8e245b934ed3d92a7a12d91
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium095_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium095",
+  "question": "Which is larger: the number of SSE-listed private enterprises in the industry where Yihai Changjin Business Company operates, or the number of BSE-listed private enterprises in China's Transportation, Storage and Postal Services industry?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Yihai Changjin Business Company: Leasing and Business Services",
+    "Leasing and Business Services: number of SSE-listed private enterprises = 11",
+    "Transportation, Storage and Postal Services (China): number of BSE-listed private enterprises = 2"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Yihai Changjin Business Company belongs to the Leasing and Business Services industry.",
+    "Extract from national_industry_status.csv that the number of SSE-listed private enterprises in the Leasing and Business Services industry is 11.",
+    "Extract from national_industry_status.csv that the number of BSE-listed private enterprises in the Transportation, Storage and Postal Services industry is 2.",
+    "Compare the counts: 11 > 2, so Yihai Changjin Business Company is higher."
+  ],
+  "steps_num": 4,
+  "answer": "Yihai Changjin Business Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Yihai Changjin Business Company": "Leasing and Business Services",
+    "Number of SSE-listed private enterprises in the Leasing and Business Services industry": 11,
+    "Number of BSE-listed private enterprises in the Transportation, Storage and Postal Services industry": 2,
+    "Comparison result (larger one)": "Yihai Changjin Business Company"
+  },
+  "reference": [
+    {
+      "national_industry_status": "e4578b09-ad56-4edf-adc8-9cd6b54149b7"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    },
+    {
+      "national_industry_status": "c5e98e13-e208-406f-aa43-9e029e2c4de6"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium096_result.json b/assets/qa_raw/enterprise_industry_analysis/medium096_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e3e4b257e8bb10acc4068983acec7d259bffbd6
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium096_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium096",
+  "question": "Which is larger: the number of HKEX-listed central state-owned enterprises in the industry where Zhongche Yuanze Shipbuilding Company operates, or the number of HKEX-listed private enterprises in China's Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing industry?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhongche Yuanze Shipbuilding Company: Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing: number of HKEX-listed central state-owned enterprises = 7",
+    "Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing (China): number of HKEX-listed private enterprises = 13"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Zhongche Yuanze Shipbuilding Company belongs to the Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing industry.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed central state-owned enterprises in this industry is 7.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed private enterprises in China's Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing industry is 13.",
+    "Compare the counts: 13 > 7, so the industry side is larger."
+  ],
+  "steps_num": 4,
+  "answer": "industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Zhongche Yuanze Shipbuilding Company": "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Number of HKEX-listed central state-owned enterprises in Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing": 7,
+    "Number of HKEX-listed private enterprises in Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing": 13,
+    "Comparison result (larger one)": "industry"
+  },
+  "reference": [
+    {
+      "national_industry_status": "4e4122cb-0cfc-47c7-8dde-7296312f31a7"
+    },
+    {
+      "company_profile": "b5c4bcce-8ecd-4a6e-b6ca-e5b6eb41fb8e"
+    },
+    {
+      "national_industry_status": "cf5aa639-7c24-475d-be3b-7b10669c525f"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium097_result.json b/assets/qa_raw/enterprise_industry_analysis/medium097_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d213e4eeaee3e1a40ee52cc7ac947f25dadb4c46
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium097_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium097",
+  "question": "Which is larger: the number of HKEX-listed private enterprises in the industry where Zhongche Yuanze Shipbuilding Company operates, or the number of SSE-listed private enterprises in China's Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing industry?",
+  "guidelines": "The answer must be either \"the number of HKEX-listed private enterprises in the industry where Zhongche Yuanze Shipbuilding Company operates\", \"China\", or \"Equal\". Output only the answer itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhongche Yuanze Shipbuilding Company: Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing: number of HKEX-listed private enterprises = 5",
+    "Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing (China): number of SSE-listed private enterprises = 5"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Zhongche Yuanze Shipbuilding Company belongs to the Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing industry.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed private enterprises in this industry is 5.",
+    "Extract from national_industry_status.csv that the number of SSE-listed private enterprises in China's Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing industry is 5.",
+    "Compare the counts: 5 = 5, so the result is Equal."
+  ],
+  "steps_num": 4,
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Zhongche Yuanze Shipbuilding Company": "Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing",
+    "Number of HKEX-listed private enterprises in Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing": 5,
+    "Number of SSE-listed private enterprises in Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing": 5,
+    "Comparison result": "Equal"
+  },
+  "reference": [
+    {
+      "national_industry_status": "4e4122cb-0cfc-47c7-8dde-7296312f31a7"
+    },
+    {
+      "company_profile": "b5c4bcce-8ecd-4a6e-b6ca-e5b6eb41fb8e"
+    },
+    {
+      "national_industry_status": "cf5aa639-7c24-475d-be3b-7b10669c525f"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium098_result.json b/assets/qa_raw/enterprise_industry_analysis/medium098_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea4d68a28ffebd51d23fcefa29ea94f07819d905
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium098_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium098",
+  "question": "Which is larger: the number of SZSE-listed enterprises in the industry where Yihai Changjin Business Company operates, or the number of SZSE-listed private enterprises in China's Conglomerates industry?",
+  "guidelines": "The answer must be either the company name or \"industry\". Output only the name, without any explanation or description. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Yihai Changjin Business Company: Leasing and Business Services",
+    "Leasing and Business Services: number of SZSE-listed enterprises = 44",
+    "Conglomerates (China): number of SZSE-listed private enterprises = 4"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Yihai Changjin Business Company belongs to the Leasing and Business Services industry.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed enterprises in the Leasing and Business Services industry is 44.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed private enterprises in the Conglomerates industry is 4.",
+    "Compare the counts: 44 > 4, so Yihai Changjin Business Company is higher."
+  ],
+  "steps_num": 4,
+  "answer": "Yihai Changjin Business Company",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Yihai Changjin Business Company": "Leasing and Business Services",
+    "Number of SZSE-listed enterprises in the Leasing and Business Services industry": 44,
+    "Number of SZSE-listed private enterprises in the Conglomerates industry": 4,
+    "Comparison result (larger one)": "Yihai Changjin Business Company"
+  },
+  "reference": [
+    {
+      "national_industry_status": "e4578b09-ad56-4edf-adc8-9cd6b54149b7"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    },
+    {
+      "national_industry_status": "fee7bf00-36e3-4f1a-a296-f5612ad8cedd"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium099_result.json b/assets/qa_raw/enterprise_industry_analysis/medium099_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..80f492f6522a5cca6ae44f3a43bfcde94eb2e540
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium099_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium099",
+  "question": "Yi Hai Chang Jin Shang Wu Co., Ltd.industry's Number of SZSE-listed enterprises and China ConglomeratesindustryPrivate enterpriseShenzhen Stock Exchange countcompared with which unitsgreater？",
+  "guidelines": "The answer must a company name or \"industry\", Output onlyname, without any explanation or description.Wu, answer\"No relevant data found\".",
+  "evidence": [
+    "Yi Hai Chang Jin Shang Wu Co., Ltd.industry：Business Services",
+    "Business Services：Number of SZSE-listed enterprises 44",
+    "Conglomerates（China）：Private enterprise_Number of SZSE-listed enterprises 4"
+  ],
+  "steps": [
+    "Extracted from company_profile.csv: Yi Hai Chang Jin Shang Wu Co., Ltd.industry = Business Services",
+    "national_industry_status.csv in extractBusiness Services Number of SZSE-listed enterprises= 44",
+    "national_industry_status.csv in extractConglomeratesindustry Private enterprise_Number of SZSE-listed enterprises= 4",
+    "Comparecount：44 > 4, Yi Hai Chang Jin Shang Wu Co., Ltd.higher"
+  ],
+  "steps_num": 4,
+  "answer": "Yi Hai Chang Jin Shang Wu Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Yi Hai Chang Jin Shang Wu Co., Ltd.industry": "Business Services",
+    "Business Services Number of SZSE-listed enterprises": 44,
+    "Conglomerates Private enterprise_Number of SZSE-listed enterprises": 4,
+    "（greater）": "Yi Hai Chang Jin Shang Wu Co., Ltd."
+  },
+  "reference": [
+    {
+      "national_industry_status": "e4578b09-ad56-4edf-adc8-9cd6b54149b7"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    },
+    {
+      "national_industry_status": "fee7bf00-36e3-4f1a-a296-f5612ad8cedd"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium100_result.json b/assets/qa_raw/enterprise_industry_analysis/medium100_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..324d8c4f74bba861d7ba48415a24928fb9fa5b73
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium100_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium100",
+  "question": "What is the difference between the median year-on-year change in R&D expenditure for the industry where Biyuan Zhize Urban Development Company operates and that of the Pharmaceutical Manufacturing industry in China?",
+  "guidelines": "The answer must be a single number with two decimal places. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Biyuan Zhize Urban Development Company: Real Estate",
+    "Real Estate: median year-on-year change in R&D expenditure = 1.68 %",
+    "Pharmaceutical Manufacturing (China): median year-on-year change in R&D expenditure = 11.55 %"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Biyuan Zhize Urban Development Company belongs to the Real Estate industry.",
+    "Extract from national_industry_status.csv that the median year-on-year change in R&D expenditure for the Real Estate industry is 1.68 %.",
+    "Extract from national_industry_status.csv that the median year-on-year change in R&D expenditure for the Pharmaceutical Manufacturing industry is 11.55 %.",
+    "Calculate the gap: 1.68 - 11.55 = -9.87."
+  ],
+  "steps_num": 4,
+  "answer": -9.87,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry of Biyuan Zhize Urban Development Company": "Real Estate",
+    "Median year-on-year change in R&D expenditure for the Real Estate industry": 1.68,
+    "Median year-on-year change in R&D expenditure for the Pharmaceutical Manufacturing industry": 11.55,
+    "Gap": -9.87
+  },
+  "reference": [
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "1497d311-a090-4fdb-9af8-30fee577d417"
+    },
+    {
+      "national_industry_status": "9c71278e-97a3-4867-826d-e139f1dffc24"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium101_result.json b/assets/qa_raw/enterprise_industry_analysis/medium101_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e89c5dc65b378e951f1a5066a6315bbb467949b
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium101_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium101",
+  "question": "What is the difference between the number of SSE-listed foreign-funded enterprises in the industry corresponding to Biyuan Zhize Urban Development Company and the number of HKEX-listed state-owned research institute enterprises in China's Pharmaceutical Manufacturing industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry corresponding to Biyuan Zhize Urban Development Company: Real Estate",
+    "Real Estate: number of SSE-listed foreign-funded enterprises = 6",
+    "Pharmaceutical Manufacturing (China): number of HKEX-listed state-owned research institute enterprises = 1"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that the corresponding industry of Biyuan Zhize Urban Development Company is Real Estate.",
+    "Extract from national_industry_status.csv that the number of SSE-listed foreign-funded enterprises in the Real Estate industry is 6.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed state-owned research institute enterprises in the Pharmaceutical Manufacturing industry is 1.",
+    "Calculate the difference: 6 - 1 = 5.0."
+  ],
+  "steps_num": 4,
+  "answer": 5.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Industry corresponding to Biyuan Zhize Urban Development Company": "Real Estate",
+    "Number of SSE-listed foreign-funded enterprises in the Real Estate industry": 6,
+    "Number of HKEX-listed state-owned research institute enterprises in the Pharmaceutical Manufacturing industry": 1,
+    "Difference": 5.0
+  },
+  "reference": [
+    {
+      "national_industry_status": "27f2a242-57e9-4b4a-b933-e02981cf17d2"
+    },
+    {
+      "company_profile": "1497d311-a090-4fdb-9af8-30fee577d417"
+    },
+    {
+      "national_industry_status": "9c71278e-97a3-4867-826d-e139f1dffc24"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium102_result.json b/assets/qa_raw/enterprise_industry_analysis/medium102_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a51d75dcc0e94d2c3c387d7170a6fcc9cca1e2ed
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium102_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium102",
+  "question": "Which is larger: the total number of Consumer Electronics and Electrical industry enterprises in the province where Changqiao Jinchuang Technology Company is located, or the number of HKEX-listed foreign-funded enterprises in China's Conglomerates industry?",
+  "guidelines": "The answer must be either \"the total number of Consumer Electronics and Electrical industry enterprises in the province where Changqiao Jinchuang Technology Company is located\" or \"the number of HKEX-listed foreign-funded enterprises in China's Conglomerates industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Changqiao Jinchuang Technology Company: Gansu Province",
+    "Gansu Province - Consumer Electronics and Electrical industry: total number of enterprises = 0",
+    "Conglomerates (China): number of HKEX-listed foreign-funded enterprises = 5"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Changqiao Jinchuang Technology Company is located in Gansu Province and belongs to the Consumer Electronics and Electrical industry.",
+    "Extract from regional_industry_status.csv that the total number of enterprises in Gansu Province - Consumer Electronics and Electrical industry is 0.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed foreign-funded enterprises in the Conglomerates industry is 5.",
+    "Compare the counts: 5 > 0, so the number of HKEX-listed foreign-funded enterprises in China's Conglomerates industry is larger."
+  ],
+  "steps_num": 4,
+  "answer": "the number of HKEX-listed foreign-funded enterprises in China's Conglomerates industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Changqiao Jinchuang Technology Company": "Gansu Province",
+    "Total number of enterprises in Gansu Province - Consumer Electronics and Electrical industry": 0,
+    "Number of HKEX-listed foreign-funded enterprises in the Conglomerates industry": 5,
+    "Comparison result (larger one)": "the number of HKEX-listed foreign-funded enterprises in China's Conglomerates industry"
+  },
+  "reference": [
+    {
+      "regional_industry_status": "3b2f0630-2cdb-49c4-9fae-264d4f9e4ff4"
+    },
+    {
+      "company_profile": "58eb73a2-c225-4c83-9324-b515fec3a76f"
+    },
+    {
+      "national_industry_status": "fee7bf00-36e3-4f1a-a296-f5612ad8cedd"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium103_result.json b/assets/qa_raw/enterprise_industry_analysis/medium103_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..362c178fb1c02d5b5b6a7526ea14d54aa412a06b
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium103_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium103",
+  "question": "Which is larger: the total number of Consumer Electronics and Electrical industry enterprises in the province where Changqiao Jinchuang Technology Company is located, or the number of HKEX-listed private enterprises in China's Conglomerates industry?",
+  "guidelines": "The answer must be either \"the total number of Consumer Electronics and Electrical industry enterprises in the province where Changqiao Jinchuang Technology Company is located\" or \"the number of HKEX-listed private enterprises in China's Conglomerates industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Changqiao Jinchuang Technology Company: Gansu Province",
+    "Gansu Province - Consumer Electronics and Electrical industry: total number of enterprises = 0",
+    "Conglomerates (China): number of HKEX-listed private enterprises = 12"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Changqiao Jinchuang Technology Company is located in Gansu Province and belongs to the Consumer Electronics and Electrical industry.",
+    "Extract from regional_industry_status.csv that the total number of enterprises in Gansu Province - Consumer Electronics and Electrical industry is 0.",
+    "Extract from national_industry_status.csv that the number of HKEX-listed private enterprises in the Conglomerates industry is 12.",
+    "Compare the counts: 12 > 0, so the number of HKEX-listed private enterprises in China's Conglomerates industry is larger."
+  ],
+  "steps_num": 4,
+  "answer": "the number of HKEX-listed private enterprises in China's Conglomerates industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Changqiao Jinchuang Technology Company": "Gansu Province",
+    "Total number of enterprises in Gansu Province - Consumer Electronics and Electrical industry": 0,
+    "Number of HKEX-listed private enterprises in the Conglomerates industry": 12,
+    "Comparison result (larger one)": "the number of HKEX-listed private enterprises in China's Conglomerates industry"
+  },
+  "reference": [
+    {
+      "regional_industry_status": "3b2f0630-2cdb-49c4-9fae-264d4f9e4ff4"
+    },
+    {
+      "company_profile": "58eb73a2-c225-4c83-9324-b515fec3a76f"
+    },
+    {
+      "national_industry_status": "fee7bf00-36e3-4f1a-a296-f5612ad8cedd"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium104_result.json b/assets/qa_raw/enterprise_industry_analysis/medium104_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c0ae8c3256289e196853a4160ccaabba4c0ce729
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium104_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium104",
+  "question": "Which value is larger: the number of HKEX-listed foreign-funded enterprises in the province where Jinzhi Hongsheng Asset Management Company is located, or the number of SSE-listed central state-owned enterprises in China's Culture, Sports and Entertainment industry?",
+  "guidelines": "The answer must be either \"the number of HKEX-listed foreign-funded enterprises in the province where Jinzhi Hongsheng Asset Management Company is located\", \"the number of SSE-listed central state-owned enterprises in China's Culture, Sports and Entertainment industry\", or \"Equal\". Output only the answer itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Jinzhi Hongsheng Asset Management Company: Shanghai Municipality",
+    "Shanghai Municipality - Financial industry: number of HKEX-listed foreign-funded enterprises = 2",
+    "Culture, Sports and Entertainment (China): number of SSE-listed central state-owned enterprises = 2"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Jinzhi Hongsheng Asset Management Company is located in Shanghai Municipality and belongs to the Financial industry.",
+    "Extract from regional_industry_status.csv that the number of HKEX-listed foreign-funded enterprises in Shanghai Municipality - Financial industry is 2.",
+    "Extract from national_industry_status.csv that the number of SSE-listed central state-owned enterprises in China's Culture, Sports and Entertainment industry is 2.",
+    "Compare the values: 2 = 2, so the result is Equal."
+  ],
+  "steps_num": 4,
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Jinzhi Hongsheng Asset Management Company": "Shanghai Municipality",
+    "Number of HKEX-listed foreign-funded enterprises in Shanghai Municipality - Financial industry": 2,
+    "Number of SSE-listed central state-owned enterprises in Culture, Sports and Entertainment": 2,
+    "Comparison result (larger one)": "Equal"
+  },
+  "reference": [
+    {
+      "regional_industry_status": "6d7fb68b-9d0a-4cfb-bab2-abac83115eed"
+    },
+    {
+      "company_profile": "62612789-d0a3-4b19-bbee-abe8a89ab1bd"
+    },
+    {
+      "national_industry_status": "9e59028a-f6d1-4a30-ae4b-ce2b5174093e"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium105_result.json b/assets/qa_raw/enterprise_industry_analysis/medium105_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c7a81499312cf799460fc51617f78291eb3f34ee
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium105_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium105",
+  "question": "What is the difference between the mean number of provincial or ministerial Science and Technology Progress Awards in the province where Jinzhi Hongsheng Asset Management Company is located and the mean number of the same awards in China's Culture, Sports and Entertainment industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Jinzhi Hongsheng Asset Management Company: Shanghai Municipality",
+    "Shanghai Municipality - Financial industry: mean number of provincial or ministerial Science and Technology Progress Awards = 3",
+    "Culture, Sports and Entertainment (China): mean number of provincial or ministerial Science and Technology Progress Awards = 0"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Jinzhi Hongsheng Asset Management Company is located in Shanghai Municipality and belongs to the Financial industry.",
+    "Extract from regional_industry_status.csv that the mean number of provincial or ministerial Science and Technology Progress Awards in Shanghai Municipality - Financial industry is 3.",
+    "Extract from national_industry_status.csv that the mean number of provincial or ministerial Science and Technology Progress Awards in the Culture, Sports and Entertainment industry is 0.",
+    "Calculate the difference: 3 - 0 = 3.0."
+  ],
+  "steps_num": 4,
+  "answer": 3.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Jinzhi Hongsheng Asset Management Company": "Shanghai Municipality",
+    "Mean number of provincial or ministerial Science and Technology Progress Awards in Shanghai Municipality - Financial industry": 3,
+    "Mean number of provincial or ministerial Science and Technology Progress Awards in Culture, Sports and Entertainment": 0,
+    "Difference": 3.0
+  },
+  "reference": [
+    {
+      "regional_industry_status": "6d7fb68b-9d0a-4cfb-bab2-abac83115eed"
+    },
+    {
+      "company_profile": "62612789-d0a3-4b19-bbee-abe8a89ab1bd"
+    },
+    {
+      "national_industry_status": "9e59028a-f6d1-4a30-ae4b-ce2b5174093e"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium106_result.json b/assets/qa_raw/enterprise_industry_analysis/medium106_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3c8df422a913081af18a2de3f0b3c6472cafe4d1
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium106_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium106",
+  "question": "What is the difference between the total number of Consumer Electronics and Electrical industry enterprises in the province where Shiyang Jinjin Electrical Appliances Company is located and the number of SZSE-listed enterprises in China's Rubber and Plastic Products industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Shiyang Jinjin Electrical Appliances Company: Inner Mongolia Autonomous Region",
+    "Inner Mongolia Autonomous Region - Consumer Electronics and Electrical industry: total number of enterprises = 0",
+    "Rubber and Plastic Products industry (China): number of SZSE-listed enterprises = 68"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Shiyang Jinjin Electrical Appliances Company is located in the Inner Mongolia Autonomous Region and belongs to the Consumer Electronics and Electrical industry.",
+    "Extract from regional_industry_status.csv that the total number of enterprises in Inner Mongolia Autonomous Region - Consumer Electronics and Electrical industry is 0.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed enterprises in the Rubber and Plastic Products industry is 68.",
+    "Calculate the difference: 0 - 68 = -68.0."
+  ],
+  "steps_num": 4,
+  "answer": -68.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Shiyang Jinjin Electrical Appliances Company": "Inner Mongolia Autonomous Region",
+    "Total number of enterprises in Inner Mongolia Autonomous Region - Consumer Electronics and Electrical industry": 0,
+    "Number of SZSE-listed enterprises in the Rubber and Plastic Products industry": 68,
+    "Difference": -68.0
+  },
+  "reference": [
+    {
+      "regional_industry_status": "fa024fe2-f4d4-401e-9630-e58759c828a0"
+    },
+    {
+      "company_profile": "c4791202-0687-425c-9012-35538f3a300c"
+    },
+    {
+      "national_industry_status": "82707f38-7a84-4096-a41e-d04d8382f558"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium107_result.json b/assets/qa_raw/enterprise_industry_analysis/medium107_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..56e878142d48d9242b4c7aa85882de7ddc85927f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium107_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium107",
+  "question": "What is the difference between the total number of Consumer Electronics and Electrical industry enterprises in the province where Shiyang Jinjin Electrical Appliances Company is located and the number of enterprises in China's Rubber and Plastic Products industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Shiyang Jinjin Electrical Appliances Company: Inner Mongolia Autonomous Region",
+    "Inner Mongolia Autonomous Region - Consumer Electronics and Electrical industry: total number of enterprises = 0",
+    "Rubber and Plastic Products industry (China): number of enterprises = 107"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Shiyang Jinjin Electrical Appliances Company is located in the Inner Mongolia Autonomous Region and belongs to the Consumer Electronics and Electrical industry.",
+    "Extract from regional_industry_status.csv that the total number of enterprises in Inner Mongolia Autonomous Region - Consumer Electronics and Electrical industry is 0.",
+    "Extract from national_industry_status.csv that the number of enterprises in the Rubber and Plastic Products industry is 107.",
+    "Calculate the difference: 0 - 107 = -107.0."
+  ],
+  "steps_num": 4,
+  "answer": -107.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Shiyang Jinjin Electrical Appliances Company": "Inner Mongolia Autonomous Region",
+    "Total number of enterprises in Inner Mongolia Autonomous Region - Consumer Electronics and Electrical industry": 0,
+    "Number of enterprises in the Rubber and Plastic Products industry": 107,
+    "Difference": -107.0
+  },
+  "reference": [
+    {
+      "regional_industry_status": "fa024fe2-f4d4-401e-9630-e58759c828a0"
+    },
+    {
+      "company_profile": "c4791202-0687-425c-9012-35538f3a300c"
+    },
+    {
+      "national_industry_status": "82707f38-7a84-4096-a41e-d04d8382f558"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium108_result.json b/assets/qa_raw/enterprise_industry_analysis/medium108_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..9da89dc1c74c65791b7dc2ce93cf863946633f6a
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium108_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium108",
+  "question": "Which value is larger: the total number of industry enterprises in the province where Zhongke Keshu Software Company is located, or the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry?",
+  "guidelines": "The answer must be either \"the total number of industry enterprises in the province where Zhongke Keshu Software Company is located\" or \"the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Zhongke Keshu Software Company: Guangxi Zhuang Autonomous Region",
+    "Guangxi Zhuang Autonomous Region - Information Transmission, Software and IT Services: total number of enterprises = 4",
+    "Communication Transmission Equipment (China): number of SZSE-listed private enterprises = 50"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Zhongke Keshu Software Company is located in Guangxi Zhuang Autonomous Region and belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from regional_industry_status.csv that the total number of enterprises in Guangxi Zhuang Autonomous Region - Information Transmission, Software and IT Services is 4.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed private enterprises in the Communication Transmission Equipment industry is 50.",
+    "Compare the values: 50 > 4, so the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry is larger."
+  ],
+  "steps_num": 4,
+  "answer": "the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Zhongke Keshu Software Company": "Guangxi Zhuang Autonomous Region",
+    "Total number of enterprises in Guangxi Zhuang Autonomous Region - Information Transmission, Software and IT Services": 4,
+    "Number of SZSE-listed private enterprises in the Communication Transmission Equipment industry": 50,
+    "Comparison result (larger one)": "the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry"
+  },
+  "reference": [
+    {
+      "regional_industry_status": "4b18b7b8-6343-4468-a8dc-88fd19b1cdb8"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "national_industry_status": "22fbfe9c-7c9b-4c5c-82a5-46baa3b08f2e"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium109_result.json b/assets/qa_raw/enterprise_industry_analysis/medium109_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c2e5361f9b04599f303dbaf6719fab565086a92a
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium109_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium109",
+  "question": "Which is larger: the number of SSE-listed private enterprises in the Information Transmission, Software and IT Services industry in the province where Zhongke Keshu Software Company is located, or the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry?",
+  "guidelines": "The answer must be either \"the number of SSE-listed private enterprises in the Information Transmission, Software and IT Services industry in the province where Zhongke Keshu Software Company is located\" or \"the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry\". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Zhongke Keshu Software Company: Guangxi Zhuang Autonomous Region",
+    "Guangxi Zhuang Autonomous Region - Information Transmission, Software and IT Services: number of SSE-listed private enterprises = 2",
+    "Communication Transmission Equipment (China): number of SZSE-listed private enterprises = 50"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Zhongke Keshu Software Company is located in Guangxi Zhuang Autonomous Region and belongs to the Information Transmission, Software and IT Services industry.",
+    "Extract from regional_industry_status.csv that the number of SSE-listed private enterprises in Guangxi Zhuang Autonomous Region - Information Transmission, Software and IT Services is 2.",
+    "Extract from national_industry_status.csv that the number of SZSE-listed private enterprises in the Communication Transmission Equipment industry is 50.",
+    "Compare the counts: 50 > 2, so the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry is larger."
+  ],
+  "steps_num": 4,
+  "answer": "the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Zhongke Keshu Software Company": "Guangxi Zhuang Autonomous Region",
+    "Number of SSE-listed private enterprises in Guangxi Zhuang Autonomous Region - Information Transmission, Software and IT Services": 2,
+    "Number of SZSE-listed private enterprises in the Communication Transmission Equipment industry": 50,
+    "Comparison result (larger one)": "the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry"
+  },
+  "reference": [
+    {
+      "regional_industry_status": "4b18b7b8-6343-4468-a8dc-88fd19b1cdb8"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "national_industry_status": "22fbfe9c-7c9b-4c5c-82a5-46baa3b08f2e"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium110_result.json b/assets/qa_raw/enterprise_industry_analysis/medium110_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4ce607368c5eaa6820cc97b6f3193b9db87458a3
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium110_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium110",
+  "question": "What is the difference between the total annual number of China patent grants in the province where Xingkuwen Arts and Crafts Company is located and that in China's Chemical Fiber Manufacturing industry?",
+  "guidelines": "The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Xingkuwen Arts and Crafts Company: Zhejiang Province",
+    "Zhejiang Province - Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing: total annual number of China patent grants = 193",
+    "Chemical Fiber Manufacturing (China): total annual number of China patent grants = 1507"
+  ],
+  "steps": [
+    "Extract from company_profile.csv that Xingkuwen Arts and Crafts Company is located in Zhejiang Province and belongs to the Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing industry.",
+    "Extract from regional_industry_status.csv that the total annual number of China patent grants in Zhejiang Province - Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing is 193.",
+    "Extract from national_industry_status.csv that the total annual number of China patent grants in the Chemical Fiber Manufacturing industry is 1507.",
+    "Calculate the gap: 193 - 1507 = -1314.0."
+  ],
+  "steps_num": 4,
+  "answer": -1314.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "milestone": {
+    "Province of Xingkuwen Arts and Crafts Company": "Zhejiang Province",
+    "Total annual number of China patent grants in Zhejiang Province - Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing": 193,
+    "Total annual number of China patent grants in the Chemical Fiber Manufacturing industry": 1507,
+    "Gap": -1314.0
+  },
+  "reference": [
+    {
+      "regional_industry_status": "d91beb15-f36a-4783-8c2a-ab1e21b2bcda"
+    },
+    {
+      "company_profile": "8ac66182-6508-416b-b0a3-b7b5f71175eb"
+    },
+    {
+      "national_industry_status": "e53c0704-dbd1-451f-82b2-2a0df9488daf"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_analysis/medium111_result.json b/assets/qa_raw/enterprise_industry_analysis/medium111_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3c18c766c4385425c2b53b9ff3cdbb761acad81b
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_analysis/medium111_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium111",
+  "question": "Which is larger: the number of SZSE-listed enterprises in the province where Xingkuwen Arts and Crafts Company is located, or the number of SSE-listed private enterprises in China's Chemical Fiber Manufacturing industry?",
+  "guidelines": "The answer must be either \"the number of SZSE-listed enterprises in the province where Xingkuwen Arts and Crafts Company is located\" or \"the number of SSE-listed private enterprises in China's Chemical Fiber Manufacturing industry\". Output only one of these, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Xingkuwen Arts and Crafts Company (data in this dataset) is located in Zhejiang Province",
+    "Zhejiang Province (data in this dataset), Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing: number of SZSE-listed enterprises = 3",
+    "National (data in this dataset), Chemical Fiber Manufacturing: number of SSE-listed private enterprises = 7"
+  ],
+  "answer": "the number of SSE-listed private enterprises in China's Chemical Fiber Manufacturing industry",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_analysis"
+  },
+  "steps": [
+    "Determine from company_profile.csv that Xingkuwen Arts and Crafts Company is located in Zhejiang Province.",
+    "Extract from regional_industry_status.csv that the number of SZSE-listed enterprises in Zhejiang Province, Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing is 3.",
+    "Extract from national_industry_status.csv that the number of SSE-listed private enterprises in the Chemical Fiber Manufacturing industry is 7.",
+    "Compare 3 and 7; 7 is larger, so output \"the number of SSE-listed private enterprises in China's Chemical Fiber Manufacturing industry\"."
+  ],
+  "steps_num": 4,
+  "milestone": {
+    "Province of Xingkuwen Arts and Crafts Company": "Zhejiang Province",
+    "Number of SZSE-listed enterprises in Zhejiang Province, Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing": 3,
+    "Number of SSE-listed private enterprises in Chemical Fiber Manufacturing": 7,
+    "Comparison result (larger one)": "the number of SSE-listed private enterprises in China's Chemical Fiber Manufacturing industry"
+  },
+  "reference": [
+    {
+      "regional_industry_status": "d91beb15-f36a-4783-8c2a-ab1e21b2bcda"
+    },
+    {
+      "company_profile": "8ac66182-6508-416b-b0a3-b7b5f71175eb"
+    },
+    {
+      "national_industry_status": "e53c0704-dbd1-451f-82b2-2a0df9488daf"
+    }
+  ]
+}
\ No newline at end of file
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/easy067_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/easy067_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..138d90e74c728ae9235cfe313baea100dc9262b5
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/easy067_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy067",
+  "question": "“江西省人民政府印发关于做优做强我省锂电新能源产业若干政策措施的通知”对众白昌锦商贸公司的行业是否适用？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "evidence": [
+    "众白昌锦商贸公司所在省份为上海市",
+    "政策《江西省人民政府印发关于做优做强我省锂电新能源产业若干政策措施的通知》适用区域为江西省"
+  ],
+  "milestone": {
+    "众白昌锦商贸公司所在省份": "上海市",
+    "政策适用区域": "江西省",
+    "是否适用": "否"
+  },
+  "steps": [
+    "从company_profile.csv中抽取众白昌锦商贸公司所在省份为上海市",
+    "从policy_resource.csv中抽取政策《江西省人民政府印发关于做优做强我省锂电新能源产业若干政策措施的通知》的适用区域为江西省",
+    "判断企业所在省份是否在政策适用区域内，不在则输出「否」"
+  ],
+  "steps_num": 3,
+  "answer": "否",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "company_profile": "e9bfa955-c28c-4e84-aa66-a4728bea39df"
+    },
+    {
+      "policy_resource": "ed302f58-847e-46ae-8c19-6e4989b55232"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/easy068_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/easy068_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..722b5e12574b4208b7b6fa7b01d2212d362f1b48
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/easy068_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy068",
+  "question": "大花表仪医疗科技公司所在行业是否会受到“合肥市促进“两强一增”行动若干政策”的影响？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "evidence": [
+    "大花表仪医疗科技公司所在省份为四川省",
+    "政策《合肥市促进“两强一增”行动若干政策》适用区域为合肥市"
+  ],
+  "milestone": {
+    "大花表仪医疗科技公司所在省份": "四川省",
+    "政策适用区域": "合肥市",
+    "是否适用": "否"
+  },
+  "steps": [
+    "从company_profile.csv中抽取大花表仪医疗科技公司所在省份为四川省",
+    "从policy_resource.csv中抽取政策《合肥市促进“两强一增”行动若干政策》的适用区域为合肥市",
+    "判断企业所在省份是否在政策适用区域内，不在则输出「否」"
+  ],
+  "steps_num": 3,
+  "answer": "否",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "company_profile": "2b695e04-ab4e-4b46-ab33-61011693d5b1"
+    },
+    {
+      "policy_resource": "85b2dea8-e5f8-433b-b361-ec609e9156b7"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/easy069_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/easy069_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e4cc48fe52b27b23979f04fd736b8e6110f1dc7
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/easy069_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy069",
+  "question": "丽群汇通零售公司所在行业是否会受到“教育部办公厅　工业和信息化部办公厅　国家知识产权局办公室关于组织开展“千校万企”协同创新伙伴行动的通知”的影响？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "evidence": [
+    "丽群汇通零售公司所在行业为批发和零售业",
+    "政策《教育部办公厅 工业和信息化部办公厅 国家知识产权局办公室关于组织开展“千校万企”协同创新伙伴行动的通知》适用行业为教育、科学研究和技术服务业"
+  ],
+  "milestone": {
+    "丽群汇通零售公司所在行业": "批发和零售业",
+    "政策适用行业": "教育、科学研究和技术服务业",
+    "是否适用": "否"
+  },
+  "steps": [
+    "从company_profile.csv中抽取丽群汇通零售公司所在行业为批发和零售业",
+    "从policy_resource.csv中抽取政策《教育部办公厅 工业和信息化部办公厅 国家知识产权局办公室关于组织开展“千校万企”协同创新伙伴行动的通知》的适用行业为教育、科学研究和技术服务业",
+    "判断企业所在行业是否命中政策适用行业，不命中则输出「否」"
+  ],
+  "steps_num": 3,
+  "answer": "否",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "company_profile": "54bb17f1-141b-4e11-9be4-b88f888ff947"
+    },
+    {
+      "policy_resource": "60c74b14-c137-403d-8c78-172ecdfbfc8e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/easy070_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/easy070_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c05602c6ef3881679cd44ff2cf49e6c1a8854480
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/easy070_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy070",
+  "question": "“关于印发重庆市促进大中小企业融通发展工作方案（2022—2025年）的通知”对乐动乐博娱乐用品公司的行业是否适用？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "evidence": [
+    "乐动乐博娱乐用品公司所在省份为广东省",
+    "政策《关于印发重庆市促进大中小企业融通发展工作方案（2022—2025年）的通知》适用区域为重庆市"
+  ],
+  "milestone": {
+    "乐动乐博娱乐用品公司所在省份": "广东省",
+    "政策适用区域": "重庆市",
+    "是否适用": "否"
+  },
+  "steps": [
+    "从company_profile.csv中抽取乐动乐博娱乐用品公司所在省份为广东省",
+    "从policy_resource.csv中抽取政策《关于印发重庆市促进大中小企业融通发展工作方案（2022—2025年）的通知》的适用区域为重庆市",
+    "判断企业所在省份是否在政策适用区域内，不在则输出「否」"
+  ],
+  "steps_num": 3,
+  "answer": "否",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "company_profile": "9c529a65-97b8-44b4-ac73-82e62f4100a8"
+    },
+    {
+      "policy_resource": "3efeed8a-28fc-45a7-a276-214e75648157"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/easy071_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/easy071_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..603b4de46065f4b81c6b8efd02ce136676a72756
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/easy071_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy071",
+  "question": "“科技部等九部门关于印发《“十四五” 东西部科技合作实施方案》的通知”对亚玮工泽机床公司的行业是否适用？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "evidence": [
+    "亚玮工泽机床公司所在行业为通用设备制造业",
+    "政策《科技部等九部门关于印发《“十四五” 东西部科技合作实施方案》的通知》适用行业为科学研究和技术服务业"
+  ],
+  "milestone": {
+    "亚玮工泽机床公司所在行业": "通用设备制造业",
+    "政策适用行业": "科学研究和技术服务业",
+    "是否适用": "否"
+  },
+  "steps": [
+    "从company_profile.csv中抽取亚玮工泽机床公司所在行业为通用设备制造业",
+    "从policy_resource.csv中抽取政策《科技部等九部门关于印发《“十四五” 东西部科技合作实施方案》的通知》适用行业为科学研究和技术服务业",
+    "判断企业所在行业是否命中政策适用行业，不命中则输出「否」"
+  ],
+  "steps_num": 3,
+  "answer": "否",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "company_profile": "e62e0dd2-147f-4bf4-a208-d0edb3fd0878"
+    },
+    {
+      "policy_resource": "ec155f47-c7e9-48b8-9f80-7a17a1bfe727"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/easy072_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/easy072_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ca1beec1006b29bebb1e5b8ba01f9bb58358083
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/easy072_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy072",
+  "question": "果融泽鸿资产管理公司所在行业是否受益于“自治区发展改革委关于印发《宁夏回族自治区氢能产业发展规划》的通知”？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "evidence": [
+    "果融泽鸿资产管理公司所在省份为山东省",
+    "政策《自治区发展改革委关于印发《宁夏回族自治区氢能产业发展规划》的通知》适用区域为宁夏回族自治区"
+  ],
+  "milestone": {
+    "果融泽鸿资产管理公司所在省份": "山东省",
+    "政策适用区域": "宁夏回族自治区",
+    "是否适用": "否"
+  },
+  "steps": [
+    "从company_profile.csv中抽取果融泽鸿资产管理公司所在省份为山东省",
+    "从policy_resource.csv中抽取政策《自治区发展改革委关于印发《宁夏回族自治区氢能产业发展规划》的通知》适用区域为宁夏回族自治区",
+    "判断企业所在省份是否在政策适用区域内，不在则输出「否」"
+  ],
+  "steps_num": 3,
+  "answer": "否",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "company_profile": "e10a59f6-9ef1-4c63-9f08-f757e15cadc8"
+    },
+    {
+      "policy_resource": "af3bfab9-9006-4264-9362-60058a4ecbe6"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/easy073_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/easy073_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d83723ce1852df4154ad24207b45842167ee29b8
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/easy073_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy073",
+  "question": "“商务部等14部门关于开展内外贸一体化试点的通知”是否可以推进物丽昌源批发公司所属行业发展？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "evidence": [
+    "物丽昌源批发公司所在行业为批发和零售业",
+    "政策《商务部等14部门关于开展内外贸一体化试点的通知》适用行业包含批发和零售业"
+  ],
+  "milestone": {
+    "物丽昌源批发公司所在行业": "批发和零售业",
+    "政策适用行业": "批发和零售业；互联网和相关服务；交通运输、仓储和邮政业；租赁和商务服务业；电信、广播电视和卫星传输服务；居民服务、修理和其他服务业",
+    "是否适用": "是"
+  },
+  "steps": [
+    "从company_profile.csv中抽取物丽昌源批发公司所在行业为批发和零售业",
+    "从policy_resource.csv中抽取政策《商务部等14部门关于开展内外贸一体化试点的通知》适用行业列表",
+    "判断企业所在行业是否命中政策适用行业，命中则输出「是」"
+  ],
+  "steps_num": 3,
+  "answer": "是",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "company_profile": "4a50579b-d523-4bca-8dcd-ee5defe6d541"
+    },
+    {
+      "policy_resource": "8f8daf8a-fc47-446c-b1f0-e95403c32f36"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/easy074_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/easy074_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e9559f51504e7a81cac868ce9abe94cac126779
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/easy074_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "easy074",
+  "question": "“广东省人民政府关于印发中国（韶关）等8个  跨境电子商务综合试验区实施方案的通知”对恒通达达信息技术公司的行业是否适用？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "evidence": [
+    "恒通达达信息技术公司所在行业为批发和零售业，所在省份为广东省",
+    "政策《广东省人民政府关于印发中国（韶关）等8个 跨境电子商务综合试验区实施方案的通知》适用行业为批发和零售业，适用区域为广东省"
+  ],
+  "milestone": {
+    "恒通达达信息技术公司所在行业": "批发和零售业",
+    "恒通达达信息技术公司所在省份": "广东省",
+    "政策适用行业": "批发和零售业",
+    "政策适用省份": "广东省",
+    "是否适用": "是"
+  },
+  "steps": [
+    "从company_profile.csv中抽取恒通达达信息技术公司所在行业为批发和零售业且所在省份为广东省",
+    "从policy_resource.csv中抽取政策《广东省人民政府关于印发中国（韶关）等8个 跨境电子商务综合试验区实施方案的通知》的适用行业和适用省份",
+    "判断企业行业与省份是否同时命中政策适用范围，命中则输出「是」"
+  ],
+  "steps_num": 3,
+  "answer": "是",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "company_profile": "e20932eb-d86f-4e5a-936c-110af8517faa"
+    },
+    {
+      "policy_resource": "dda403d0-76ca-4e7e-a415-97b10891d457"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/easy075_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/easy075_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..551a6541b455337f25b5d60c41d4b594e6b57463
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/easy075_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy075",
+  "question": "众课科数软件公司所在行业是否受益于“民政部、中央政法委、中央网信办、发展改革委、工业和信息化部、公安部、财政部、住房城乡建设部、农业农村部印发《关于深入推进智慧社区建设的意见》的通知”？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "evidence": [
+    "众课科数软件公司所在行业为信息传输、软件和信息技术服务业",
+    "政策《民政部、中央政法委、中央网信办、发展改革委、工业和信息化部、公安部、财政部、住房城乡建设部、农业农村部印发〈关于深入推进智慧社区建设的意见〉的通知》适用行业包含信息传输、软件和信息技术服务业"
+  ],
+  "milestone": {
+    "众课科数软件公司所在行业": "信息传输、软件和信息技术服务业",
+    "政策适用行业": "信息传输、软件和信息技术服务业；互联网和相关服务；电信、广播电视和卫星传输服务",
+    "是否适用": "是"
+  },
+  "steps": [
+    "从company_profile.csv中抽取众课科数软件公司所在行业为信息传输、软件和信息技术服务业",
+    "从policy_resource.csv中抽取政策《民政部、中央政法委、中央网信办、发展改革委、工业和信息化部、公安部、财政部、住房城乡建设部、农业农村部印发〈关于深入推进智慧社区建设的意见〉的通知》的适用行业列表",
+    "判断企业所在行业是否命中政策适用行业，命中则输出「是」"
+  ],
+  "steps_num": 3,
+  "answer": "是",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "policy_resource": "78db1ada-cafd-4c7e-be6c-12bb0575d403"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/easy076_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/easy076_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..519f30a54495578c5e169ca2a92871f4fd657a6c
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/easy076_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "easy076",
+  "question": "“四部门关于公布农业、建筑、医疗、矿山领域机器人典型应用场景名单的通知”对以山泽辰医疗器械公司的行业是否适用？",
+  "guidelines": "答案必须是\"是\"或\"否\"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答\"未查询到相关数据\"。",
+  "evidence": [
+    "以山泽辰医疗器械公司所在行业为医药制造业",
+    "政策《四部门关于公布农业、建筑、医疗、矿山领域机器人典型应用场景名单的通知》适用行业为农、林、牧、渔业；建筑业；卫生和社会工作；采矿业"
+  ],
+  "milestone": {
+    "以山泽辰医疗器械公司所在行业": "医药制造业",
+    "政策适用行业": "农、林、牧、渔业；建筑业；卫生和社会工作；采矿业",
+    "是否适用": "否"
+  },
+  "steps": [
+    "从company_profile.csv中抽取以山泽辰医疗器械公司所在行业为医药制造业",
+    "从policy_resource.csv中抽取政策《四部门关于公布农业、建筑、医疗、矿山领域机器人典型应用场景名单的通知》的适用行业列表",
+    "判断企业所在行业是否命中政策适用行业，不命中则输出「否」"
+  ],
+  "steps_num": 3,
+  "answer": "否",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "easy",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "company_profile": "16c1efd1-fb26-4851-8f3d-0cd794cb5351"
+    },
+    {
+      "policy_resource": "d849b18e-d1b8-40ba-8c7b-66432e04620a"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium001_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium001_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a96dd0f35be14db7c565db2fdee94c05203af8dc
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium001_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium001",
+  "question": "What is the gap between the number of central ministry/agency policies issued by the Ministry of Commerce in the ministerial policies for the industry of Wu Li Hui Da Lian Suo Co., Ltd. and the number of local policies issued by the Hainan Province Department of Industry and Information Technology for the industry of Xing Ku Wen Gong Yi Mei Shu Pin Co., Ltd.?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Wu Li Hui Da Lian Suo Co., Ltd.: Wholesale and Retail",
+    "Wholesale and Retail ministerial policies - Ministry of Commerce - number of policies: 3",
+    "Industry of Xing Ku Wen Gong Yi Mei Shu Pin Co., Ltd.: Cultural, Arts, Sports and Entertainment Goods Manufacturing",
+    "Cultural, Arts, Sports and Entertainment Goods Manufacturing local policies - Hainan Province Department of Industry and Information Technology - number of policies: 1"
+  ],
+  "milestone": {
+    "Industry of Wu Li Hui Da Lian Suo Co., Ltd.": "Wholesale and Retail",
+    "Wholesale and Retail ministerial policies_Ministry of Commerce number of policies": 3,
+    "Industry of Xing Ku Wen Gong Yi Mei Shu Pin Co., Ltd.": "Cultural, Arts, Sports and Entertainment Goods Manufacturing",
+    "Cultural, Arts, Sports and Entertainment Goods Manufacturing local policies_Hainan Province Department of Industry and Information Technology number of policies": 1,
+    "Gap (Ministry of Commerce number of policies - Hainan Province Department of Industry and Information Technology number of policies)": 2.0
+  },
+  "answer": 2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that the industry of Wu Li Hui Da Lian Suo Co., Ltd. is Wholesale and Retail",
+    "Extract from policy_release_status.csv that in ministerial policies for Wholesale and Retail, the number of policies issued by the Ministry of Commerce is 3",
+    "Extract from company_profile.csv that the industry of Xing Ku Wen Gong Yi Mei Shu Pin Co., Ltd. is Cultural, Arts, Sports and Entertainment Goods Manufacturing",
+    "Extract from policy_release_status.csv that in local policies for Cultural, Arts, Sports and Entertainment Goods Manufacturing, the number of policies issued by the Hainan Province Department of Industry and Information Technology is 1",
+    "Compute the gap: 3 - 1 = 2.0"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "1f977ef9-1e22-45fc-8781-0f97b90702d3"
+    },
+    {
+      "company_profile": "24d73db3-b3be-4f9d-87e8-a8f7bcd7b0f9"
+    },
+    {
+      "policy_release_status": "ae96a1ad-d893-4619-aac3-ae9056fffb7e"
+    },
+    {
+      "company_profile": "8ac66182-6508-416b-b0a3-b7b5f71175eb"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium002_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium002_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c0ed022e629503beff58dd195aacef42b51b86b1
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium002_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium002",
+  "question": "What is the difference between the number of policies issued by the Gansu Province General Office of the People's Government in the local policies for the Wholesale and Retail industry and the number of policies issued by the Ministry of Transport in the ministerial policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without any unit, comma, or explanatory text. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Wu Li Hui Da Lian Suo Co., Ltd.: Wholesale and Retail",
+    "Wholesale and Retail local policies — Gansu Province General Office of the People's Government — number of policies: 1",
+    "Industry of Xing Ku Wen Gong Yi Mei Shu Pin Co., Ltd.: Cultural, Arts, Sports and Entertainment Goods Manufacturing",
+    "Cultural, Arts, Sports and Entertainment Goods Manufacturing ministerial policies — Ministry of Transport — number of policies: 1"
+  ],
+  "milestone": {
+    "Industry of Wu Li Hui Da Lian Suo Co., Ltd.": "Wholesale and Retail",
+    "Wholesale and Retail local policies_Gansu Province General Office of the People's Government number of policies": 1,
+    "Industry of Xing Ku Wen Gong Yi Mei Shu Pin Co., Ltd.": "Cultural, Arts, Sports and Entertainment Goods Manufacturing",
+    "Cultural, Arts, Sports and Entertainment Goods Manufacturing ministerial policies_Ministry of Transport number of policies": 1,
+    "Difference (Gansu Province General Office of the People's Government number of policies - Ministry of Transport number of policies)": 0.0
+  },
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Wu Li Hui Da Lian Suo Co., Ltd.'s industry is Wholesale and Retail",
+    "From policy_release_status.csv, extract that in the local policies for Wholesale and Retail, the number of policies issued by the Gansu Province General Office of the People's Government is 1",
+    "From company_profile.csv, extract that Xing Ku Wen Gong Yi Mei Shu Pin Co., Ltd.'s industry is Cultural, Arts, Sports and Entertainment Goods Manufacturing",
+    "From policy_release_status.csv, extract that in the ministerial policies for Cultural, Arts, Sports and Entertainment Goods Manufacturing, the number of policies issued by the Ministry of Transport is 1",
+    "Compute the difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "1f977ef9-1e22-45fc-8781-0f97b90702d3"
+    },
+    {
+      "company_profile": "24d73db3-b3be-4f9d-87e8-a8f7bcd7b0f9"
+    },
+    {
+      "policy_release_status": "ae96a1ad-d893-4619-aac3-ae9056fffb7e"
+    },
+    {
+      "company_profile": "8ac66182-6508-416b-b0a3-b7b5f71175eb"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium003_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium003_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..40e6c194c8aa49adff3cfe2fb38426f417110ffb
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium003_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium003",
+  "question": "Which is greater: the number of policies issued by the General Administration of Customs in the ministerial policies for the industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd., or the number of policies issued by the General Office of the State Council in the State Council policies for the industry of Bei Kong Ze Jing Water Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.: Financial Services",
+    "Financial Services ministerial policies_General Administration of Customs policy count = 1",
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.: Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management State Council policies_General Office of the State Council policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_General Administration of Customs policy count": 1,
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.": "Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management State Council policies_General Office of the State Council policy count": 1,
+    "Comparison result": "Both are equal; per question requirement, output Bei Kong Ze Jing Water Co., Ltd."
+  },
+  "answer": "Bei Kong Ze Jing Water Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Ying Tai Sheng Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in Financial Services ministerial policies, the General Administration of Customs policy count is 1",
+    "Extract from company_profile.csv that Bei Kong Ze Jing Water Co., Ltd.'s industry is Water Conservancy, Environment and Public Facilities Management",
+    "Extract from policy_release_status.csv that in Water Conservancy, Environment and Public Facilities Management State Council policies, the General Office of the State Council policy count is 1",
+    "Compare 1 and 1; when equal, output \"Bei Kong Ze Jing Water Co., Ltd.\" as required by the question"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "b3ffe2d5-c5b8-4499-b7e7-c6955e84e897"
+    },
+    {
+      "policy_release_status": "fbe8fd5e-a3b5-48a2-aef1-e94b5d88cb16"
+    },
+    {
+      "company_profile": "b32ab5eb-bfc5-4320-8527-8c31fd09ac36"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium004_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium004_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..cbd133bfde99d4b6a5b634c68744f50b4e8d710b
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium004_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium004",
+  "question": "Which is larger: the number of central ministry/agency policies issued by the Ministry of Agriculture and Rural Affairs in the ministerial policies for the Financial Services industry, or the number of central ministry/agency policies issued by the National Health Commission in the ministerial policies for the Water Conservancy, Environment and Public Facilities Management industry?",
+  "guidelines": "The answer must be a company name; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.: Financial Services",
+    "Financial Services ministerial policies — Ministry of Agriculture and Rural Affairs — number of policies: 1",
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.: Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management ministerial policies — National Health Commission — number of policies: 2"
+  ],
+  "milestone": {
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_Ministry of Agriculture and Rural Affairs number of policies": 1,
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.": "Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management ministerial policies_National Health Commission number of policies": 2,
+    "Comparison result": "2 is larger; output Bei Kong Ze Jing Water Co., Ltd."
+  },
+  "answer": "Bei Kong Ze Jing Water Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Hua Ying Tai Sheng Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "From policy_release_status.csv, extract that in the ministerial policies for Financial Services, the number of policies issued by the Ministry of Agriculture and Rural Affairs is 1",
+    "From company_profile.csv, extract that Bei Kong Ze Jing Water Co., Ltd.'s industry is Water Conservancy, Environment and Public Facilities Management",
+    "From policy_release_status.csv, extract that in the ministerial policies for Water Conservancy, Environment and Public Facilities Management, the number of policies issued by the National Health Commission is 2",
+    "Compare 1 and 2; 2 is larger; output \"Bei Kong Ze Jing Water Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "b3ffe2d5-c5b8-4499-b7e7-c6955e84e897"
+    },
+    {
+      "policy_release_status": "fbe8fd5e-a3b5-48a2-aef1-e94b5d88cb16"
+    },
+    {
+      "company_profile": "b32ab5eb-bfc5-4320-8527-8c31fd09ac36"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium005_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium005_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..66af6e339e68935bc0e964a4b3c48cb6415e4e14
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium005_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium005",
+  "question": "Is the number of central ministry/agency policies issued by the Development and Reform Commission in the ministerial policies for the industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. higher than the number of local policies issued by the Shenzhen Municipality Bureau of Commerce in the local policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd.?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies — Development and Reform Commission — number of policies: 2",
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.: Leasing and Business Services",
+    "Leasing and Business Services local policies — Shenzhen Municipality Bureau of Commerce — number of policies: 1"
+  ],
+  "milestone": {
+    "Industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies_Development and Reform Commission number of policies": 2,
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Leasing and Business Services local policies_Shenzhen Municipality Bureau of Commerce number of policies": 1,
+    "Whether higher (Development and Reform Commission number of policies > Shenzhen Municipality Bureau of Commerce number of policies)": "Yes"
+  },
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "From policy_release_status.csv, extract that in the ministerial policies for Information Transmission, Software and IT Services, the number of policies issued by the Development and Reform Commission is 2",
+    "From company_profile.csv, extract that Yi Hai Chang Jin Shang Wu Co., Ltd.'s industry is Leasing and Business Services",
+    "From policy_release_status.csv, extract that in the local policies for Leasing and Business Services, the number of policies issued by the Shenzhen Municipality Bureau of Commerce is 1",
+    "Determine whether 2 is greater than 1; if so, output \"Yes\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    },
+    {
+      "company_profile": "d6ab2642-2691-4f63-a197-6a4c29026c0d"
+    },
+    {
+      "policy_release_status": "b11a2b60-c999-4fc9-9581-5a6601001790"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium006_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium006_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..09d9f21acc127aef3f445dfe2ef9bb37f7e4a05b
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium006_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium006",
+  "question": "Which is greater: the number of local policies issued by the Yunnan Province Communications Administration in the local policies for the industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd., or the number of central ministry/agency policies issued by the Development and Reform Commission in the ministerial policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd.?",
+  "guidelines": "The answer must be either a company name or \"Equal\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies — Yunnan Province Communications Administration — number of policies: 1",
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.: Leasing and Business Services",
+    "Leasing and Business Services ministerial policies — Development and Reform Commission — number of policies: 1"
+  ],
+  "milestone": {
+    "Industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies_Yunnan Province Communications Administration number of policies": 1,
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Leasing and Business Services ministerial policies_Development and Reform Commission number of policies": 1,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "From policy_release_status.csv, extract that in the local policies for Information Transmission, Software and IT Services, the number of policies issued by the Yunnan Province Communications Administration is 1",
+    "From company_profile.csv, extract that Yi Hai Chang Jin Shang Wu Co., Ltd.'s industry is Leasing and Business Services",
+    "From policy_release_status.csv, extract that in the ministerial policies for Leasing and Business Services, the number of policies issued by the Development and Reform Commission is 1",
+    "Compare 1 and 1; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    },
+    {
+      "company_profile": "d6ab2642-2691-4f63-a197-6a4c29026c0d"
+    },
+    {
+      "policy_release_status": "b11a2b60-c999-4fc9-9581-5a6601001790"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium007_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium007_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..07a1eee2dc96408c74c9844a401b47d5f6f2a0bb
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium007_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium007",
+  "question": "What is the difference between the number of policies issued by the Ministry of Industry and Information Technology in the ministerial policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd. and the number of local policies issued by the Guangdong Provincial People's Government for the industry of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.: Leasing and Business Services",
+    "Leasing and Business Services ministerial policies_Ministry of Industry and Information Technology policy count = 2",
+    "Industry of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.: Health and Social Work",
+    "Health and Social Work local policies_Guangdong Provincial People's Government policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Leasing and Business Services ministerial policies_Ministry of Industry and Information Technology policy count": 2,
+    "Industry of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.": "Health and Social Work",
+    "Health and Social Work local policies_Guangdong Provincial People's Government policy count": 1,
+    "Difference (Ministry of Industry and Information Technology policy count - Guangdong Provincial People's Government policy count)": 1.0
+  },
+  "answer": 1.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Yi Hai Chang Jin Shang Wu Co., Ltd.'s industry is Leasing and Business Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Leasing and Business Services, the Ministry of Industry and Information Technology policy count is 2",
+    "Extract from company_profile.csv that Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.'s industry is Health and Social Work",
+    "Extract from policy_release_status.csv that in the local policies for Health and Social Work, the Guangdong Provincial People's Government policy count is 1",
+    "Calculate the difference: 2 - 1 = 1.0"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "b11a2b60-c999-4fc9-9581-5a6601001790"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    },
+    {
+      "policy_release_status": "eb7febde-e468-4a11-ad1d-2d9e5daab31a"
+    },
+    {
+      "company_profile": "ca56f5f4-4ea0-433d-9eb2-cf2c630fc69d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium008_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium008_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..dadd17694456f83a23e144d993d6d7d7f50f6bf2
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium008_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium008",
+  "question": "Which is greater: the number of local policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd., or the number of policies issued by the General Office of the State Council in the State Council policies for the industry of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"industry\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.: Leasing and Business Services",
+    "Leasing and Business Services local policies policy count = 12",
+    "Industry of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.: Health and Social Work",
+    "Health and Social Work State Council policies - General Office of the State Council - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Leasing and Business Services local policies policy count": 12,
+    "Industry of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.": "Health and Social Work",
+    "Health and Social Work State Council policies_General Office of the State Council policy count": 1,
+    "Comparison result": "12 is larger; output Yi Hai Chang Jin Shang Wu Co., Ltd."
+  },
+  "answer": "Yi Hai Chang Jin Shang Wu Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Yi Hai Chang Jin Shang Wu Co., Ltd.'s industry is Leasing and Business Services",
+    "Extract from policy_release_status.csv that the local policies policy count for Leasing and Business Services is 12",
+    "Extract from company_profile.csv that Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.'s industry is Health and Social Work",
+    "Extract from policy_release_status.csv that in Health and Social Work State Council policies, the policy count issued by the General Office of the State Council is 1",
+    "Compare 12 and 1; 12 is larger; output \"Yi Hai Chang Jin Shang Wu Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "b11a2b60-c999-4fc9-9581-5a6601001790"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    },
+    {
+      "policy_release_status": "eb7febde-e468-4a11-ad1d-2d9e5daab31a"
+    },
+    {
+      "company_profile": "ca56f5f4-4ea0-433d-9eb2-cf2c630fc69d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium009_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium009_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..eac7d63bde7cde697a7ddd189c2087f919887051
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium009_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium009",
+  "question": "Which is greater: the number of local policies issued by the Guangdong Province General Office of the People's Government in the local policies for the industry of Zhang Qiao Jin Chuang Technology Co., Ltd., or the number of central ministry/agency policies issued by the National Cryptography Administration in the ministerial policies for the industry of Heng Li Ke Zhi Ruan Jian Co., Ltd.?",
+  "guidelines": "The answer must be either \"Number of local policies issued by the Guangdong Province General Office of the People's Government for the industry of Zhang Qiao Jin Chuang Technology Co., Ltd.\" or \"Number of central ministry/agency policies issued by the National Cryptography Administration for the industry of Heng Li Ke Zhi Ruan Jian Co., Ltd.\" or \"Equal\". Output only the selected answer text, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhang Qiao Jin Chuang Technology Co., Ltd.: Consumer Electronics and Electrical Equipment",
+    "Consumer Electronics and Electrical Equipment local policies — Guangdong Province General Office of the People's Government — number of policies: 1",
+    "Industry of Heng Li Ke Zhi Ruan Jian Co., Ltd.: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies — National Cryptography Administration — number of policies: 1"
+  ],
+  "milestone": {
+    "Industry of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Consumer Electronics and Electrical Equipment",
+    "Consumer Electronics and Electrical Equipment local policies_Guangdong Province General Office of the People's Government number of policies": 1,
+    "Industry of Heng Li Ke Zhi Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies_National Cryptography Administration number of policies": 1,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhang Qiao Jin Chuang Technology Co., Ltd.'s industry is Consumer Electronics and Electrical Equipment",
+    "From policy_release_status.csv, extract that in the local policies for Consumer Electronics and Electrical Equipment, the number of policies issued by the Guangdong Province General Office of the People's Government is 1",
+    "From company_profile.csv, extract that Heng Li Ke Zhi Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "From policy_release_status.csv, extract that in the ministerial policies for Information Transmission, Software and IT Services, the number of policies issued by the National Cryptography Administration is 1",
+    "Compare 1 and 1; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "d3414a8b-ce02-4229-b6f0-34e0aa208f3c"
+    },
+    {
+      "company_profile": "58eb73a2-c225-4c83-9324-b515fec3a76f"
+    },
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    },
+    {
+      "company_profile": "6c1bc3d3-0763-4686-bb01-a61e146eee63"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium010_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium010_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0135f8c62a06b3b68f4a85c7d7f92ca43f0e2b75
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium010_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium010",
+  "question": "Which is greater: the number of local policies issued by the Hainan Province Department of Finance for the industry of Zhang Qiao Jin Chuang Technology Co., Ltd., or the number of central ministry/agency policies issued by the National Health Commission for the industry of Heng Li Ke Zhi Ruan Jian Co., Ltd.?",
+  "guidelines": "The answer must be a company name or \"Equal\". Output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhang Qiao Jin Chuang Technology Co., Ltd.: Consumer Electronics and Electrical Equipment",
+    "Consumer Electronics and Electrical Equipment local policies - Hainan Province Department of Finance - policy count = 1",
+    "Industry of Heng Li Ke Zhi Ruan Jian Co., Ltd.: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies - National Health Commission - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Consumer Electronics and Electrical Equipment",
+    "Consumer Electronics and Electrical Equipment local policies_Hainan Province Department of Finance policy count": 1,
+    "Industry of Heng Li Ke Zhi Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies_National Health Commission policy count": 1,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhang Qiao Jin Chuang Technology Co., Ltd.'s industry is Consumer Electronics and Electrical Equipment",
+    "Extract from policy_release_status.csv that in the local policies for Consumer Electronics and Electrical Equipment, the policy count issued by the Hainan Province Department of Finance is 1",
+    "Extract from company_profile.csv that Heng Li Ke Zhi Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Information Transmission, Software and IT Services, the policy count issued by the National Health Commission is 1",
+    "Compare 1 and 1; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "d3414a8b-ce02-4229-b6f0-34e0aa208f3c"
+    },
+    {
+      "company_profile": "58eb73a2-c225-4c83-9324-b515fec3a76f"
+    },
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    },
+    {
+      "company_profile": "6c1bc3d3-0763-4686-bb01-a61e146eee63"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium011_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium011_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d3d42f66434da44c08e14a73127beb0fb490e988
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium011_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium011",
+  "question": "What is the difference between the number of central ministry/agency policies issued by the State Administration of Foreign Exchange in the ministerial policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. and the number of local policies issued by the Guangdong Provincial Committee of the CPC for the Financial Services industry in Guangdong Province where Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies - State Administration of Foreign Exchange - policy count = 1",
+    "Province of Hui Jin Jin Rui Cai Fu Management Co., Ltd.: Guangdong Province",
+    "Guangdong Province Financial Services local policies - Guangdong Provincial Committee of the CPC - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies_State Administration of Foreign Exchange policy count": 1,
+    "Province of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Financial Services local policies_Guangdong Provincial Committee of the CPC policy count": 1,
+    "Difference (State Administration of Foreign Exchange policy count - Guangdong Provincial Committee of the CPC policy count)": 0.0
+  },
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "From policy_release_status.csv, extract that in the ministerial policies for Information Transmission, Software and IT Services, the policy count issued by the State Administration of Foreign Exchange is 1",
+    "From company_profile.csv, extract that Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located in Guangdong Province and its industry is Financial Services",
+    "From policy_release_status.csv, filter by province = Guangdong Province and industry = Financial Services, and extract that in local policies, the policy count issued by the Guangdong Provincial Committee of the CPC is 1",
+    "Compute the difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "policy_release_status": "12b8a701-365a-430a-b6db-13c043557d2b"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium012_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium012_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6c7c43fdb3eda608b20b6a326bbcb1c35e73ca1
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium012_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium012",
+  "question": "Which is greater: the number of central ministry/agency policies issued by the People's Bank of China in the ministerial policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd., or the number of local policies issued by the Guangdong Provincial Committee of the CPC for the Financial Services industry in Guangdong Province where Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"Equal\"; output only the selected answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies - People's Bank of China - policy count = 1",
+    "Province of Hui Jin Jin Rui Cai Fu Management Co., Ltd.: Guangdong Province",
+    "Guangdong Province Financial Services local policies - Guangdong Provincial Committee of the CPC - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies_People's Bank of China policy count": 1,
+    "Province of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Financial Services local policies_Guangdong Provincial Committee of the CPC policy count": 1,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "From policy_release_status.csv, extract that in the ministerial policies for Information Transmission, Software and IT Services, the policy count issued by the People's Bank of China is 1",
+    "From company_profile.csv, extract that Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located in Guangdong Province and its industry is Financial Services",
+    "From policy_release_status.csv, filter by province = Guangdong Province and industry = Financial Services, and extract that in local policies, the policy count issued by the Guangdong Provincial Committee of the CPC is 1",
+    "Compare 1 and 1; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "policy_release_status": "12b8a701-365a-430a-b6db-13c043557d2b"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium013_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium013_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..24b803c2434aa68deafe6af4dd22813f57e8b405
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium013_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium013",
+  "question": "Which is greater: the number of local policies issued by the Yunnan Province Development and Reform Commission for the industry of Bei Kong Ze Jing Water Co., Ltd., or the number of local policies issued by the Shanghai Municipality Science and Technology Commission for the Health and Social Work industry in Shanghai Municipality where Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"Equal\"; output only the selected answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.: Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management local policies - Yunnan Province Development and Reform Commission - policy count = 2",
+    "Province of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.: Shanghai Municipality",
+    "Shanghai Municipality Health and Social Work local policies - Shanghai Municipality Science and Technology Commission - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.": "Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management local policies_Yunnan Province Development and Reform Commission policy count": 2,
+    "Province of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Health and Social Work local policies_Shanghai Municipality Science and Technology Commission policy count": 1,
+    "Comparison result": "2 is greater; output Bei Kong Ze Jing Water Co., Ltd."
+  },
+  "answer": "Bei Kong Ze Jing Water Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Bei Kong Ze Jing Water Co., Ltd.'s industry is Water Conservancy, Environment and Public Facilities Management",
+    "Extract from policy_release_status.csv that in the local policies for Water Conservancy, Environment and Public Facilities Management, the policy count issued by the Yunnan Province Development and Reform Commission is 2",
+    "Extract from company_profile.csv that Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd. is located in Shanghai Municipality and its industry is Health and Social Work",
+    "Extract from policy_release_status.csv that in the local policies for Health and Social Work in Shanghai Municipality, the policy count issued by the Shanghai Municipality Science and Technology Commission is 1",
+    "Compare 2 and 1; 2 is greater; output \"Bei Kong Ze Jing Water Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "fbe8fd5e-a3b5-48a2-aef1-e94b5d88cb16"
+    },
+    {
+      "company_profile": "b32ab5eb-bfc5-4320-8527-8c31fd09ac36"
+    },
+    {
+      "policy_release_status": "27e08f44-5d2e-4681-addd-d501571e641f"
+    },
+    {
+      "company_profile": "ca56f5f4-4ea0-433d-9eb2-cf2c630fc69d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium014_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium014_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..469e3f4de0fef3d6529029f647905198825a2c80
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium014_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium014",
+  "question": "Which is higher: the number of central ministry/agency policies issued by the Ministry of Agriculture and Rural Affairs in the ministerial policies for the industry of Bei Kong Ze Jing Water Co., Ltd., or the number of local policies issued by the Shanghai Municipality Health Commission for the Health and Social Work industry in Shanghai Municipality where Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.: Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management ministerial policies - Ministry of Agriculture and Rural Affairs - policy count = 2",
+    "Province of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.: Shanghai Municipality",
+    "Shanghai Municipality Health and Social Work local policies - Shanghai Municipality Health Commission - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.": "Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management ministerial policies_Ministry of Agriculture and Rural Affairs policy count": 2,
+    "Province of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Health and Social Work local policies_Shanghai Municipality Health Commission policy count": 1,
+    "Comparison result": "2 is higher; output Bei Kong Ze Jing Water Co., Ltd."
+  },
+  "answer": "Bei Kong Ze Jing Water Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Bei Kong Ze Jing Water Co., Ltd.'s industry is Water Conservancy, Environment and Public Facilities Management",
+    "Extract from policy_release_status.csv that in the ministerial policies for this industry, the policy count issued by the Ministry of Agriculture and Rural Affairs is 2",
+    "Extract from company_profile.csv that Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd. is located in Shanghai Municipality and its industry is Health and Social Work",
+    "Extract from policy_release_status.csv that in the local policies for Health and Social Work in Shanghai Municipality, the policy count issued by the Shanghai Municipality Health Commission is 1",
+    "Compare 2 and 1; 2 is higher; output \"Bei Kong Ze Jing Water Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "fbe8fd5e-a3b5-48a2-aef1-e94b5d88cb16"
+    },
+    {
+      "company_profile": "b32ab5eb-bfc5-4320-8527-8c31fd09ac36"
+    },
+    {
+      "policy_release_status": "27e08f44-5d2e-4681-addd-d501571e641f"
+    },
+    {
+      "company_profile": "ca56f5f4-4ea0-433d-9eb2-cf2c630fc69d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium015_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium015_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..eab0ae5c17d679eb8281a697e39486f53b0195b8
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium015_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium015",
+  "question": "Which is greater: the number of central ministry/agency policies issued by the State Administration of Foreign Exchange in the ministerial policies for the industry of Wu Li Hui Da Lian Suo Co., Ltd., or the number of local policies for the Water Conservancy, Environment and Public Facilities Management industry in Guangdong Province where Bei Kong Ze Jing Water Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"Equal\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Wu Li Hui Da Lian Suo Co., Ltd.: Wholesale and Retail",
+    "Wholesale and Retail ministerial policies - State Administration of Foreign Exchange - policy count = 1",
+    "Province of Bei Kong Ze Jing Water Co., Ltd.: Guangdong Province",
+    "Guangdong Province Water Conservancy, Environment and Public Facilities Management local policies policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Wu Li Hui Da Lian Suo Co., Ltd.": "Wholesale and Retail",
+    "Wholesale and Retail ministerial policies_State Administration of Foreign Exchange policy count": 1,
+    "Province of Bei Kong Ze Jing Water Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Water Conservancy, Environment and Public Facilities Management local policies policy count": 1,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Wu Li Hui Da Lian Suo Co., Ltd.'s industry is Wholesale and Retail",
+    "From policy_release_status.csv, extract that in the ministerial policies for Wholesale and Retail, the policy count issued by the State Administration of Foreign Exchange is 1",
+    "From company_profile.csv, extract that Bei Kong Ze Jing Water Co., Ltd. is located in Guangdong Province and its industry is Water Conservancy, Environment and Public Facilities Management",
+    "From policy_release_status.csv, filter by province = Guangdong Province and industry = Water Conservancy, Environment and Public Facilities Management, and extract that the local policies policy count is 1",
+    "Compare 1 and 1; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "1f977ef9-1e22-45fc-8781-0f97b90702d3"
+    },
+    {
+      "company_profile": "24d73db3-b3be-4f9d-87e8-a8f7bcd7b0f9"
+    },
+    {
+      "policy_release_status": "cf35b82a-2d81-4b2e-a02b-96f40e65e597"
+    },
+    {
+      "company_profile": "b32ab5eb-bfc5-4320-8527-8c31fd09ac36"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium016_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium016_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..470170a1b90d25d15b9daf3a1225b07ee0785c58
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium016_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium016",
+  "question": "Which is greater: the number of local policies issued by the Shenzhen Municipal People's Government for the industry of Wu Li Hui Da Lian Suo Co., Ltd., or the total number of policies for the Water Conservancy, Environment and Public Facilities Management industry in Guangdong Province where Bei Kong Ze Jing Water Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"Equal\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Wu Li Hui Da Lian Suo Co., Ltd.: Wholesale and Retail",
+    "Wholesale and Retail local policies - Shenzhen Municipal People's Government - policy count = 1",
+    "Province of Bei Kong Ze Jing Water Co., Ltd.: Guangdong Province",
+    "Guangdong Province Water Conservancy, Environment and Public Facilities Management total policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Wu Li Hui Da Lian Suo Co., Ltd.": "Wholesale and Retail",
+    "Wholesale and Retail local policies_Shenzhen Municipal People's Government policy count": 1,
+    "Province of Bei Kong Ze Jing Water Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Water Conservancy, Environment and Public Facilities Management total policy count": 1,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Wu Li Hui Da Lian Suo Co., Ltd.'s industry is Wholesale and Retail",
+    "From policy_release_status.csv, extract that in the local policies for Wholesale and Retail, the policy count issued by the Shenzhen Municipal People's Government is 1",
+    "From company_profile.csv, extract that Bei Kong Ze Jing Water Co., Ltd. is located in Guangdong Province and its industry is Water Conservancy, Environment and Public Facilities Management",
+    "From policy_release_status.csv, extract that the total policy count for Water Conservancy, Environment and Public Facilities Management in Guangdong Province is 1",
+    "Compare 1 and 1; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "1f977ef9-1e22-45fc-8781-0f97b90702d3"
+    },
+    {
+      "company_profile": "24d73db3-b3be-4f9d-87e8-a8f7bcd7b0f9"
+    },
+    {
+      "policy_release_status": "cf35b82a-2d81-4b2e-a02b-96f40e65e597"
+    },
+    {
+      "company_profile": "b32ab5eb-bfc5-4320-8527-8c31fd09ac36"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium017_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium017_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2bc769151532952762e30857485c365e825dfd41
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium017_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium017",
+  "question": "Which is greater: the number of central ministry/agency policies issued by the National Health Commission in the ministerial policies for the industry of Bei Kong Ze Jing Water Co., Ltd., or the number of local policies issued by the Guangdong Provincial Committee of the CPC for the Financial Services industry in Guangdong Province where Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.: Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management ministerial policies - National Health Commission - policy count = 2",
+    "Province of Hui Jin Jin Rui Cai Fu Management Co., Ltd.: Guangdong Province",
+    "Guangdong Province Financial Services local policies - Guangdong Provincial Committee of the CPC - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.": "Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management ministerial policies_National Health Commission policy count": 2,
+    "Province of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Financial Services local policies_Guangdong Provincial Committee of the CPC policy count": 1,
+    "Comparison result": "2 is greater; output Bei Kong Ze Jing Water Co., Ltd."
+  },
+  "answer": "Bei Kong Ze Jing Water Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Bei Kong Ze Jing Water Co., Ltd.'s industry is Water Conservancy, Environment and Public Facilities Management",
+    "Extract from policy_release_status.csv that in the ministerial policies for this industry, the policy count issued by the National Health Commission is 2",
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located in Guangdong Province and its industry is Financial Services",
+    "Extract from policy_release_status.csv that in Guangdong Province Financial Services local policies, the policy count issued by the Guangdong Provincial Committee of the CPC is 1",
+    "Compare 2 and 1; 2 is greater; output \"Bei Kong Ze Jing Water Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "fbe8fd5e-a3b5-48a2-aef1-e94b5d88cb16"
+    },
+    {
+      "company_profile": "b32ab5eb-bfc5-4320-8527-8c31fd09ac36"
+    },
+    {
+      "policy_release_status": "12b8a701-365a-430a-b6db-13c043557d2b"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium018_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium018_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d45dc26f977c80dd877496072ad8d39c0bfee228
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium018_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium018",
+  "question": "Which is greater: the number of local policies issued by the Liaoning Province People's Government for the industry of Bei Kong Ze Jing Water Co., Ltd., or the number of local policies issued by the Guangdong Province General Office of the People's Government for the Financial Services industry in Guangdong Province where Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.: Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management local policies - Liaoning Province People's Government - policy count = 1",
+    "Province of Hui Jin Jin Rui Cai Fu Management Co., Ltd.: Guangdong Province",
+    "Guangdong Province Financial Services local policies - Guangdong Province General Office of the People's Government - policy count = 2"
+  ],
+  "milestone": {
+    "Industry of Bei Kong Ze Jing Water Co., Ltd.": "Water Conservancy, Environment and Public Facilities Management",
+    "Water Conservancy, Environment and Public Facilities Management local policies_Liaoning Province People's Government policy count": 1,
+    "Province of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Financial Services local policies_Guangdong Province General Office of the People's Government policy count": 2,
+    "Comparison result": "2 is greater; output Hui Jin Jin Rui Cai Fu Management Co., Ltd."
+  },
+  "answer": "Hui Jin Jin Rui Cai Fu Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Bei Kong Ze Jing Water Co., Ltd.'s industry is Water Conservancy, Environment and Public Facilities Management",
+    "Extract from policy_release_status.csv that in the local policies for this industry, the policy count issued by the Liaoning Province People's Government is 1",
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located in Guangdong Province and its industry is Financial Services",
+    "Extract from policy_release_status.csv that in the local policies for Financial Services in Guangdong Province, the policy count issued by the Guangdong Province General Office of the People's Government is 2",
+    "Compare 1 and 2; 2 is greater; output \"Hui Jin Jin Rui Cai Fu Management Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "fbe8fd5e-a3b5-48a2-aef1-e94b5d88cb16"
+    },
+    {
+      "company_profile": "b32ab5eb-bfc5-4320-8527-8c31fd09ac36"
+    },
+    {
+      "policy_release_status": "12b8a701-365a-430a-b6db-13c043557d2b"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium019_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium019_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..42a8d61f8c5977cbe2a770fb7fa2b5fe4d3a6a92
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium019_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium019",
+  "question": "What is the difference between the number of local policies issued by the Shaanxi Province Development and Reform Commission for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. and the total number of policies for the Conglomerates industry in Guangdong Province where Hua Cheng Sheng Yuan Integrated Development Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies - Shaanxi Province Development and Reform Commission - policy count = 1",
+    "Province of Hua Cheng Sheng Yuan Integrated Development Co., Ltd.: Guangdong Province",
+    "Guangdong Province Conglomerates total policy count = 2"
+  ],
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies_Shaanxi Province Development and Reform Commission policy count": 1,
+    "Province of Hua Cheng Sheng Yuan Integrated Development Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Conglomerates total policy count": 2,
+    "Difference (Shaanxi Province Development and Reform Commission policy count - total policy count)": -1.0
+  },
+  "answer": -1.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for this industry in Shaanxi Province, the policy count issued by the Shaanxi Province Development and Reform Commission is 1",
+    "Extract from company_profile.csv that Hua Cheng Sheng Yuan Integrated Development Co., Ltd. is located in Guangdong Province and its industry is Conglomerates",
+    "Extract from policy_release_status.csv that the total policy count for Conglomerates in Guangdong Province is 2",
+    "Calculate the difference: 1 - 2 = -1.0"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "policy_release_status": "bd9a173a-9396-4aba-b4a2-059bc01cec3d"
+    },
+    {
+      "company_profile": "41fc40f8-35ba-4a45-9c72-56aa081d8caa"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium020_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium020_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e6244dbd19168217fd1353c6c3e7125da66fa39
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium020_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium020",
+  "question": "Which is greater: the number of central ministry/agency policies issued by the Ministry of Housing and Urban-Rural Development in the ministerial policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd., or the number of local policies for the Conglomerates industry in Guangdong Province where Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"Equal\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies - Ministry of Housing and Urban-Rural Development - policy count = 2",
+    "Province of Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd.: Guangdong Province",
+    "Guangdong Province Conglomerates local policies policy count = 2"
+  ],
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies_Ministry of Housing and Urban-Rural Development policy count": 2,
+    "Province of Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Conglomerates local policies policy count": 2,
+    "Comparison result": "Equal"
+  },
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for this industry, the policy count issued by the Ministry of Housing and Urban-Rural Development is 2",
+    "Extract from company_profile.csv that Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. is located in Guangdong Province and its industry is Conglomerates",
+    "Extract from policy_release_status.csv that in Guangdong Province, the Conglomerates local policies policy count is 2",
+    "Compare 2 and 2; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "policy_release_status": "bd9a173a-9396-4aba-b4a2-059bc01cec3d"
+    },
+    {
+      "company_profile": "41fc40f8-35ba-4a45-9c72-56aa081d8caa"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium021_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium021_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..60067e1b1d1e6df007364d9bb384b97c02231784
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium021_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium021",
+  "question": "Which is greater: the number of specific local policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd., or the number of similar local policies for the industry of Bao Xin Hui Hui Wang Luo Co., Ltd. in its province?",
+  "guidelines": "The answer must be a company name or \"industry\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies - Sichuan Province People's Government - policy count = 2",
+    "Province of Bao Xin Hui Hui Wang Luo Co., Ltd.: Beijing Municipality; industry: Information Transmission, Software and IT Services",
+    "Beijing Municipality Information Transmission, Software and IT Services local policies - Beijing Municipality Bureau of Economy and Information Technology - policy count = 7"
+  ],
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies_Sichuan Province People's Government policy count": 2,
+    "Province of Bao Xin Hui Hui Wang Luo Co., Ltd.": "Beijing Municipality",
+    "Beijing Municipality Information Transmission, Software and IT Services local policies_Beijing Municipality Bureau of Economy and Information Technology policy count": 7,
+    "Comparison result": "7 is greater; output Bao Xin Hui Hui Wang Luo Co., Ltd."
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for this industry, the policy count issued by the Sichuan Province People's Government is 2",
+    "Extract from company_profile.csv that Bao Xin Hui Hui Wang Luo Co., Ltd. is located in Beijing Municipality and its industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for this industry in Beijing Municipality, the policy count issued by the Beijing Municipality Bureau of Economy and Information Technology is 7",
+    "Compare 2 and 7; 7 is greater; output \"Bao Xin Hui Hui Wang Luo Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "answer": "Bao Xin Hui Hui Wang Luo Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "policy_release_status": "6fbacb2a-b768-4b54-a394-b87639684041"
+    },
+    {
+      "company_profile": "fe635f7d-bee1-4e2e-9ed6-3923871590d0"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium022_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium022_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4606cbad425d8b924ebc77bb819124a02f05f551
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium022_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium022",
+  "question": "Between Zhong Ke Ke Shu Ruan Jian Co., Ltd. and Bao Xin Hui Hui Wang Luo Co., Ltd., which obtains a greater number of local policy supports?",
+  "guidelines": "The answer must be a company name or \"Equal\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies - Shenzhen Municipality Science and Technology Innovation Commission - policy count = 3",
+    "Province of Bao Xin Hui Hui Wang Luo Co., Ltd.: Beijing Municipality; industry: Information Transmission, Software and IT Services",
+    "Beijing Municipality Information Transmission, Software and IT Services local policies - Beijing Municipality Science and Technology Commission - policy count = 3"
+  ],
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies_Shenzhen Municipality Science and Technology Innovation Commission policy count": 3,
+    "Province of Bao Xin Hui Hui Wang Luo Co., Ltd.": "Beijing Municipality",
+    "Beijing Municipality Information Transmission, Software and IT Services local policies_Beijing Municipality Science and Technology Commission policy count": 3,
+    "Comparison result": "Equal"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for this industry, the policy count issued by the Shenzhen Municipality Science and Technology Innovation Commission is 3",
+    "Extract from company_profile.csv that Bao Xin Hui Hui Wang Luo Co., Ltd. is located in Beijing Municipality and its industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for this industry in Beijing Municipality, the policy count issued by the Beijing Municipality Science and Technology Commission is 3",
+    "Compare 3 and 3; if equal, output \"Equal\""
+  ],
+  "steps_num": 5,
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "policy_release_status": "6fbacb2a-b768-4b54-a394-b87639684041"
+    },
+    {
+      "company_profile": "fe635f7d-bee1-4e2e-9ed6-3923871590d0"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium023_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium023_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..1999d24e92ddafc3c9f0c7e595da2abcd6d91f21
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium023_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium023",
+  "question": "What is the difference between the number of local policies issued by the Hunan Province General Office of the People's Government for the industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. and the total number of policies for the Consumer Electronics and Electrical Equipment industry in Guangdong Province where Zhang Qiao Jin Chuang Technology Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.: Real Estate",
+    "Real Estate local policies - Hunan Province General Office of the People's Government - policy count = 1",
+    "Province of Zhang Qiao Jin Chuang Technology Co., Ltd.: Guangdong Province; industry: Consumer Electronics and Electrical Equipment",
+    "Guangdong Province Consumer Electronics and Electrical Equipment total policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.": "Real Estate",
+    "Real Estate local policies_Hunan Province General Office of the People's Government policy count": 1,
+    "Province of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Consumer Electronics and Electrical Equipment total policy count": 1,
+    "Difference (Hunan Province General Office of the People's Government policy count - total policy count)": 0.0
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.'s industry is Real Estate",
+    "Extract from policy_release_status.csv that in the local policies for Real Estate, the policy count issued by the Hunan Province General Office of the People's Government is 1",
+    "Extract from company_profile.csv that Zhang Qiao Jin Chuang Technology Co., Ltd. is located in Guangdong Province and its industry is Consumer Electronics and Electrical Equipment",
+    "Extract from policy_release_status.csv that the total policy count for Consumer Electronics and Electrical Equipment in Guangdong Province is 1",
+    "Calculate the difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 5,
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "7e08e8dd-76ba-4683-9e6b-4b48d94f0d3c"
+    },
+    {
+      "company_profile": "3f2c07ea-c142-43aa-ab56-ee76d8aaa4e5"
+    },
+    {
+      "policy_release_status": "558608bf-1ac0-4a8b-8285-011f4eb6f845"
+    },
+    {
+      "company_profile": "58eb73a2-c225-4c83-9324-b515fec3a76f"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium024_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium024_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..24209b53235415f2711fab7801b86a30e027d843
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium024_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium024",
+  "question": "Which is greater: the number of central ministry/agency policies issued by the Ministry of Science and Technology in the ministerial policies for the industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd., or the number of local policies for the Consumer Electronics and Electrical Equipment industry in Guangdong Province where Zhang Qiao Jin Chuang Technology Co., Ltd. is located?",
+  "guidelines": "The answer must be a company name or \"industry\"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.: Real Estate",
+    "Real Estate ministerial policies - Ministry of Science and Technology - policy count = 1",
+    "Province of Zhang Qiao Jin Chuang Technology Co., Ltd.: Guangdong Province; industry: Consumer Electronics and Electrical Equipment",
+    "Guangdong Province Consumer Electronics and Electrical Equipment local policies policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.": "Real Estate",
+    "Real Estate ministerial policies_Ministry of Science and Technology policy count": 1,
+    "Province of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Consumer Electronics and Electrical Equipment local policies policy count": 1,
+    "Comparison result": "Equal; per question requirement, output Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd."
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.'s industry is Real Estate",
+    "Extract from policy_release_status.csv that in the ministerial policies for Real Estate, the policy count issued by the Ministry of Science and Technology is 1",
+    "Extract from company_profile.csv that Zhang Qiao Jin Chuang Technology Co., Ltd. is located in Guangdong Province and its industry is Consumer Electronics and Electrical Equipment",
+    "Extract from policy_release_status.csv that the local policies policy count for Consumer Electronics and Electrical Equipment in Guangdong Province is 1",
+    "Compare 1 and 1; if equal, output \"Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.\" as required by the question"
+  ],
+  "steps_num": 5,
+  "answer": "Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "7e08e8dd-76ba-4683-9e6b-4b48d94f0d3c"
+    },
+    {
+      "company_profile": "3f2c07ea-c142-43aa-ab56-ee76d8aaa4e5"
+    },
+    {
+      "policy_release_status": "558608bf-1ac0-4a8b-8285-011f4eb6f845"
+    },
+    {
+      "company_profile": "58eb73a2-c225-4c83-9324-b515fec3a76f"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium025_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium025_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..139b41d07679d8f560c1ff02f9184ecb207e4191
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium025_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium025",
+  "question": "What is the difference between the number of local policies issued by the Anhui Province People's Government for the industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd. and the total number of local policies for the Consumer Electronics and Electrical Equipment industry in Guangdong Province where Zhang Qiao Jin Chuang Technology Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.: Financial Services",
+    "Financial Services local policies - Anhui Province People's Government - policy count = 1",
+    "Province of Zhang Qiao Jin Chuang Technology Co., Ltd.: Guangdong Province; industry: Consumer Electronics and Electrical Equipment",
+    "Guangdong Province Consumer Electronics and Electrical Equipment local policies policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services local policies_Anhui Province People's Government policy count": 1,
+    "Province of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Consumer Electronics and Electrical Equipment local policies policy count": 1,
+    "Difference (Anhui Province People's Government policy count - local policies policy count)": 0.0
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Ying Tai Sheng Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the local policies for Financial Services, the policy count issued by the Anhui Province People's Government is 1",
+    "Extract from company_profile.csv that Zhang Qiao Jin Chuang Technology Co., Ltd. is located in Guangdong Province and its industry is Consumer Electronics and Electrical Equipment",
+    "Extract from policy_release_status.csv that the local policies policy count for Consumer Electronics and Electrical Equipment in Guangdong Province is 1",
+    "Calculate the difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 5,
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "b3ffe2d5-c5b8-4499-b7e7-c6955e84e897"
+    },
+    {
+      "policy_release_status": "558608bf-1ac0-4a8b-8285-011f4eb6f845"
+    },
+    {
+      "company_profile": "58eb73a2-c225-4c83-9324-b515fec3a76f"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium026_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium026_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ab7116f63d2d4633e751dd1a292041a4eec47dee
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium026_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium026",
+  "question": "What is the difference between the number of central ministry/agency policies issued by the General Office of the China Banking and Insurance Regulatory Commission in the ministerial policies for the industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd. and the number of local policies for the Consumer Electronics and Electrical Equipment industry in Guangdong Province where Zhang Qiao Jin Chuang Technology Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.: Financial Services",
+    "Financial Services ministerial policies - General Office of the China Banking and Insurance Regulatory Commission - policy count = 1",
+    "Province of Zhang Qiao Jin Chuang Technology Co., Ltd.: Guangdong Province; industry: Consumer Electronics and Electrical Equipment",
+    "Guangdong Province Consumer Electronics and Electrical Equipment local policies policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_General Office of the China Banking and Insurance Regulatory Commission policy count": 1,
+    "Province of Zhang Qiao Jin Chuang Technology Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Consumer Electronics and Electrical Equipment local policies policy count": 1,
+    "Difference (General Office of the China Banking and Insurance Regulatory Commission policy count - local policies policy count)": 0.0
+  },
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Ying Tai Sheng Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Financial Services, the policy count issued by the General Office of the China Banking and Insurance Regulatory Commission is 1",
+    "Extract from company_profile.csv that Zhang Qiao Jin Chuang Technology Co., Ltd. is located in Guangdong Province and its industry is Consumer Electronics and Electrical Equipment",
+    "Extract from policy_release_status.csv that the local policies policy count for Consumer Electronics and Electrical Equipment in Guangdong Province is 1",
+    "Calculate the difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "b3ffe2d5-c5b8-4499-b7e7-c6955e84e897"
+    },
+    {
+      "policy_release_status": "558608bf-1ac0-4a8b-8285-011f4eb6f845"
+    },
+    {
+      "company_profile": "58eb73a2-c225-4c83-9324-b515fec3a76f"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium027_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium027_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff8283665846186d7102a1835ca81aceb51648f2
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium027_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium027",
+  "question": "Between the number of local policies issued by the General Office of the People's Government of Henan Province in the local policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd. and the number of local policies issued by the Shanghai Municipality Finance Bureau in the local policies for the industry of Lang Ji Hui Ruan Technology Co., Ltd. in Shanghai, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.: Leasing and Business Services",
+    "Leasing and Business Services local policies - Henan Province General Office of the People's Government - policy count = 1",
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.: Shanghai Municipality; industry: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies in Shanghai Municipality - Shanghai Municipality Finance Bureau - policy count = 2"
+  ],
+  "milestone": {
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Leasing and Business Services local policies_Henan Province General Office of the People's Government policy count": 1,
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.": "Shanghai Municipality",
+    "Information Transmission, Software and IT Services local policies in Shanghai Municipality_Shanghai Municipality Finance Bureau policy count": 2,
+    "Comparison result": "2 is greater; output Lang Ji Hui Ruan Technology Co., Ltd."
+  },
+  "steps": [
+    "Extract from company_profile.csv that Yi Hai Chang Jin Shang Wu Co., Ltd.'s industry is Leasing and Business Services",
+    "Extract from policy_release_status.csv that in the local policies for Leasing and Business Services, the policy count issued by Henan Province General Office of the People's Government is 1",
+    "Extract from company_profile.csv that Lang Ji Hui Ruan Technology Co., Ltd. is located in Shanghai Municipality and its industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for Information Transmission, Software and IT Services in Shanghai Municipality, the policy count issued by the Shanghai Municipality Finance Bureau is 2",
+    "Compare 1 and 2; since 2 is greater, output \"Lang Ji Hui Ruan Technology Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "answer": "Lang Ji Hui Ruan Technology Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "b11a2b60-c999-4fc9-9581-5a6601001790"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    },
+    {
+      "policy_release_status": "0a1d883b-42a1-4d7a-938a-43e0301f00ac"
+    },
+    {
+      "company_profile": "e9d63626-4106-4f18-8022-b5396454e690"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium028_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium028_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..045f88f459760b5724ddcfb65dfbd8efd94e0186
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium028_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium028",
+  "question": "What is the difference between the number of central ministry/agency policies issued by the China Federation of Logistics & Purchasing in the ministerial policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd. and the number of local policies for the Information Transmission, Software and IT Services industry in Shanghai Municipality where Lang Ji Hui Ruan Technology Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.: Leasing and Business Services",
+    "Leasing and Business Services ministerial policies - China Federation of Logistics & Purchasing - policy count = 1",
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.: Shanghai Municipality; industry: Information Transmission, Software and IT Services",
+    "Shanghai Municipality Information Transmission, Software and IT Services local policies policy count = 15"
+  ],
+  "milestone": {
+    "Industry of Yi Hai Chang Jin Shang Wu Co., Ltd.": "Leasing and Business Services",
+    "Leasing and Business Services ministerial policies_China Federation of Logistics & Purchasing policy count": 1,
+    "Province of Lang Ji Hui Ruan Technology Co., Ltd.": "Shanghai Municipality",
+    "Shanghai Municipality Information Transmission, Software and IT Services local policies policy count": 15,
+    "Difference (China Federation of Logistics & Purchasing policy count - local policies policy count)": -14.0
+  },
+  "answer": -14.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Yi Hai Chang Jin Shang Wu Co., Ltd.'s industry is Leasing and Business Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Leasing and Business Services, the policy count issued by the China Federation of Logistics & Purchasing is 1",
+    "Extract from company_profile.csv that Lang Ji Hui Ruan Technology Co., Ltd. is located in Shanghai Municipality and its industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that the local policies policy count for Information Transmission, Software and IT Services in Shanghai Municipality is 15",
+    "Calculate the difference: 1 - 15 = -14.0"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "b11a2b60-c999-4fc9-9581-5a6601001790"
+    },
+    {
+      "company_profile": "6663bda5-77a9-4884-b0c3-c7ae19f910a8"
+    },
+    {
+      "policy_release_status": "0a1d883b-42a1-4d7a-938a-43e0301f00ac"
+    },
+    {
+      "company_profile": "e9d63626-4106-4f18-8022-b5396454e690"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium029_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium029_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..287bc45b2230d081156dd06c330f74ce460a8f0a
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium029_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium029",
+  "question": "Between the number of central ministry/agency policies in the ministerial policies for the industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd. and the number of local policies in the local policies for the industry of Hua Cheng Sheng Yuan Integrated Development Co., Ltd. in Guangdong Province, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd.: Real Estate",
+    "Real Estate ministerial policies policy count = 5",
+    "Province of Hua Cheng Sheng Yuan Integrated Development Co., Ltd.: Guangdong Province; industry: Integrated",
+    "Guangdong Province Integrated local policies policy count = 2"
+  ],
+  "milestone": {
+    "Industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd.": "Real Estate",
+    "Real Estate ministerial policies policy count": 5,
+    "Province of Hua Cheng Sheng Yuan Integrated Development Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Integrated local policies policy count": 2,
+    "Comparison result": "5 is greater; output Zhao Ye Hua Chang Real Estate Development Co., Ltd."
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhao Ye Hua Chang Real Estate Development Co., Ltd.'s industry is Real Estate",
+    "Extract from policy_release_status.csv that the ministerial policies policy count for Real Estate is 5",
+    "Extract from company_profile.csv that Hua Cheng Sheng Yuan Integrated Development Co., Ltd. is located in Guangdong Province and its industry is Integrated",
+    "Extract from policy_release_status.csv that the local policies policy count for Integrated in Guangdong Province is 2",
+    "Compare 5 and 2; since 5 is greater, output \"Zhao Ye Hua Chang Real Estate Development Co., Ltd.\""
+  ],
+  "steps_num": 5,
+  "answer": "Zhao Ye Hua Chang Real Estate Development Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "7e08e8dd-76ba-4683-9e6b-4b48d94f0d3c"
+    },
+    {
+      "company_profile": "3f2c07ea-c142-43aa-ab56-ee76d8aaa4e5"
+    },
+    {
+      "policy_release_status": "bd9a173a-9396-4aba-b4a2-059bc01cec3d"
+    },
+    {
+      "company_profile": "41fc40f8-35ba-4a45-9c72-56aa081d8caa"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium030_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium030_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..be29fa076c48cd87121e25aae379c76416ff913c
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium030_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium030",
+  "question": "What is the difference between the number of central ministry/agency policies issued by the Ministry of Housing and Urban-Rural Development in the ministerial policies for the industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd. and the number of policies for the Integrated industry in Guangdong Province where Hua Cheng Sheng Yuan Integrated Development Co., Ltd. is located?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd.: Real Estate",
+    "Real Estate ministerial policies - Ministry of Housing and Urban-Rural Development - policy count = 1",
+    "Province of Hua Cheng Sheng Yuan Integrated Development Co., Ltd.: Guangdong Province; industry: Integrated",
+    "Guangdong Province Integrated policy count = 2"
+  ],
+  "milestone": {
+    "Industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd.": "Real Estate",
+    "Real Estate ministerial policies_Ministry of Housing and Urban-Rural Development policy count": 1,
+    "Province of Hua Cheng Sheng Yuan Integrated Development Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Integrated policy count": 2,
+    "Difference (Ministry of Housing and Urban-Rural Development policy count - policy count)": -1.0
+  },
+  "answer": -1.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhao Ye Hua Chang Real Estate Development Co., Ltd.'s industry is Real Estate",
+    "Extract from policy_release_status.csv that in the ministerial policies for Real Estate, the policy count issued by the Ministry of Housing and Urban-Rural Development is 1",
+    "Extract from company_profile.csv that Hua Cheng Sheng Yuan Integrated Development Co., Ltd. is located in Guangdong Province and its industry is Integrated",
+    "Extract from policy_release_status.csv that the policy count for Integrated in Guangdong Province is 2",
+    "Calculate the difference: 1 - 2 = -1.0"
+  ],
+  "steps_num": 5,
+  "reference": [
+    {
+      "policy_release_status": "7e08e8dd-76ba-4683-9e6b-4b48d94f0d3c"
+    },
+    {
+      "company_profile": "3f2c07ea-c142-43aa-ab56-ee76d8aaa4e5"
+    },
+    {
+      "policy_release_status": "bd9a173a-9396-4aba-b4a2-059bc01cec3d"
+    },
+    {
+      "company_profile": "41fc40f8-35ba-4a45-9c72-56aa081d8caa"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium031_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium031_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4b573a1954e5ac1ec6a659e7215c71312dd6e035
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium031_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium031",
+  "question": "What is the difference between the number of central ministry/agency policies issued by the Ministry of Culture and Tourism in the ministerial policies for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the number of local policies issued by the Chengdu Municipality Bureau of Economy and Information Technology in the local policies for the Commercial Electrical Machinery and Equipment Manufacturing industry?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.: Financial Services",
+    "Financial Services ministerial policies - Ministry of Culture and Tourism - policy count = 1",
+    "Commercial Electrical Machinery and Equipment Manufacturing local policies - Chengdu Municipality Bureau of Economy and Information Technology - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_Ministry of Culture and Tourism policy count": 1,
+    "Commercial Electrical Machinery and Equipment Manufacturing local policies_Chengdu Municipality Bureau of Economy and Information Technology policy count": 1,
+    "Difference (Ministry of Culture and Tourism policy count - Chengdu Municipality Bureau of Economy and Information Technology policy count)": 0.0
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Financial Services, the policy count issued by the Ministry of Culture and Tourism is 1",
+    "Extract from policy_release_status.csv that in the local policies for Commercial Electrical Machinery and Equipment Manufacturing, the policy count issued by the Chengdu Municipality Bureau of Economy and Information Technology is 1",
+    "Calculate the difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    },
+    {
+      "policy_release_status": "3d7eef2c-40cc-4055-a862-336f60c2302e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium032_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium032_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c22cbd3b2c062390bdb0cd1f76a6a01e786095e
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium032_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium032",
+  "question": "Between the number of local policies related to Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the corresponding number of local policies for the Commercial Electrical Machinery and Equipment Manufacturing industry, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.: Financial Services",
+    "Financial Services local policies - Henan Province General Office of the People's Government - policy count = 2",
+    "Commercial Electrical Machinery and Equipment Manufacturing local policies - Hainan Province Department of Industry and Information Technology - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services local policies_Henan Province General Office of the People's Government policy count": 2,
+    "Commercial Electrical Machinery and Equipment Manufacturing local policies_Hainan Province Department of Industry and Information Technology policy count": 1,
+    "Comparison result": "2 is greater; output Hui Jin Jin Rui Cai Fu Management Co., Ltd."
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the local policies for Financial Services, the policy count issued by Henan Province General Office of the People's Government is 2",
+    "Extract from policy_release_status.csv that in the local policies for Commercial Electrical Machinery and Equipment Manufacturing, the policy count issued by the Hainan Province Department of Industry and Information Technology is 1",
+    "Compare 2 and 1; since 2 is greater, output \"Hui Jin Jin Rui Cai Fu Management Co., Ltd.\""
+  ],
+  "steps_num": 4,
+  "answer": "Hui Jin Jin Rui Cai Fu Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    },
+    {
+      "policy_release_status": "3d7eef2c-40cc-4055-a862-336f60c2302e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium033_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium033_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea73dccb2ccf86a03968f04c4bb698383d963834
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium033_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium033",
+  "question": "Between the number of local policies issued by the General Office of the Guangxi Zhuang Autonomous Regional People's Government in the local policies for the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. and the number of local policies issued by the Guizhou Province Department of Housing and Urban-Rural Development in the local policies for the Information Transmission, Software and IT Services industry, which is greater?",
+  "guidelines": "The answer must be \"Equal\", a company name, or the word \"industry\"; output only one word or company name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.: Education",
+    "Education local policies - General Office of the Guangxi Zhuang Autonomous Regional People's Government - policy count = 1",
+    "Information Transmission, Software and IT Services local policies - Guizhou Province Department of Housing and Urban-Rural Development - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.": "Education",
+    "Education local policies_General Office of the Guangxi Zhuang Autonomous Regional People's Government policy count": 1,
+    "Information Transmission, Software and IT Services local policies_Guizhou Province Department of Housing and Urban-Rural Development policy count": 1,
+    "Comparison result": "Equal"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.'s industry is Education",
+    "Extract from policy_release_status.csv that in the local policies for Education, the policy count issued by the General Office of the Guangxi Zhuang Autonomous Regional People's Government is 1",
+    "Extract from policy_release_status.csv that in the local policies for Information Transmission, Software and IT Services, the policy count issued by the Guizhou Province Department of Housing and Urban-Rural Development is 1",
+    "Compare 1 and 1; if they are equal, output \"Equal\""
+  ],
+  "steps_num": 4,
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "3f9efae4-8dfb-4ec7-bea9-87bba16a53d7"
+    },
+    {
+      "company_profile": "d5008c9a-7492-4ce4-b428-22e6861c68a7"
+    },
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium034_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium034_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc6c9debe4b69f91f55803b6faf9c4f85d68371c
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium034_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium034",
+  "question": "Between the number of policies issued by the Yunnan Province Department of Industry and Information Technology in the local policies for the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. and the number of policies issued by the Provincial General Office of the People's Government in the local policies for the Information Transmission, Software and IT Services industry, which is greater?",
+  "guidelines": "The answer must be \"Number of policies for the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. issued by the Yunnan Province Department of Industry and Information Technology\", \"Number of policies for the Information Transmission, Software and IT Services industry issued by the Provincial General Office of the People's Government\", or \"Equal\". Output only the specified answer text, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.: Education",
+    "Education local policies - Yunnan Province Department of Industry and Information Technology - policy count = 1",
+    "Information Transmission, Software and IT Services local policies - Provincial General Office of the People's Government - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.": "Education",
+    "Education local policies_Yunnan Province Department of Industry and Information Technology policy count": 1,
+    "Information Transmission, Software and IT Services local policies_Provincial General Office of the People's Government policy count": 1,
+    "Comparison result": "Equal"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.'s industry is Education",
+    "Extract from policy_release_status.csv that in the local policies for Education, the policy count issued by the Yunnan Province Department of Industry and Information Technology is 1",
+    "Extract from policy_release_status.csv that in the local policies for Information Transmission, Software and IT Services, the policy count issued by the Provincial General Office of the People's Government is 1",
+    "Compare 1 and 1; if they are equal, output \"Equal\""
+  ],
+  "steps_num": 4,
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "3f9efae4-8dfb-4ec7-bea9-87bba16a53d7"
+    },
+    {
+      "company_profile": "d5008c9a-7492-4ce4-b428-22e6861c68a7"
+    },
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium035_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium035_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a8aa1d319756acc5d465f0681ef5daf19ccbe71d
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium035_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium035",
+  "question": "What is the difference between the number of local policies issued by the Shanghai Municipality Commission of Economy and Information Technology in the local policies for the industry of Rui Xing Jian Kang Zhi Yao Co., Ltd. and the number of local policies issued by the Liaoning Province People's Government in the local policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Rui Xing Jian Kang Zhi Yao Co., Ltd.: Pharmaceutical Manufacturing",
+    "Pharmaceutical Manufacturing local policies - Shanghai Municipality Commission of Economy and Information Technology - policy count = 3",
+    "Cultural, Arts, Sports and Entertainment Goods Manufacturing local policies - Liaoning Province People's Government - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Rui Xing Jian Kang Zhi Yao Co., Ltd.": "Pharmaceutical Manufacturing",
+    "Pharmaceutical Manufacturing local policies_Shanghai Municipality Commission of Economy and Information Technology policy count": 3,
+    "Cultural, Arts, Sports and Entertainment Goods Manufacturing local policies_Liaoning Province People's Government policy count": 1,
+    "Difference (Shanghai Municipality Commission of Economy and Information Technology policy count - Liaoning Province People's Government policy count)": 2.0
+  },
+  "steps": [
+    "Extract from company_profile.csv that Rui Xing Jian Kang Zhi Yao Co., Ltd.'s industry is Pharmaceutical Manufacturing",
+    "Extract from policy_release_status.csv that in the local policies for Pharmaceutical Manufacturing, the policy count issued by the Shanghai Municipality Commission of Economy and Information Technology is 3",
+    "Extract from policy_release_status.csv that in the local policies for Cultural, Arts, Sports and Entertainment Goods Manufacturing, the policy count issued by the Liaoning Province People's Government is 1",
+    "Calculate the difference: 3 - 1 = 2.0"
+  ],
+  "steps_num": 4,
+  "answer": 2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "5cf78286-0cc2-4d3f-bdda-e401124cca92"
+    },
+    {
+      "company_profile": "4c3a2a9c-278d-458c-8242-de58a40d50a1"
+    },
+    {
+      "policy_release_status": "ae96a1ad-d893-4619-aac3-ae9056fffb7e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium036_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium036_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..627a7048c8166f7dfecfcd7566dde71dfdd3bf34
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium036_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium036",
+  "question": "Between the number of local policies issued by the Shanghai Municipality Commission of Economy and Information Technology in the local policies for the industry of Rui Xing Jian Kang Zhi Yao Co., Ltd. and the number of central ministry/agency policies issued by the Ministry of Agriculture and Rural Affairs in the ministerial policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Rui Xing Jian Kang Zhi Yao Co., Ltd.: Pharmaceutical Manufacturing",
+    "Pharmaceutical Manufacturing local policies - Shanghai Municipality Commission of Economy and Information Technology - policy count = 3",
+    "Cultural, Arts, Sports and Entertainment Goods Manufacturing ministerial policies - Ministry of Agriculture and Rural Affairs - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Rui Xing Jian Kang Zhi Yao Co., Ltd.": "Pharmaceutical Manufacturing",
+    "Pharmaceutical Manufacturing local policies_Shanghai Municipality Commission of Economy and Information Technology policy count": 3,
+    "Cultural, Arts, Sports and Entertainment Goods Manufacturing ministerial policies_Ministry of Agriculture and Rural Affairs policy count": 1,
+    "Comparison result": "3 is greater; output Rui Xing Jian Kang Zhi Yao Co., Ltd."
+  },
+  "steps": [
+    "Extract from company_profile.csv that Rui Xing Jian Kang Zhi Yao Co., Ltd.'s industry is Pharmaceutical Manufacturing",
+    "Extract from policy_release_status.csv that in the local policies for Pharmaceutical Manufacturing, the policy count issued by the Shanghai Municipality Commission of Economy and Information Technology is 3",
+    "Extract from policy_release_status.csv that in the ministerial policies for Cultural, Arts, Sports and Entertainment Goods Manufacturing, the policy count issued by the Ministry of Agriculture and Rural Affairs is 1",
+    "Compare 3 and 1; since 3 is greater, output \"Rui Xing Jian Kang Zhi Yao Co., Ltd.\""
+  ],
+  "steps_num": 4,
+  "answer": "Rui Xing Jian Kang Zhi Yao Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "5cf78286-0cc2-4d3f-bdda-e401124cca92"
+    },
+    {
+      "company_profile": "4c3a2a9c-278d-458c-8242-de58a40d50a1"
+    },
+    {
+      "policy_release_status": "ae96a1ad-d893-4619-aac3-ae9056fffb7e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium037_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium037_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4916af7ed814d12d4549c7517fcf89694b0a1251
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium037_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium037",
+  "question": "Between the number of local policies issued by the Secretariat of the Office of Nansha Development Zone Administrative Committee, Guangzhou in the local policies for the industry of Bao Xin Hui Hui Wang Luo Co., Ltd. and the number of local policies issued by the Shanghai Municipality Finance Bureau in the local policies for the General Equipment Manufacturing industry, which is greater?",
+  "guidelines": "The answer must be \"Number of local policies for the industry of Bao Xin Hui Hui Wang Luo Co., Ltd. issued by the Secretariat of the Office of Nansha Development Zone Administrative Committee, Guangzhou\" or \"Number of local policies for the General Equipment Manufacturing industry issued by the Shanghai Municipality Finance Bureau\". Output only the policy title, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Bao Xin Hui Hui Wang Luo Co., Ltd.: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies - Secretariat of the Office of Nansha Development Zone Administrative Committee, Guangzhou - policy count = 1",
+    "General Equipment Manufacturing local policies - Shanghai Municipality Finance Bureau - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Bao Xin Hui Hui Wang Luo Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies_Secretariat of the Office of Nansha Development Zone Administrative Committee, Guangzhou policy count": 1,
+    "General Equipment Manufacturing local policies_Shanghai Municipality Finance Bureau policy count": 1,
+    "Comparison result": "Equal; according to the question requirement, output General Equipment Manufacturing"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Bao Xin Hui Hui Wang Luo Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for Information Transmission, Software and IT Services, the policy count issued by the Secretariat of the Office of Nansha Development Zone Administrative Committee, Guangzhou is 1",
+    "Extract from policy_release_status.csv that in the local policies for General Equipment Manufacturing, the policy count issued by the Shanghai Municipality Finance Bureau is 1",
+    "Compare 1 and 1; since they are equal, output \"General Equipment Manufacturing\" according to the question requirement"
+  ],
+  "steps_num": 4,
+  "answer": "General Equipment Manufacturing",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    },
+    {
+      "company_profile": "fe635f7d-bee1-4e2e-9ed6-3923871590d0"
+    },
+    {
+      "policy_release_status": "3fb33313-f509-461e-95be-1ad114aef5f1"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium038_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium038_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..21dd413c69e4f9937b387586a72ffc06d58b7ce4
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium038_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium038",
+  "question": "What is the difference between the number of central ministry/agency policies issued by the General Office of the China National Intellectual Property Administration in the ministerial policies for the industry of Bao Xin Hui Hui Wang Luo Co., Ltd. and the number of local policies issued by the Sichuan Province People's Government in the local policies for the General Equipment Manufacturing industry?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Bao Xin Hui Hui Wang Luo Co., Ltd.: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies - General Office of the China National Intellectual Property Administration - policy count = 2",
+    "General Equipment Manufacturing local policies - Sichuan Province People's Government - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Bao Xin Hui Hui Wang Luo Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services ministerial policies_General Office of the China National Intellectual Property Administration policy count": 2,
+    "General Equipment Manufacturing local policies_Sichuan Province People's Government policy count": 1,
+    "Difference (General Office of the China National Intellectual Property Administration policy count - Sichuan Province People's Government policy count)": 1.0
+  },
+  "steps": [
+    "Extract from company_profile.csv that Bao Xin Hui Hui Wang Luo Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Information Transmission, Software and IT Services, the policy count issued by the General Office of the China National Intellectual Property Administration is 2",
+    "Extract from policy_release_status.csv that in the local policies for General Equipment Manufacturing, the policy count issued by the Sichuan Province People's Government is 1",
+    "Calculate the difference: 2 - 1 = 1.0"
+  ],
+  "steps_num": 4,
+  "answer": 1.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    },
+    {
+      "company_profile": "fe635f7d-bee1-4e2e-9ed6-3923871590d0"
+    },
+    {
+      "policy_release_status": "3fb33313-f509-461e-95be-1ad114aef5f1"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium039_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium039_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5be95b491a91c593fc96a41c4d40f5939f317409
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium039_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium039",
+  "question": "Between the number of local policies issued by the Hefei Municipality Office of the People's Government in the local policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. and the number of local policies issued by the Shandong Province People's Government in the local policies for the Scientific Research and Technical Services industry, which is greater?",
+  "guidelines": "The answer must be \"Number of local policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. issued by the Hefei Municipality Office of the People's Government\" or \"Number of local policies for the Scientific Research and Technical Services industry issued by the Shandong Province People's Government\". Output only the policy title, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies - Hefei Municipality Office of the People's Government - policy count = 2",
+    "Scientific Research and Technical Services local policies - Shandong Province People's Government - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies_Hefei Municipality Office of the People's Government policy count": 2,
+    "Scientific Research and Technical Services local policies_Shandong Province People's Government policy count": 1,
+    "Comparison result": "2 is greater; according to the question statement, output Scientific Research and Technical Services"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for Information Transmission, Software and IT Services, the policy count issued by the Hefei Municipality Office of the People's Government is 2",
+    "Extract from policy_release_status.csv that in the local policies for Scientific Research and Technical Services, the policy count issued by the Shandong Province People's Government is 1",
+    "Compare 2 and 1; although 2 is greater, follow the question's requirement and output \"Scientific Research and Technical Services\""
+  ],
+  "steps_num": 4,
+  "answer": "Scientific Research and Technical Services",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "policy_release_status": "2760c4e4-c3b3-4e72-8ff7-cea178ce1503"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium040_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium040_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3721c0bb4d039b5df458ee619ba244ecae06bd79
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium040_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium040",
+  "question": "What is the difference between the number of local policies issued by the Guangzhou Development District Bureau of Economy and Information Technology in the local policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. and the number of local policies issued by the Yunnan Province General Office of the People's Government in the local policies for the Scientific Research and Technical Services industry?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.: Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies - Guangzhou Development District Bureau of Economy and Information Technology - policy count = 2",
+    "Scientific Research and Technical Services local policies - Yunnan Province General Office of the People's Government - policy count = 2"
+  ],
+  "milestone": {
+    "Industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd.": "Information Transmission, Software and IT Services",
+    "Information Transmission, Software and IT Services local policies_Guangzhou Development District Bureau of Economy and Information Technology policy count": 2,
+    "Scientific Research and Technical Services local policies_Yunnan Province General Office of the People's Government policy count": 2,
+    "Difference (Guangzhou Development District Bureau of Economy and Information Technology policy count - Yunnan Province General Office of the People's Government policy count)": 0.0
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhong Ke Ke Shu Ruan Jian Co., Ltd.'s industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for Information Transmission, Software and IT Services, the policy count issued by the Guangzhou Development District Bureau of Economy and Information Technology is 2",
+    "Extract from policy_release_status.csv that in the local policies for Scientific Research and Technical Services, the policy count issued by the Yunnan Province General Office of the People's Government is 2",
+    "Calculate the difference: 2 - 2 = 0.0"
+  ],
+  "steps_num": 4,
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "5bec0ffb-41d8-4699-ad95-003164a2e21f"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "policy_release_status": "2760c4e4-c3b3-4e72-8ff7-cea178ce1503"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium041_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium041_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c716ab668fa60d58638c0f4bd04de0e534dd2f2
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium041_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium041",
+  "question": "Between the number of central ministry/agency policies issued by the Ministry of Public Security for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the number of local policies for the Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing industry in Anhui Province, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.: Financial Services",
+    "Financial Services ministerial policies - Ministry of Public Security - policy count = 1",
+    "Anhui Province Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing local policies policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_Ministry of Public Security policy count": 1,
+    "Anhui Province Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing local policies policy count": 1,
+    "Comparison result": "Equal; according to the question requirement, output Hui Jin Jin Rui Cai Fu Management Co., Ltd."
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Financial Services, the policy count issued by the Ministry of Public Security is 1",
+    "Extract from policy_release_status.csv that in the local policies for Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing in Anhui Province, the policy count is 1",
+    "Compare 1 and 1; since they are equal, output \"Hui Jin Jin Rui Cai Fu Management Co., Ltd.\" according to the question requirement"
+  ],
+  "steps_num": 4,
+  "answer": "Hui Jin Jin Rui Cai Fu Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    },
+    {
+      "policy_release_status": "dc0f2b26-e63d-4012-a032-1162ad6a8e45"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium042_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium042_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ed4ba6ff5aaea52de366dff5b71f639bcdf57e6
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium042_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium042",
+  "question": "Between the number of central ministry/agency policies issued by the General Administration of Sport of China in the ministerial policies for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the total number of policies for the Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing industry in Anhui Province, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.: Financial Services",
+    "Financial Services ministerial policies - General Administration of Sport of China - policy count = 1",
+    "Anhui Province Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_General Administration of Sport of China policy count": 1,
+    "Anhui Province Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing policy count": 1,
+    "Comparison result": "Equal; according to the question requirement, output Hui Jin Jin Rui Cai Fu Management Co., Ltd."
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Financial Services, the policy count issued by the General Administration of Sport of China is 1",
+    "Extract from policy_release_status.csv that in Anhui Province, the policy count for the Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing industry is 1",
+    "Compare 1 and 1; since they are equal, output \"Hui Jin Jin Rui Cai Fu Management Co., Ltd.\" according to the question requirement"
+  ],
+  "steps_num": 4,
+  "answer": "Hui Jin Jin Rui Cai Fu Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    },
+    {
+      "policy_release_status": "dc0f2b26-e63d-4012-a032-1162ad6a8e45"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium043_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium043_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ab2904455a5ee955e6a1145779e0f0d7b29f766
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium043_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium043",
+  "question": "Between the number of local policies issued by the Hunan Province People's Government in the local policies for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the number of policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry in Guangdong Province, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.: Financial Services",
+    "Financial Services local policies - Hunan Province People's Government - policy count = 1",
+    "Guangdong Province Cultural, Arts, Sports and Entertainment Goods Manufacturing policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services local policies_Hunan Province People's Government policy count": 1,
+    "Guangdong Province Cultural, Arts, Sports and Entertainment Goods Manufacturing policy count": 1,
+    "Comparison result": "Equal; according to the question requirement, output Hui Jin Jin Rui Cai Fu Management Co., Ltd."
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the local policies for Financial Services, the policy count issued by the Hunan Province People's Government is 1",
+    "Extract from policy_release_status.csv that in Guangdong Province, the policy count for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry is 1",
+    "Compare 1 and 1; since they are equal, output \"Hui Jin Jin Rui Cai Fu Management Co., Ltd.\" according to the question requirement"
+  ],
+  "steps_num": 4,
+  "answer": "Hui Jin Jin Rui Cai Fu Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    },
+    {
+      "policy_release_status": "33beb942-e725-47bb-9002-15bd75938e4d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium044_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium044_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0675cbe07c84e1b97cbaf443b050de04b0c2ac67
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium044_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium044",
+  "question": "Between the number of central ministry/agency policies issued by the State Administration of Foreign Exchange in the ministerial policies for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the number of local policies issued by the Shenzhen Municipality People's Government in the local policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry in Guangdong Province, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.: Financial Services",
+    "Financial Services ministerial policies - State Administration of Foreign Exchange - policy count = 1",
+    "Guangdong Province Cultural, Arts, Sports and Entertainment Goods Manufacturing local policies - Shenzhen Municipality People's Government - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_State Administration of Foreign Exchange policy count": 1,
+    "Guangdong Province Cultural, Arts, Sports and Entertainment Goods Manufacturing local policies_Shenzhen Municipality People's Government policy count": 1,
+    "Comparison result": "Equal; according to the question requirement, output Hui Jin Jin Rui Cai Fu Management Co., Ltd."
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hui Jin Jin Rui Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Financial Services, the policy count issued by the State Administration of Foreign Exchange is 1",
+    "Extract from policy_release_status.csv that in the local policies for Cultural, Arts, Sports and Entertainment Goods Manufacturing in Guangdong Province, the policy count issued by the Shenzhen Municipality People's Government is 1",
+    "Compare 1 and 1; since they are equal, output \"Hui Jin Jin Rui Cai Fu Management Co., Ltd.\" according to the question requirement"
+  ],
+  "steps_num": 4,
+  "answer": "Hui Jin Jin Rui Cai Fu Management Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "f3c336ac-af8f-4c74-aed7-cc746cbb5826"
+    },
+    {
+      "policy_release_status": "33beb942-e725-47bb-9002-15bd75938e4d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium045_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium045_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..17ca5dd251e7f91e108f3ca3c3f9371b6c72a876
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium045_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium045",
+  "question": "Is the number of central ministry/agency policies issued by the Ministry of Housing and Urban-Rural Development in the ministerial policies for the industry of Zhao Ye Ze Jin Real Estate Holdings Co., Ltd. greater than the total number of policies for the Metal Smelting and Rolling Processing industry in Gansu Province?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Zhao Ye Ze Jin Real Estate Holdings Co., Ltd.: Real Estate",
+    "Real Estate ministerial policies - Ministry of Housing and Urban-Rural Development - policy count = 1",
+    "Gansu Province Metal Smelting and Rolling Processing industry policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Zhao Ye Ze Jin Real Estate Holdings Co., Ltd.": "Real Estate",
+    "Real Estate ministerial policies_Ministry of Housing and Urban-Rural Development policy count": 1,
+    "Gansu Province Metal Smelting and Rolling Processing industry policy count": 1,
+    "Whether greater (Ministry of Housing and Urban-Rural Development policy count > industry policy count)": "No"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Zhao Ye Ze Jin Real Estate Holdings Co., Ltd.'s industry is Real Estate",
+    "Extract from policy_release_status.csv that in the ministerial policies for Real Estate, the policy count issued by the Ministry of Housing and Urban-Rural Development is 1",
+    "Extract from policy_release_status.csv that in Gansu Province, the policy count for the Metal Smelting and Rolling Processing industry is 1",
+    "Determine whether 1 is greater than 1; since it is not, output \"No\""
+  ],
+  "steps_num": 4,
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "7e08e8dd-76ba-4683-9e6b-4b48d94f0d3c"
+    },
+    {
+      "company_profile": "df986c37-1dd4-4c6e-a419-04cc2fd2a9ca"
+    },
+    {
+      "policy_release_status": "65465faa-8ce3-4c73-885b-d1ecd6298f26"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium046_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium046_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d393cbe16e9942b6e3a57232a2da361022a4f3c8
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium046_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium046",
+  "question": "Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.industry's Ministry of Housing and Urban-Rural DevelopmentNumber of policies (indicator) and Gansu Province Development and Reform CommissionNumber of policies (indicator)what is the gap?",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "evidence": [
+    "Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.industry = Real Estate",
+    "Real Estateministerial policies_Ministry of Housing and Urban-Rural DevelopmentNumber of policies (indicator)=1",
+    "Gansu Provincelocal policies_Gansu ProvinceDevelopment and Reform CommissionNumber of policies (indicator)=1"
+  ],
+  "milestone": {
+    "Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.industry": "Real Estate",
+    "Real Estateministerial policies_Ministry of Housing and Urban-Rural DevelopmentNumber of policies (indicator)": 1,
+    "Gansu Provincelocal policies_Gansu ProvinceDevelopment and Reform CommissionNumber of policies (indicator)": 1,
+    "Difference(Ministry of Housing and Urban-Rural DevelopmentNumber of policies (indicator)-Gansu ProvinceDevelopment and Reform CommissionNumber of policies (indicator))": 0.0
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.industry = Real Estate",
+    "Extracted from policy_release_status.csv: Real Estateministerial policies in Ministry of Housing and Urban-Rural DevelopmentNumber of policies (indicator)=1",
+    "policy_release_status.csv in by province=Gansu Province、industry=extractlocal policies in Gansu ProvinceDevelopment and Reform CommissionNumber of policies (indicator)=1",
+    "Calculate difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "7e08e8dd-76ba-4683-9e6b-4b48d94f0d3c"
+    },
+    {
+      "company_profile": "df986c37-1dd4-4c6e-a419-04cc2fd2a9ca"
+    },
+    {
+      "policy_release_status": "65465faa-8ce3-4c73-885b-d1ecd6298f26"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium047_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium047_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c13cab10fcdc62a0f4b0c60d51ec30416b2d4e35
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium047_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium047",
+  "question": "Between the number of local policies issued by the Chengdu Municipality Bureau of Economy and Information Technology in the local policies for the industry of Hua Xin Yuan Shi New Materials Co., Ltd. and the number of local policies issued by the Gansu Province General Office of the People's Government in the local policies for the Pharmaceutical Manufacturing industry in Gansu Province, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"Equal\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hua Xin Yuan Shi New Materials Co., Ltd.: Non-metallic Mineral Products",
+    "Non-metallic Mineral Products local policies - Chengdu Municipality Bureau of Economy and Information Technology - policy count = 1",
+    "Pharmaceutical Manufacturing local policies in Gansu Province - Gansu Province General Office of the People's Government - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Hua Xin Yuan Shi New Materials Co., Ltd.": "Non-metallic Mineral Products",
+    "Non-metallic Mineral Products local policies_Chengdu Municipality Bureau of Economy and Information Technology policy count": 1,
+    "Pharmaceutical Manufacturing local policies_Gansu Province General Office of the People's Government policy count": 1,
+    "Comparison result": "Equal"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Xin Yuan Shi New Materials Co., Ltd.'s industry is Non-metallic Mineral Products",
+    "Extract from policy_release_status.csv that in the local policies for Non-metallic Mineral Products, the policy count issued by the Chengdu Municipality Bureau of Economy and Information Technology is 1",
+    "Extract from policy_release_status.csv that in the local policies for Pharmaceutical Manufacturing in Gansu Province, the policy count issued by the Gansu Province General Office of the People's Government is 1",
+    "Compare 1 and 1; if they are equal, output \"Equal\""
+  ],
+  "steps_num": 4,
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "cd90a091-df5d-424c-a346-8964d348fe65"
+    },
+    {
+      "company_profile": "f150e113-74af-4929-b33d-7b30a892e86d"
+    },
+    {
+      "policy_release_status": "0c41d7a1-4f57-436d-9ea4-0b9e7e18eb45"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium048_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium048_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c5ffaad019fe64bd06081162b3041c29d4bd6aa
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium048_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium048",
+  "question": "Between the number of policies issued by the Shandong Province Department of Industry and Information Technology in the local policies for the industry of Hua Xin Yuan Shi Xin Cai Liao Co., Ltd. and the number of policies for the Pharmaceutical Manufacturing industry in Gansu Province, which is higher?",
+  "guidelines": "The answer must be a company name or the word \"Equal\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hua Xin Yuan Shi Xin Cai Liao Co., Ltd.: Non-metallic Mineral Products",
+    "Non-metallic Mineral Products local policies - Shandong Province Department of Industry and Information Technology - policy count = 1",
+    "Gansu Province Pharmaceutical Manufacturing policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Hua Xin Yuan Shi Xin Cai Liao Co., Ltd.": "Non-metallic Mineral Products",
+    "Non-metallic Mineral Products local policies_Shandong Province Department of Industry and Information Technology policy count": 1,
+    "Gansu Province Pharmaceutical Manufacturing policy count": 1,
+    "Comparison result": "Equal"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Xin Yuan Shi Xin Cai Liao Co., Ltd.'s industry is Non-metallic Mineral Products",
+    "Extract from policy_release_status.csv that in the local policies for Non-metallic Mineral Products, the policy count issued by the Shandong Province Department of Industry and Information Technology is 1",
+    "Extract from policy_release_status.csv that in Gansu Province, the policy count for the Pharmaceutical Manufacturing industry is 1",
+    "Compare 1 and 1; if they are equal, output \"Equal\""
+  ],
+  "steps_num": 4,
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "cd90a091-df5d-424c-a346-8964d348fe65"
+    },
+    {
+      "company_profile": "f150e113-74af-4929-b33d-7b30a892e86d"
+    },
+    {
+      "policy_release_status": "0c41d7a1-4f57-436d-9ea4-0b9e7e18eb45"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium049_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium049_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b429e3d5ece90788e80ceec6fab902b81bcd45a
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium049_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium049",
+  "question": "Between the number of local policies for the industry of Wan Hui Jin Sheng Real Estate Development Co., Ltd. and the number of local policies issued by the Jiangxi Province People's Government for the Commercial Electrical Machinery and Equipment Manufacturing industry in Jiangxi Province, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Wan Hui Jin Sheng Real Estate Development Co., Ltd.: Real Estate",
+    "Real Estate local policies policy count = 2",
+    "Jiangxi Province Commercial Electrical Machinery and Equipment Manufacturing local policies - Jiangxi Province People's Government - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Wan Hui Jin Sheng Real Estate Development Co., Ltd.": "Real Estate",
+    "Real Estate local policies policy count": 2,
+    "Jiangxi Province Commercial Electrical Machinery and Equipment Manufacturing local policies_Jiangxi Province People's Government policy count": 1,
+    "Comparison result": "2 is greater; output Wan Hui Jin Sheng Real Estate Development Co., Ltd."
+  },
+  "steps": [
+    "Extract from company_profile.csv that Wan Hui Jin Sheng Real Estate Development Co., Ltd.'s industry is Real Estate",
+    "Extract from policy_release_status.csv that the local policies policy count for Real Estate is 2",
+    "Extract from policy_release_status.csv that in the local policies for Commercial Electrical Machinery and Equipment Manufacturing in Jiangxi Province, the policy count issued by the Jiangxi Province People's Government is 1",
+    "Compare 2 and 1; since 2 is greater, output \"Wan Hui Jin Sheng Real Estate Development Co., Ltd.\""
+  ],
+  "steps_num": 4,
+  "answer": "Wan Hui Jin Sheng Real Estate Development Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "7e08e8dd-76ba-4683-9e6b-4b48d94f0d3c"
+    },
+    {
+      "company_profile": "3ef2095b-702e-465f-a901-fdc23ba53048"
+    },
+    {
+      "policy_release_status": "6786ccf7-d12c-4afb-abde-fd28560aa5e3"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium050_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium050_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a413637f1649066582012eb6cc670a01655a8ff4
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium050_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium050",
+  "question": "Wan Hui Jin Sheng Fang Di Chan Kai Fa Co., Ltd.industry's Shandong ProvinceNumber of policies (indicator) and Jiangxi Province Number of policies (indicator)how much？",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "evidence": [
+    "Wan Hui Jin Sheng Fang Di Chan Kai Fa Co., Ltd.industry = Real Estate",
+    "Real Estatelocal policies_Shandong ProvincepersonsNumber of policies (indicator)=1",
+    "Jiangxi Provincelocal policies_Jiangxi ProvincepersonsNumber of policies (indicator)=1"
+  ],
+  "milestone": {
+    "Wan Hui Jin Sheng Fang Di Chan Kai Fa Co., Ltd.industry": "Real Estate",
+    "Real Estatelocal policies_Shandong ProvincepersonsNumber of policies (indicator)": 1,
+    "Jiangxi Provincelocal policies_Jiangxi ProvincepersonsNumber of policies (indicator)": 1,
+    "Difference(Shandong ProvincepersonsNumber of policies (indicator)-Jiangxi ProvincepersonsNumber of policies (indicator))": 0.0
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Wan Hui Jin Sheng Fang Di Chan Kai Fa Co., Ltd.industry = Real Estate",
+    "Extracted from policy_release_status.csv: Real Estatelocal policies in Shandong ProvincepersonsNumber of policies (indicator)=1",
+    "policy_release_status.csv in by province=Jiangxi Province、industry=extractlocal policies in Jiangxi ProvincepersonsNumber of policies (indicator)=1",
+    "Calculate difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "7e08e8dd-76ba-4683-9e6b-4b48d94f0d3c"
+    },
+    {
+      "company_profile": "3ef2095b-702e-465f-a901-fdc23ba53048"
+    },
+    {
+      "policy_release_status": "6786ccf7-d12c-4afb-abde-fd28560aa5e3"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium051_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium051_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..428cefbee1199117a6be65ce9cd30d659f281d07
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium051_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium051",
+  "question": "Wu Li Hui Da Chain Co., Ltd.industry's ministerial policies_Number of policies (indicator) and China Shenzhen CitypersonsNumber of policies (indicator)compared with difference how much？",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "evidence": [
+    "Wu Li Hui Da Chain Co., Ltd.industry = Retail",
+    "Retailministerial policies_Number of policies (indicator)=1",
+    "ChinaConglomerateslocal policies_Shenzhen CitypersonsNumber of policies (indicator)=1"
+  ],
+  "milestone": {
+    "Wu Li Hui Da Chain Co., Ltd.industry": "Retail",
+    "Retailministerial policies_Number of policies (indicator)": 1,
+    "ChinaConglomerateslocal policies_Shenzhen CitypersonsNumber of policies (indicator)": 1,
+    "Difference(Number of policies (indicator)-Shenzhen CitypersonsNumber of policies (indicator))": 0.0
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Wu Li Hui Da Chain Co., Ltd.industry = Retail",
+    "Extracted from policy_release_status.csv: Retailministerial policies in Number of policies (indicator)=1",
+    "Extracted from policy_release_status.csv: ChinaConglomerateslocal policies in Shenzhen CitypersonsNumber of policies (indicator)=1",
+    "Calculate difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "1f977ef9-1e22-45fc-8781-0f97b90702d3"
+    },
+    {
+      "company_profile": "24d73db3-b3be-4f9d-87e8-a8f7bcd7b0f9"
+    },
+    {
+      "policy_release_status": "69a94ba4-0c65-466a-9a53-0c1dd38a2375"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium052_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium052_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e35f10143ef25aa6f6713eed2d7ffccca1efadb8
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium052_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium052",
+  "question": "Between the number of local policies issued by the Sichuan Province People's Government in the local policies for the industry of Wu Li Hui Da Chain Co., Ltd. and the number of central ministry/agency policies in the China conglomerates category, which is greater?",
+  "guidelines": "The answer must be a company name or the word \"industry\"; output only the name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Wu Li Hui Da Chain Co., Ltd.: Wholesale and Retail",
+    "Wholesale and Retail local policies - Sichuan Province People's Government - policy count = 1",
+    "China conglomerates central ministry/agency policies policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Wu Li Hui Da Chain Co., Ltd.": "Wholesale and Retail",
+    "Wholesale and Retail local policies_Sichuan Province People's Government policy count": 1,
+    "China conglomerates central ministry/agency policies policy count": 1,
+    "Comparison result": "Equal; according to the question requirement, output Wu Li Hui Da Chain Co., Ltd."
+  },
+  "steps": [
+    "Extract from company_profile.csv that Wu Li Hui Da Chain Co., Ltd.'s industry is Wholesale and Retail",
+    "Extract from policy_release_status.csv that in the local policies for Wholesale and Retail, the policy count issued by the Sichuan Province People's Government is 1",
+    "Extract from policy_release_status.csv that in the China conglomerates central ministry/agency policies, the policy count is 1",
+    "Compare 1 and 1; since they are equal, output \"Wu Li Hui Da Chain Co., Ltd.\" according to the question requirement"
+  ],
+  "steps_num": 4,
+  "answer": "Wu Li Hui Da Chain Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "1f977ef9-1e22-45fc-8781-0f97b90702d3"
+    },
+    {
+      "company_profile": "24d73db3-b3be-4f9d-87e8-a8f7bcd7b0f9"
+    },
+    {
+      "policy_release_status": "69a94ba4-0c65-466a-9a53-0c1dd38a2375"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium053_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium053_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..86d10907e23e02571494e2056509b6beeb9fc657
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium053_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium053",
+  "question": "Between the number of policies issued by the State Administration of Foreign Exchange in the ministerial policies for the industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd. and the number of local policies issued by the Chongqing Municipality General Office of the People's Government in the local policies for the Financial Services industry in China, which is greater?",
+  "guidelines": "The answer must be \"Equal\", a company name, or the word \"industry\"; output only one word or company name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.: Financial Services",
+    "Financial Services ministerial policies - State Administration of Foreign Exchange - policy count = 1",
+    "China Financial Services local policies - Chongqing Municipality General Office of the People's Government - policy count = 2"
+  ],
+  "milestone": {
+    "Industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_State Administration of Foreign Exchange policy count": 1,
+    "China Financial Services local policies_Chongqing Municipality General Office of the People's Government policy count": 2,
+    "Comparison result": "2 is greater; according to the question requirement, output Chongqing Municipality General Office of the People's Government"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Ying Tai Sheng Cai Fu Management Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Financial Services, the policy count issued by the State Administration of Foreign Exchange is 1",
+    "Extract from policy_release_status.csv that in the China Financial Services local policies, the policy count issued by the Chongqing Municipality General Office of the People's Government is 2",
+    "Compare 1 and 2; since 2 is greater, output \"Chongqing Municipality General Office of the People's Government\""
+  ],
+  "steps_num": 4,
+  "answer": "Chongqing Municipality General Office of the People's Government",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "b3ffe2d5-c5b8-4499-b7e7-c6955e84e897"
+    },
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium054_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium054_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..caa5a71890fcb56929e10e376f19949b6ce84d7e
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium054_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium054",
+  "question": "Hua Ying Tai Sheng Wealth Management Co., Ltd.industry's Shanghai Municipalitypersons Number of policies (indicator) and China Number of policies issued by Ministry of Educationcompared with which unitsgreater？",
+  "guidelines": "The answer must a company name or \"industry\"\"Equal\", Output onlyname, without any explanation or description.Wu, answer\"No relevant data found\".",
+  "evidence": [
+    "Hua Ying Tai Sheng Wealth Management Co., Ltd.industry = Financial Industry",
+    "Financial Industrylocal policies_Shanghai MunicipalitypersonsNumber of policies (indicator)=1",
+    "ChinaFinancial Industryministerial policies_EducationNumber of policies (indicator)=1"
+  ],
+  "milestone": {
+    "Hua Ying Tai Sheng Wealth Management Co., Ltd.industry": "Financial Industry",
+    "Financial Industrylocal policies_Shanghai MunicipalitypersonsNumber of policies (indicator)": 1,
+    "ChinaFinancial Industryministerial policies_EducationNumber of policies (indicator)": 1,
+    "Comparison result": "Equal"
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Hua Ying Tai Sheng Wealth Management Co., Ltd.industry = Financial Industry",
+    "Extracted from policy_release_status.csv: Financial Industrylocal policies in Shanghai MunicipalitypersonsNumber of policies (indicator)=1",
+    "Extracted from policy_release_status.csv: ChinaFinancial Industryministerial policies in EducationNumber of policies (indicator)=1",
+    "Compare1 and 1, Equalthen output\"Equal\""
+  ],
+  "steps_num": 4,
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "b3ffe2d5-c5b8-4499-b7e7-c6955e84e897"
+    },
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium055_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium055_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c451aa4256213a20a0f16a24b6b88179da63b946
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium055_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium055",
+  "question": "Between the number of policies issued by the Ministry of Public Security in the ministerial policies for the industry of Tong Tong Ze Hong Securities Co., Ltd. and the number of policies issued by the State Administration of Foreign Exchange in the ministerial policies for the Leasing and Business Services industry in China, which is greater?",
+  "guidelines": "The answer must be \"Equal\", a company name, or the word \"industry\"; output only one word or company name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Industry of Tong Tong Ze Hong Securities Co., Ltd.: Financial Services",
+    "Financial Services ministerial policies - Ministry of Public Security - policy count = 1",
+    "China Leasing and Business Services ministerial policies - State Administration of Foreign Exchange - policy count = 1"
+  ],
+  "milestone": {
+    "Industry of Tong Tong Ze Hong Securities Co., Ltd.": "Financial Services",
+    "Financial Services ministerial policies_Ministry of Public Security policy count": 1,
+    "China Leasing and Business Services ministerial policies_State Administration of Foreign Exchange policy count": 1,
+    "Comparison result": "Equal"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Tong Tong Ze Hong Securities Co., Ltd.'s industry is Financial Services",
+    "Extract from policy_release_status.csv that in the ministerial policies for Financial Services, the policy count issued by the Ministry of Public Security is 1",
+    "Extract from policy_release_status.csv that in the ministerial policies for Leasing and Business Services in China, the policy count issued by the State Administration of Foreign Exchange is 1",
+    "Compare 1 and 1; if they are equal, output \"Equal\""
+  ],
+  "steps_num": 4,
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "63debe9f-0257-4192-a552-fe35fb82a435"
+    },
+    {
+      "policy_release_status": "b11a2b60-c999-4fc9-9581-5a6601001790"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium056_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium056_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..91d3784ce825cfe6b8768c6bca0b1e8c687c48ec
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium056_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium056",
+  "question": "Tong Tong Ze Hong Zheng Quan Co., Ltd.industry's local policiesGuangzhou CitypersonsNumber of policies (indicator) and China ministerial policiesNational Development and Reform CommissionNumber of policies (indicator)compared with difference how much？",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "evidence": [
+    "Tong Tong Ze Hong Zheng Quan Co., Ltd.industry = Financial Industry",
+    "Financial Industrylocal policies_Guangzhou CitypersonsNumber of policies (indicator)=1",
+    "ChinaBusiness Servicesministerial policies_National Development and Reform CommissionNumber of policies (indicator)=1"
+  ],
+  "milestone": {
+    "Tong Tong Ze Hong Zheng Quan Co., Ltd.industry": "Financial Industry",
+    "Financial Industrylocal policies_Guangzhou CitypersonsNumber of policies (indicator)": 1,
+    "ChinaBusiness Servicesministerial policies_National Development and Reform CommissionNumber of policies (indicator)": 1,
+    "Difference(Guangzhou CitypersonsNumber of policies (indicator)-National Development and Reform CommissionNumber of policies (indicator))": 0.0
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Tong Tong Ze Hong Zheng Quan Co., Ltd.industry = Financial Industry",
+    "Extracted from policy_release_status.csv: Financial Industrylocal policies in Guangzhou CitypersonsNumber of policies (indicator)=1",
+    "Extracted from policy_release_status.csv: ChinaBusiness Servicesministerial policies in National Development and Reform CommissionNumber of policies (indicator)=1",
+    "Calculate difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "63debe9f-0257-4192-a552-fe35fb82a435"
+    },
+    {
+      "policy_release_status": "b11a2b60-c999-4fc9-9581-5a6601001790"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium057_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium057_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..9fafc1866fb99f3ce03716daba9876c66600ed8e
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium057_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium057",
+  "question": "Hua Ying Tai Sheng Wealth Management Co., Ltd.industry's Number of policies (indicator) and China Sichuan ProvinceNumber of policies (indicator)what is the gap?",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "evidence": [
+    "Hua Ying Tai Sheng Wealth Management Co., Ltd.industry = Financial Industry",
+    "Financial Industryministerial policies_Number of policies (indicator)=3",
+    "ChinaTransportation、Postal Serviceslocal policies_Sichuan ProvinceNumber of policies (indicator)=1"
+  ],
+  "milestone": {
+    "Hua Ying Tai Sheng Wealth Management Co., Ltd.industry": "Financial Industry",
+    "Financial Industryministerial policies_Number of policies (indicator)": 3,
+    "ChinaTransportation、Postal Serviceslocal policies_Sichuan ProvinceNumber of policies (indicator)": 1,
+    "Difference(Number of policies (indicator)-Sichuan ProvinceNumber of policies (indicator))": 2.0
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Hua Ying Tai Sheng Wealth Management Co., Ltd.industry = Financial Industry",
+    "Extracted from policy_release_status.csv: Financial Industryministerial policies in Number of policies (indicator)=3",
+    "Extracted from policy_release_status.csv: ChinaTransportation、Postal Serviceslocal policies in Sichuan ProvinceNumber of policies (indicator)=1",
+    "Calculate difference: 3 - 1 = 2.0"
+  ],
+  "steps_num": 4,
+  "answer": 2.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "a2734b19-8347-4c63-b17b-61453aab869e"
+    },
+    {
+      "company_profile": "b3ffe2d5-c5b8-4499-b7e7-c6955e84e897"
+    },
+    {
+      "policy_release_status": "687a938e-3531-48c2-be70-77d2cce459e8"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium058_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium058_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..08cca86a3d5e288b047421cc8b1eed93a2949cbc
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium058_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium058",
+  "question": "Shi Yang Jin Jin Electrical Appliances Co., Ltd.province Guangdong ProvinceNumber of policies (indicator) and China Hainan ProvinceNumber of policies (indicator)compared with difference how much？",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "evidence": [
+    "Shi Yang Jin Jin Electrical Appliances Co., Ltd.province = Guangdong Province",
+    "Guangdong ProvinceTransportation、Postal Serviceslocal policies_Guangdong ProvinceNumber of policies (indicator)=1",
+    "ChinaFood and Beveragelocal policies_Hainan ProvinceNumber of policies (indicator)=1"
+  ],
+  "milestone": {
+    "Shi Yang Jin Jin Electrical Appliances Co., Ltd.province": "Guangdong Province",
+    "Guangdong ProvinceTransportation、Postal Serviceslocal policies_Guangdong ProvinceNumber of policies (indicator)": 1,
+    "ChinaFood and Beveragelocal policies_Hainan ProvinceNumber of policies (indicator)": 1,
+    "Difference(Guangdong ProvinceNumber of policies (indicator)-Hainan ProvinceNumber of policies (indicator))": 0.0
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Shi Yang Jin Jin Electrical Appliances Co., Ltd.province = Guangdong Province and industry = Transportation、Postal Services",
+    "policy_release_status.csv in by province=Guangdong Province、industry=Transportation、Postal Servicesextractlocal policies in Guangdong ProvinceNumber of policies (indicator)=1",
+    "Extracted from policy_release_status.csv: ChinaFood and Beveragelocal policies in Hainan ProvinceNumber of policies (indicator)=1",
+    "Calculate difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "2ad42eaa-d00a-43d7-b5ce-0aac5ae20ab2"
+    },
+    {
+      "company_profile": "c4791202-0687-425c-9012-35538f3a300c"
+    },
+    {
+      "policy_release_status": "1b6442a3-4d16-461b-94ca-4e65babd7a1d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium059_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium059_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f30ae643f8847432aacb710afe5b38f77cad77a5
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium059_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium059",
+  "question": "Between the number of policies issued by the Guangdong Province Development and Reform Commission in the local policies for the province where Shi Yang Jin Jin Electrical Appliances Co., Ltd. is located and the number of local policies issued by the Department of Digitalization and Future Industries in the Food and Beverage industry in China, which is greater?",
+  "guidelines": "The answer must be a company name, the word \"industry\", or \"Equal\"; output only one word or one company name, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Shi Yang Jin Jin Electrical Appliances Co., Ltd.: Guangdong Province",
+    "Guangdong Province Transportation, Warehousing and Postal Services local policies - Guangdong Province Development and Reform Commission - policy count = 1",
+    "China Food and Beverage local policies - Department of Digitalization and Future Industries - policy count = 1"
+  ],
+  "milestone": {
+    "Province of Shi Yang Jin Jin Electrical Appliances Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Transportation, Warehousing and Postal Services local policies_Guangdong Province Development and Reform Commission policy count": 1,
+    "China Food and Beverage local policies_Department of Digitalization and Future Industries policy count": 1,
+    "Comparison result": "Equal"
+  },
+  "steps": [
+    "Extract from company_profile.csv that Shi Yang Jin Jin Electrical Appliances Co., Ltd. is located in Guangdong Province and its industry is Transportation, Warehousing and Postal Services",
+    "Extract from policy_release_status.csv that in the local policies for Transportation, Warehousing and Postal Services in Guangdong Province, the policy count issued by the Guangdong Province Development and Reform Commission is 1",
+    "Extract from policy_release_status.csv that in the China Food and Beverage local policies, the policy count issued by the Department of Digitalization and Future Industries is 1",
+    "Compare 1 and 1; if they are equal, output \"Equal\""
+  ],
+  "steps_num": 4,
+  "answer": "Equal",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "2ad42eaa-d00a-43d7-b5ce-0aac5ae20ab2"
+    },
+    {
+      "company_profile": "c4791202-0687-425c-9012-35538f3a300c"
+    },
+    {
+      "policy_release_status": "1b6442a3-4d16-461b-94ca-4e65babd7a1d"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium060_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium060_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce43b621497867359d4a5a110d50898e0271cd15
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium060_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium060",
+  "question": "Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.province Number of policies (indicator) and China  in Guangdong ProvinceNumber of policies (indicator)compared with which unitsgreater？",
+  "guidelines": "The answer must a company name or \"industry\", Output onlyname, without any explanation or description.Wu, answer\"No relevant data found\".",
+  "evidence": [
+    "Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.province = Shanghai Municipality",
+    "Shanghai MunicipalityNumber of policies (indicator)=11",
+    "ChinaSemiconductor Industrylocal policies_ in Guangdong ProvinceNumber of policies (indicator)=1"
+  ],
+  "milestone": {
+    "Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.province": "Shanghai Municipality",
+    "Shanghai MunicipalityNumber of policies (indicator)": 11,
+    "ChinaSemiconductor Industrylocal policies_ in Guangdong ProvinceNumber of policies (indicator)": 1,
+    "Comparison result": "11greater，outputJian Fan Ning Ze Yang Lao Fu Wu Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.province = Shanghai Municipality and industry = ",
+    "policy_release_status.csv in by province=Shanghai Municipality、industry=extractNumber of policies (indicator)=11",
+    "Extracted from policy_release_status.csv: ChinaSemiconductor Industrylocal policies in  in Guangdong ProvinceNumber of policies (indicator)=1",
+    "Compare11 and 1, 11greater, output\"Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.\""
+  ],
+  "steps_num": 4,
+  "answer": "Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "9c2f7d86-2c84-42b8-99d1-0e8ccc483fc6"
+    },
+    {
+      "company_profile": "ca56f5f4-4ea0-433d-9eb2-cf2c630fc69d"
+    },
+    {
+      "policy_release_status": "9c744b91-d271-4bff-83e1-2ff0c2add57b"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium061_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium061_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f9134bfbb8fd3018c239576a1be2ae79d70d22f9
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium061_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium061",
+  "question": "Hua Cheng Sheng Yuan Integrated Development Co., Ltd.province Guangzhou CityHuang Pu DistrictNumber of policies (indicator) and China Number of policies (indicator)compared with which unitsgreater？",
+  "guidelines": "The answer must a company name or \"industry\", Output onlyname, without any explanation or description.Wu, answer\"No relevant data found\".",
+  "evidence": [
+    "Hua Cheng Sheng Yuan Integrated Development Co., Ltd.province = Guangdong Province",
+    "Guangdong ProvinceInformation Transmission, Software and IT Serviceslocal policies_Guangzhou CityHuang Pu DistrictNumber of policies (indicator)=2",
+    "ChinaSemiconductor Industrylocal policies_Number of policies (indicator)=1"
+  ],
+  "milestone": {
+    "Hua Cheng Sheng Yuan Integrated Development Co., Ltd.province": "Guangdong Province",
+    "Guangdong ProvinceInformation Transmission, Software and IT Serviceslocal policies_Guangzhou CityHuang Pu DistrictNumber of policies (indicator)": 2,
+    "ChinaSemiconductor Industrylocal policies_Number of policies (indicator)": 1,
+    "Comparison result": "2greater，outputHua Cheng Sheng Yuan Integrated Development Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Hua Cheng Sheng Yuan Integrated Development Co., Ltd.province = Guangdong Province and industry = Information Transmission, Software and IT Services",
+    "policy_release_status.csv in by province=Guangdong Province、industry=Information Transmission, Software and IT Servicesextractlocal policies in Guangzhou CityHuang Pu DistrictNumber of policies (indicator)=2",
+    "Extracted from policy_release_status.csv: ChinaSemiconductor Industrylocal policies in Number of policies (indicator)=1",
+    "Compare2 and 1, 2greater, output\"Hua Cheng Sheng Yuan Integrated Development Co., Ltd.\""
+  ],
+  "steps_num": 4,
+  "answer": "Hua Cheng Sheng Yuan Integrated Development Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "69b9bc85-fed0-41f4-a1bd-e3542ed816af"
+    },
+    {
+      "company_profile": "41fc40f8-35ba-4a45-9c72-56aa081d8caa"
+    },
+    {
+      "policy_release_status": "9c744b91-d271-4bff-83e1-2ff0c2add57b"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium062_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium062_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f46460fa15483f3208affeb1c043d13309c9f5d9
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium062_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium062",
+  "question": "What is the difference between the number of local policies issued by the Guangdong Provincial Committee of the CPC in the local policies for the industry of Hua Cheng Sheng Yuan Integrated Development Co., Ltd. and the number of local policies issued by the Fujian Province Department of Industry and Information Technology in the local policies for the Semiconductor Industry in China?",
+  "guidelines": "The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Hua Cheng Sheng Yuan Integrated Development Co., Ltd.: Guangdong Province",
+    "Guangdong Province Information Transmission, Software and IT Services local policies - Guangdong Provincial Committee of the CPC - policy count = 1",
+    "China Semiconductor Industry local policies - Fujian Province Department of Industry and Information Technology - policy count = 1"
+  ],
+  "milestone": {
+    "Province of Hua Cheng Sheng Yuan Integrated Development Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Information Transmission, Software and IT Services local policies_Guangdong Provincial Committee of the CPC policy count": 1,
+    "China Semiconductor Industry local policies_Fujian Province Department of Industry and Information Technology policy count": 1,
+    "Difference (Guangdong Provincial Committee of the CPC policy count - Fujian Province Department of Industry and Information Technology policy count)": 0.0
+  },
+  "steps": [
+    "Extract from company_profile.csv that Hua Cheng Sheng Yuan Integrated Development Co., Ltd. is located in Guangdong Province and its industry is Information Transmission, Software and IT Services",
+    "Extract from policy_release_status.csv that in the local policies for Information Transmission, Software and IT Services in Guangdong Province, the policy count issued by the Guangdong Provincial Committee of the CPC is 1",
+    "Extract from policy_release_status.csv that in the China Semiconductor Industry local policies, the policy count issued by the Fujian Province Department of Industry and Information Technology is 1",
+    "Calculate the difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "69b9bc85-fed0-41f4-a1bd-e3542ed816af"
+    },
+    {
+      "company_profile": "41fc40f8-35ba-4a45-9c72-56aa081d8caa"
+    },
+    {
+      "policy_release_status": "9c744b91-d271-4bff-83e1-2ff0c2add57b"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium063_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium063_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c13e20be56800613083943640357dbe89723077c
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium063_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium063",
+  "question": "Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.province Number of policies (indicator) and China TransportationNumber of policies (indicator)compared with  how much？",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "evidence": [
+    "Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.province = Guangdong Province",
+    "Guangdong ProvinceElectronicslocal policies_Guangdong ProvincepersonsNumber of policies (indicator)=1",
+    "ChinaPlastic Productsministerial policies_TransportationNumber of policies (indicator)=1"
+  ],
+  "milestone": {
+    "Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.province": "Guangdong Province",
+    "Guangdong ProvinceElectronicslocal policies_Guangdong ProvincepersonsNumber of policies (indicator)": 1,
+    "ChinaPlastic Productsministerial policies_TransportationNumber of policies (indicator)": 1,
+    "Difference(Guangdong ProvincepersonsNumber of policies (indicator)-TransportationNumber of policies (indicator))": 0.0
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.province = Guangdong Province and industry = Electronics",
+    "policy_release_status.csv in by province=Guangdong Province、industry=Electronicsextractlocal policies in Guangdong ProvincepersonsNumber of policies (indicator)=1",
+    "Extracted from policy_release_status.csv: ChinaPlastic Productsministerial policies in TransportationNumber of policies (indicator)=1",
+    "Calculate difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "558608bf-1ac0-4a8b-8285-011f4eb6f845"
+    },
+    {
+      "company_profile": "1497d311-a090-4fdb-9af8-30fee577d417"
+    },
+    {
+      "policy_release_status": "5a66b14f-0c67-4f3c-a56f-1db3efa4e764"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium064_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium064_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..97f05dadfa71211a63c0f42e28ddf399bc6ef03a
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium064_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium064",
+  "question": "Is the number of specific local policies in the province where Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd. is located the same as the number of policies issued by the Hainan Province Department of Industry and Information Technology for China's rubber and plastic products industry local policies?",
+  "guidelines": "The answer must be \"Yes\" or \"No\"; output only one word, without any explanation or description. If relevant data cannot be found, answer \"No relevant data found\".",
+  "evidence": [
+    "Province of Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.: Guangdong Province",
+    "Guangdong Province Consumer Electronics and Electrical Equipment local policies — Guangdong Province General Office of the People's Government — number of policies: 1",
+    "China rubber and plastic products industry local policies — Hainan Province Department of Industry and Information Technology — number of policies: 1"
+  ],
+  "milestone": {
+    "Province of Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.": "Guangdong Province",
+    "Guangdong Province Consumer Electronics and Electrical Equipment local policies_Guangdong Province General Office of the People's Government number of policies": 1,
+    "China rubber and plastic products industry local policies_Hainan Province Department of Industry and Information Technology number of policies": 1,
+    "Whether the same": "Yes"
+  },
+  "steps": [
+    "From company_profile.csv, extract that Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd. is in Guangdong Province and its industry is Consumer Electronics and Electrical Equipment",
+    "From policy_release_status.csv, filter by province = Guangdong Province and industry = Consumer Electronics and Electrical Equipment, and extract the number of local policies issued by the Guangdong Province General Office of the People's Government: 1",
+    "From policy_release_status.csv, extract the number of China rubber and plastic products industry local policies issued by the Hainan Province Department of Industry and Information Technology: 1",
+    "Compare whether 1 equals 1; if so, output \"Yes\""
+  ],
+  "steps_num": 4,
+  "answer": "Yes",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "558608bf-1ac0-4a8b-8285-011f4eb6f845"
+    },
+    {
+      "company_profile": "1497d311-a090-4fdb-9af8-30fee577d417"
+    },
+    {
+      "policy_release_status": "5a66b14f-0c67-4f3c-a56f-1db3efa4e764"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium065_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium065_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3725204ddd1a7f01d5f753432165315bcc244822
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium065_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium065",
+  "question": "Zhong Ke Ke Shu Software Co., Ltd.province  Number of policies (indicator) and China Sichuan Provincepersons Number of policies (indicator)what is the gap?",
+  "guidelines": "The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer\"No relevant data found\".",
+  "evidence": [
+    "Zhong Ke Ke Shu Software Co., Ltd.province = Guangdong Province",
+    "Guangdong ProvinceCommunication Transmission Equipmentlocal policies_Guangdong ProvinceNumber of policies (indicator)=1",
+    "ChinaMetal Productslocal policies_Sichuan ProvincepersonsNumber of policies (indicator)=1"
+  ],
+  "milestone": {
+    "Zhong Ke Ke Shu Software Co., Ltd.province": "Guangdong Province",
+    "Guangdong ProvinceCommunication Transmission Equipmentlocal policies_Guangdong ProvinceNumber of policies (indicator)": 1,
+    "ChinaMetal Productslocal policies_Sichuan ProvincepersonsNumber of policies (indicator)": 1,
+    "Difference(Guangdong ProvinceNumber of policies (indicator)-Sichuan ProvincepersonsNumber of policies (indicator))": 0.0
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Zhong Ke Ke Shu Software Co., Ltd.province = Guangdong Province and industry = Communication Transmission Equipment",
+    "policy_release_status.csv in by province=Guangdong Province、industry=Communication Transmission Equipmentextractlocal policies in Guangdong ProvinceNumber of policies (indicator)=1",
+    "Extracted from policy_release_status.csv: ChinaMetal Productslocal policies in Sichuan ProvincepersonsNumber of policies (indicator)=1",
+    "Calculate difference: 1 - 1 = 0.0"
+  ],
+  "steps_num": 4,
+  "answer": 0.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "f1c9e833-ad21-4fe3-a5ca-32da9783df10"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "policy_release_status": "2d5ad892-2195-4230-854d-7c1dc5913951"
+    }
+  ]
+}
diff --git a/assets/qa_raw/enterprise_industry_policy_analysis/medium066_result.json b/assets/qa_raw/enterprise_industry_policy_analysis/medium066_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..445356e93434a6b7a14e9f4a5b623df1b6537f4f
--- /dev/null
+++ b/assets/qa_raw/enterprise_industry_policy_analysis/medium066_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium066",
+  "question": "Zhong Ke Ke Shu Software Co., Ltd.province Guangzhou CitypersonsNumber of policies (indicator) and China Hunan ProvincepersonsNumber of policies (indicator)compared with which unitsgreater？",
+  "guidelines": "The answer must a company name or \"industry\", Output onlyname, without any explanation or description.Wu, answer\"No relevant data found\".",
+  "evidence": [
+    "Zhong Ke Ke Shu Software Co., Ltd.province = Guangdong Province",
+    "Guangdong ProvinceCommunication Transmission Equipmentlocal policies_Guangzhou CitypersonsNumber of policies (indicator)=2",
+    "ChinaMetal Productslocal policies_Hunan ProvincepersonsNumber of policies (indicator)=1"
+  ],
+  "milestone": {
+    "Zhong Ke Ke Shu Software Co., Ltd.province": "Guangdong Province",
+    "Guangdong ProvinceCommunication Transmission Equipmentlocal policies_Guangzhou CitypersonsNumber of policies (indicator)": 2,
+    "ChinaMetal Productslocal policies_Hunan ProvincepersonsNumber of policies (indicator)": 1,
+    "Comparison result": "2greater，outputZhong Ke Ke Shu Software Co., Ltd."
+  },
+  "steps": [
+    "Extracted from company_profile.csv: Zhong Ke Ke Shu Software Co., Ltd.province = Guangdong Province and industry = Communication Transmission Equipment",
+    "policy_release_status.csv in by province=Guangdong Province、industry=Communication Transmission Equipmentextractlocal policies in Guangzhou CitypersonsNumber of policies (indicator)=2",
+    "Extracted from policy_release_status.csv: ChinaMetal Productslocal policies in Hunan ProvincepersonsNumber of policies (indicator)=1",
+    "Compare2 and 1, 2greater, output\"Zhong Ke Ke Shu Software Co., Ltd.\""
+  ],
+  "steps_num": 4,
+  "answer": "Zhong Ke Ke Shu Software Co., Ltd.",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "enterprise_industry_policy_analysis"
+  },
+  "reference": [
+    {
+      "policy_release_status": "f1c9e833-ad21-4fe3-a5ca-32da9783df10"
+    },
+    {
+      "company_profile": "7cf1e796-303e-45d7-a538-a41ce5972b04"
+    },
+    {
+      "policy_release_status": "2d5ad892-2195-4230-854d-7c1dc5913951"
+    }
+  ]
+}
diff --git a/assets/qa_raw/hypothesis_verification/hard001_result.json b/assets/qa_raw/hypothesis_verification/hard001_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae85b4618f6dab779e0bf2bfac6b32006af07b7d
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard001_result.json
@@ -0,0 +1,42 @@
+{
+  "id": "hard001",
+  "question": "作为研发驱动型产业，医药制造业的企业创新投入与收入表现之间的关联一直备受研究者关注，而地方政策环境可能影响这种关联的强弱。请以2022年数据为基础，将医药制造业上市企业按所在省份是否出台了地方生物医药产业发展促进政策(政策名称含生物医药，且含发展或促进)分为两组，分别计算两组企业的研发投入占比与营业收入同比增减幅之间的斯皮尔曼等级相关系数，并报告两个相关系数的差值（有政策省份系数减去无政策省份系数）。",
+  "guidelines": "依次回答出台政策省份的相关系数和未出台政策省份的相关系数及两者差值。相关系数保留4位小数，差值保留2位小数。如[\"0.2356\", \"-0.1048\", \"0.34\"]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"医药\"的政策记录，找到80条医药相关政策，其中地方政策55条、部委/国务院政策25条。",
+    "从policy_resource.csv中读取并匹配上述地方政策，按“政策名称含生物医药，且含发展或促进”识别“地方生物医药产业发展促进政策”，共得到8条，政策id为：92、141、381、430、431、432、433、436。",
+    "根据这8条政策的省份归属，提取出台了生物医药产业促进政策的省份共7个：上海市（id:92）、江苏省（id:430）、云南省（id:141）、广东省（id:431）、安徽省（id:432）、浙江省（id:381、433）、天津市（id:436）。",
+    "从company_profile.csv筛选行业=\"医药制造业\"的企业，共449家。与company_operation_status.csv关联后，筛选研发投入占比和营业收入同比增减幅均非空的有效企业，共417家。",
+    "将417家有效企业按省份分为两组：有生物医药政策省份200家（上海市45家、广东省46家、浙江省45家、江苏省46家、天津市8家、云南省6家、安徽省4家），无生物医药政策省份217家。",
+    "分别计算两组企业研发投入占比与营业收入同比增减幅的斯皮尔曼等级相关系数：有政策省份r=0.1142（p=0.1075，不显著），无政策省份r=-0.2132（p=0.0016，显著）。",
+    "计算差值 = 0.1142 - (-0.2132) = 0.33。有政策省份研发投入与营收增长呈弱正相关（不显著），无政策省份呈显著负相关，差值为0.33。"
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "从policy_release_status.csv中找到80条医药相关政策。",
+    "从policy_resource.csv中分析80条政策全文，筛选出8条正文内容与\"生物医药\"的相关的地方产业促进政策，涉及7个省份。",
+    "从company_profile.csv中找到449家医药制造业企业。",
+    "从company_operation_status.csv中获取这些企业的研发投入占比和营业收入同比增减幅数据，有效企业417家。"
+  ],
+  "milestone": {
+    "医药相关政策总数（条）": 80,
+    "标题含生物医药的地方政策数（条）": 8,
+    "涉及省份数（个）": 7,
+    "医药制造业有效企业数（家）": 417,
+    "有政策省份企业数（家）": 200,
+    "无政策省份企业数（家）": 217,
+    "有政策省份斯皮尔曼系数": 0.1142,
+    "无政策省份斯皮尔曼系数": -0.2132,
+    "相关系数差值": 0.33
+  },
+  "answer": [
+    0.1142,
+    -0.2132,
+    0.33
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/hard002_result.json b/assets/qa_raw/hypothesis_verification/hard002_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6497ab15711dfa9e220174a8bf1960e26d03110
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard002_result.json
@@ -0,0 +1,33 @@
+{
+  "id": "hard002",
+  "question": "有观点认为，在推动消费电子及电气业数字化转型与智能制造升级方面，国家层面的产业政策与地方政府政策在表述深度和覆盖方向上存在系统性差异。请检验这一判断：针对所有与消费电子及电气业相关的政策文件，分别统计国家级政策（含国务院及各部委发布的政策）与地方级政策中，明确提出数字化转型或智能制造相关目标或措施的政策数量及其占各自总数的比例，并给出国家级覆盖率减去地方级覆盖率的差值（以百分点计）。",
+  "guidelines": "依次回答国家级政策覆盖率、地方级政策覆盖率、差值（国家级占比减去地方级占比）。覆盖率和差值均以百分点表示，保留2位小数。如[80.00, 71.43, 8.57]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv中筛选industry字段包含\"消费电子\"的政策记录，共找到10条消费电子及电气业相关政策。按policyClassification字段将其划分为国家级（国务院政策+部委政策）和地方级（地方政策）两类：国家级政策4条（id: 19、150、156、167），地方级政策6条（id: 15、89、96、253、303、375）。",
+    "从policy_resource.csv中读取上述10条政策的全文内容，逐条分析是否明确提出数字化转型或智能制造相关目标或措施（关键词包括：数字化转型、数字化、数智化、智能制造、智能化、工业互联网、两化融合等）。",
+    "对4条国家级政策全文进行内容分析：id=19（加快电力装备绿色低碳创新发展行动计划）正文极短，仅为转发通知，无数字化转型或智能制造相关目标措施，判定为不包含；id=150（工业和信息化部关于开展2022\"三品\"全国行活动的通知）明确提出\"加快推进数字化助力消费品工业\"三品\"战略实施\"，包含；id=156（五部门数字化助力消费品工业\"三品\"行动方案）全文以数字化为核心主题，包含；id=167（推进国家级质量标准实验室建设的指导意见）明确提出\"围绕质量管理数字化\"等数字化目标，包含。国家级政策中含数字化转型/智能制造目标的有3条。",
+    "对6条地方政策全文进行内容分析：id=15（广东省进一步促进工业经济平稳增长措施）明确提出推动企业开展\"高端化、智能化、绿色化技术改造\"，包含；id=89（江西省打造全国新兴产业培育发展高地实施方案）明确提出\"产业链核心环节数字化转型\"目标，包含；id=96（重庆市促进大中小企业融通发展工作方案）明确提出\"加快全产业链数字化、网络化转型\"，包含；id=253（成都市\"十四五\"制造业高质量发展规划）正文为简短转发通知，无数字化/智能制造相关目标措施，不包含；id=303（海南省激励企业上规模奖励资金管理实施细则）内容为企业产值达标奖励规则，无数字化转型内容，不包含；id=375（四川省承接制造业有序转移实施意见）明确提出\"推动传统劳动密集型产业向数字化、智能化、高端化转型升级\"，包含。地方政策中含数字化转型/智能制造目标的有4条。",
+    "计算国家级政策覆盖率：3 ÷ 4 × 100% = 75.00%；计算地方级政策覆盖率：4 ÷ 6 × 100% ≈ 66.67%；差值 = 75.00% - 66.67% = 8.33个百分点，国家级政策覆盖率高于地方级。"
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "从policy_release_status.csv中找到10条消费电子及电气业相关政策，其中国家级（部委政策）4条，地方级政策6条。",
+    "从policy_resource.csv中读取并分析10条政策全文内容，国家级政策中3条明确包含数字化转型或智能制造目标/措施，1条不包含；地方级政策中4条包含，2条不包含。"
+  ],
+  "milestone": {
+    "消费电子及电气业相关政策总数（条）": 10,
+    "国家级政策总数（条）": 4,
+    "国家级含数字化转型/智能制造政策数（条）": 3,
+    "地方级政策总数（条）": 6,
+    "地方级含数字化转型/智能制造政策数（条）": 4,
+    "国家级政策覆盖率（%）": 75.00,
+    "地方级政策覆盖率（%）": 66.67,
+    "覆盖率差值（百分点）": 8.33
+  },
+  "answer": [75.00, 66.67, 8.33],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
diff --git a/assets/qa_raw/hypothesis_verification/hard003_result.json b/assets/qa_raw/hypothesis_verification/hard003_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f33abac6f9c8eaff26b00137ca87da69111fb73
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard003_result.json
@@ -0,0 +1,43 @@
+{
+  "id": "hard003",
+  "question": "在2022年通用设备制造业上市企业中，计算有地方制造业创新与科技发展促进政策支撑的省份（有政策省份）与无政策省份的企业政府补贴金额与年度中国发明专利申请数之间的斯皮尔曼等级相关系数，并给出两组系数之差（有政策省份系数减无政策省份系数）。要求依次回答：有政策省份的相关系数、无政策省份的相关系数、两组系数之差（均保留4位小数）。",
+  "guidelines": "依次回答有政策省份的相关系数、无政策省份的相关系数和两者差值（有政策-无政策）。相关系数和差值均保留4位小数。如[\"0.6812\", \"0.3590\", \"0.3222\"]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选policyClassification为\"地方政策\"且industry包含\"通用设备\"的政策记录，找到36条涉及通用设备制造业的地方政策。",
+    "从policy_resource.csv筛选政策正文涉及制造业创新与科技发展促进的政策，得到8条地方制造业创新与科技发展促进政策（id:75，78，175，238，385，541，562，590），涉及7个省份：上海市、宁夏回族自治区、安徽省、广东省、新疆维吾尔自治区、福建省、陕西省。",
+    "从policy_resource.csv中读取这8条政策的全文内容，分析发现这些政策主要涉及制造业创新中心建设（上海市、福建省）、专精特新中小企业倍增培育（安徽省）、科技型企业创新发展倍增（陕西省）、创新产品研制引导（宁夏回族自治区）、创新链产业链融合发展（广东省广州市）、未来产业创新高地建设（上海市）、技术创新中心建设（新疆维吾尔自治区）和长三角科技创新共同体建设（上海市）等方向，政策内容均强调对通用设备等先进制造业企业的创新支持和补贴引导。",
+    "从company_profile.csv筛选行业为\"通用设备制造业\"的企业，共213家。",
+    "从company_operation_status.csv提取这213家企业的政府奖励资金、补贴和年度中国发明专利申请数，筛选两项指标均非空的有效企业，得到189家。其中有政策省份44家，无政策省份145家。",
+    "计算有政策省份组（44家企业）政府补贴与年度中国发明专利申请数的斯皮尔曼等级相关系数 = 0.7531。",
+    "计算无政策省份组（145家企业）政府补贴与年度中国发明专利申请数的斯皮尔曼等级相关系数 = 0.4245。",
+    "两者差值 = 0.7531 - 0.4245 = 0.3285。有政策省份的政府补贴与专利产出之间呈现更强的正相关关系，表明在出台了创新科技促进政策的省份中，政府补贴对企业专利产出的激励效应显著更强。"
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "从policy_release_status.csv中找到36条涉及通用设备制造业的地方政策，其中9条与制造业创新和科技发展促进相关，涉及7个省份。",
+    "从policy_resource.csv中分析9条政策全文，确认其内容涉及制造业创新中心建设、专精特新企业培育、科技型企业倍增、创新链产业链融合等方向。",
+    "从company_profile.csv中找到213家通用设备制造业企业。",
+    "从company_operation_status.csv中获取企业政府补贴和年度中国发明专利申请数数据，189家企业数据有效。"
+  ],
+  "milestone": {
+    "涉及通用设备的地方政策总数（条）": 36,
+    "创新科技主题地方政策数（条）": 8,
+    "涉及省份数（个）": 7,
+    "通用设备制造业企业总数（家）": 213,
+    "有效企业数（家）": 189,
+    "有政策省份有效企业数（家）": 44,
+    "无政策省份有效企业数（家）": 145,
+    "有政策省份斯皮尔曼相关系数": 0.7531,
+    "无政策省份斯皮尔曼相关系数": 0.4245
+  },
+  "answer": [
+    0.7531,
+    0.4245,
+    0.3285
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/hard004_result.json b/assets/qa_raw/hypothesis_verification/hard004_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5be785cee6cbac1f3330703f2cb2dc0a6cef7491
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard004_result.json
@@ -0,0 +1,43 @@
+{
+  "id": "hard004",
+  "question": "在2022年纺织鞋服业上市企业中对比政策支持对企业盈亏的影响，计算有地方制造业转型升级政策支撑的省份（有政策省份）与无政策省份的企业营业利润亏损占比，并给出两组占比之差（有政策省份占比减无政策省份占比，以百分点计）。要求依次回答：有政策省份亏损占比、无政策省份亏损占比、两组占比的差值（均保留2位小数）。",
+  "guidelines": "依次回答有政策省份亏损占比、无政策省份亏损占比和差值（有政策组占比减去无政策组占比）。占比以百分数表示，保留2位小数。如[38.46, 27.50, 10.96]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选涉及行业包含\"纺织鞋服业\"的地方政策，共找到14条。",
+    "从policy_resource.csv中读取这14条政策全文，对每条政策进行深度分析，判断其是否属于制造业转型升级类政策：含有\"转型升级\"、\"智能制造\"、\"绿色制造\"、\"制造业创新\"、\"数字化转型\"、\"技术改造\"、\"升级改造\"、\"高端化\"等核心政策目标词汇，且非以节能减排约束或落后产能退出为主要定位的政策。",
+    "经过政策内容分析，认定以下10条为制造业转型升级政策，并匹配policy_resource.csv中的id：广东省工业经济平稳增长措施（id=15）、福建省制造业创新中心名单（id=22）、上海市制造业创新中心建设工程实施方案（id=75）、成都市“十四五”制造业高质量发展规划（id=253）、四川省承接制造业有序转移实施意见（id=375）、广西壮族自治区强龙头壮产业行动（id=263）、山东省新旧动能转换重大产业攻关项目管理实施细则（id=274）、湖南省智能制造标杆示范行动实施方案（id=276）、河北省推进规模以上工业企业培育工作若干措施（id=409）、新疆维吾尔自治区技术创新中心建设工作指引（id=541）。涉及省份9个：广东省、福建省、上海市、四川省、广西壮族自治区、山东省、湖南省、河北省、新疆维吾尔自治区。",
+    "从company_profile.csv筛选industry='纺织鞋服业'的企业，共177家，分布于19个省份。",
+    "从company_operation_status.csv获取这177家企业的营业利润金额数据，所有企业营业利润均非空，有效企业共177家。",
+    "按是否在有政策省份（上海市、四川省、山东省、广东省、广西壮族自治区、新疆维吾尔自治区、河北省、湖南省、福建省）将177家有效企业分为两组：有政策组80家，无政策组97家。",
+    "计算各组亏损（营业利润金额<0）企业占比：有政策组亏损企业35家，占比=35/80×100%=43.75%；无政策组亏损企业31家，占比=31/97×100%=31.96%；差值=43.75%-31.96%=11.79个百分点。"
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "从policy_release_status.csv中找到14条纺织鞋服业相关地方政策。",
+    "从policy_resource.csv中分析14条政策全文，认定10条为制造业转型升级政策，涉及9个省份。",
+    "从company_profile.csv中找到纺织鞋服业企业177家，分布于19个省份。",
+    "从company_operation_status.csv中获取177家企业的营业利润数据，所有企业数据完整，有效企业共177家，其中有政策省份80家、无政策省份97家。"
+  ],
+  "milestone": {
+    "纺织鞋服业相关地方政策数（条）": 14,
+    "制造业转型升级政策数（条）": 10,
+    "有政策省份数（个）": 9,
+    "纺织鞋服业有效企业总数（家）": 177,
+    "有政策组有效企业数（家）": 80,
+    "无政策组有效企业数（家）": 97,
+    "有政策组亏损企业数（家）": 35,
+    "无政策组亏损企业数（家）": 31,
+    "有政策组亏损占比（%）": 43.75,
+    "无政策组亏损占比（%）": 31.96
+  },
+  "answer": [
+    43.75,
+    31.96,
+    11.79
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/hard005_result.json b/assets/qa_raw/hypothesis_verification/hard005_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..97ce6d7cc0edc543c3451d24f7721655d4e9b1f3
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard005_result.json
@@ -0,0 +1,48 @@
+{
+  "id": "hard005",
+  "question": "在化学原料和化学制品制造业中，碳达峰与节能减排政策的推进可能给部分企业带来额外合规成本，由此引发一种反常现象：企业获得的政府补贴较高（以全体有效企业政府补贴金额的中位数作为划定高补贴的分界点），但利润同比却在下滑。本题以2022年度该行业的上市企业为分析对象。请分别计算出台了地方碳达峰或节能减排促进政策的省份、以及未出台此类政策的省份中，反常企业占各组有效企业总数的比例，以及两组比例之差（有政策省份减无政策省份，以百分点计）。",
+  "guidelines": "依次回答有政策省份的反常企业占比、无政策省份的反常企业占比和两组占比的差值。占比和差值均以百分数表示，保留2位小数。如[32.14, 21.05, 11.09]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"化学原料\"且policyClassification为\"地方政策\"的记录，找到48条化学原料相关地方政策。",
+    "从policy_resource.csv中读取这48条政策全文，按“碳达峰/节能减排/绿色低碳转型”口径筛得12条相关政策，policy id为：575、116、163、489、106、36、7、477、268、376、512、161。",
+    "上述12条政策主要包括碳达峰实施方案、节能减排综合工作方案、清洁生产推行方案和绿色低碳转型实施意见等；对应省份共10个：上海市、四川省、宁夏回族自治区、安徽省、江西省、河南省、湖南省、甘肃省、贵州省、辽宁省。",
+    "从company_profile.csv筛选行业=\"化学原料和化学制品制造业\"的企业共364家，与company_operation_status.csv关联获取运营数据。",
+    "筛选政府奖励资金、补贴和营业利润同比增减幅均非空的有效企业，共361家。其中有政策省份99家，无政策省份262家。",
+    "计算全行业361家有效企业的政府奖励资金、补贴中位数为10050282.75元。",
+    "在有政策省份99家企业中，政府补贴高于中位数的有56家，其中营业利润同比下滑的有29家，反常企业占比=29/99×100%=29.29%。",
+    "在无政策省份262家企业中，政府补贴高于中位数的有124家，其中营业利润同比下滑的有64家，反常企业占比=64/262×100%=24.43%。",
+    "两组占比的差值=29.29%-24.43%=4.87个百分点，有政策省份的反常占比反而更高。"
+  ],
+  "steps_num": 9,
+  "evidence": [
+    "从policy_release_status.csv中找到48条化学原料和化学制品制造业相关地方政策。",
+    "从policy_resource.csv中分析48条政策全文，筛选出12条涉及碳达峰、节能减排或绿色低碳转型的政策，涉及10个省份。",
+    "从company_profile.csv中找到364家化学原料和化学制品制造业企业。",
+    "从company_operation_status.csv中获取这些企业的政府奖励资金、补贴和营业利润同比增减幅数据，有效企业361家。"
+  ],
+  "milestone": {
+    "化学原料地方政策总数（条）": 48,
+    "碳达峰/节能减排相关政策数（条）": 12,
+    "涉及省份数（个）": 10,
+    "全行业有效企业数（家）": 361,
+    "有政策省份有效企业数（家）": 99,
+    "无政策省份有效企业数（家）": 262,
+    "全行业政府补贴中位数（元）": 10050282.75,
+    "有政策省份高补贴企业数（家）": 56,
+    "有政策省份反常企业数（家）": 29,
+    "有政策省份反常占比（%）": 29.29,
+    "无政策省份高补贴企业数（家）": 124,
+    "无政策省份反常企业数（家）": 64,
+    "无政策省份反常占比（%）": 24.43
+  },
+  "answer": [
+    "29.29",
+    "24.43",
+    "4.87"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/hard006_result.json b/assets/qa_raw/hypothesis_verification/hard006_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..084de15b8499dbaee842ae7f2bbfb0bced5ddf74
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard006_result.json
@@ -0,0 +1,44 @@
+{
+  "id": "hard006",
+  "question": "在2022年铁路、船舶、航空航天和其他运输设备制造业上市企业中，计算有地方先进制造与装备产业促进政策支撑的省份（有政策省份）与无政策省份的企业总资产与累计中国发明专利授权数之间的斯皮尔曼等级相关系数，并给出两组系数之差（有政策省份系数减无政策省份系数，保留2位小数）。要求依次回答：有政策省份的相关系数、无政策省份的相关系数、两组系数的差值。",
+  "guidelines": "依次回答有政策省份的相关系数、无政策省份的相关系数和两者差值（有政策-无政策）。相关系数保留2位小数。如[\"0.63\", \"0.48\", \"0.15\"]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"铁路\"或\"船舶\"或\"航空\"的政策记录，找到46条铁路、船舶、航空航天和其他运输设备制造业相关政策。",
+    "筛选其中policyClassification为\"地方政策\"的记录，得到36条地方政策。",
+    "从policy_resource.csv中读取这36条地方政策的全文内容，分析哪些政策涉及先进制造、装备制造、智能制造、首台套或重大技术装备等内容，筛得21条含相关内容的地方政策，policy id为：42、75、78、87、89、139、153、154、175、176、189、238、274、276、303、375、385、448、476、541、590；涉及13个省份：上海市、四川省、天津市、安徽省、山东省、广东省、新疆维吾尔自治区、江西省、河南省、海南省、湖南省、福建省、陕西省。",
+    "从company_profile.csv筛选industry=\"铁路、船舶、航空航天和其他运输设备制造业\"的企业，共99家。",
+    "关联company_operation_status.csv获取总资产和累计中国发明专利授权数，排除任一指标为空的企业后，得到96家有效企业。其中有政策省份36家，无政策省份60家。",
+    "分别计算两组企业总资产与累计中国发明专利授权数的斯皮尔曼等级相关系数：有政策省份ρ=0.5589≈0.56，无政策省份ρ=0.7216≈0.72。",
+    "计算差值：0.56-0.72=-0.16。无政策省份的规模-创新相关性反而更强。"
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "从policy_release_status.csv中找到46条铁路、船舶、航空航天和其他运输设备制造业相关政策，其中36条为地方政策。",
+    "从policy_resource.csv中分析36条地方政策全文，筛选出21条含先进制造/装备制造/智能制造内容的政策，涉及13个省份。",
+    "从company_profile.csv中找到99家铁路、船舶、航空航天和其他运输设备制造业企业。",
+    "从company_operation_status.csv中获取这些企业的总资产和累计中国发明专利授权数数据，96家有效。"
+  ],
+  "milestone": {
+    "铁路船舶航空相关政策总数（条）": 46,
+    "地方政策数（条）": 36,
+    "含先进制造/装备相关内容的地方政策数（条）": 21,
+    "涉及省份数（个）": 13,
+    "行业企业总数（家）": 99,
+    "有效企业数（家）": 96,
+    "有政策省份有效企业数（家）": 36,
+    "无政策省份有效企业数（家）": 60,
+    "有政策省份斯皮尔曼相关系数": 0.56,
+    "无政策省份斯皮尔曼相关系数": 0.72,
+    "差值": -0.16
+  },
+  "answer": [
+    "0.56",
+    "0.72",
+    "-0.16"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/hard007_result.json b/assets/qa_raw/hypothesis_verification/hard007_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..13fd03bd41379e748bafcb209d3e0586e3efb2b9
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard007_result.json
@@ -0,0 +1,51 @@
+{
+  "id": "hard007",
+  "question": "在2022年非金属矿物制品业上市企业中研究政策对不同规模企业营业利润率的影响，计算已出台专项推动建材行业碳达峰或节能减排政策的省份（有政策省份）中大型企业与小型企业的平均营业利润率差距（有政策省份规模差距），以及未出台此类政策的省份（无政策省份）中同一差距（无政策省份规模差距），并计算两者的差值（有政策省份规模差距减无政策省份规模差距）。要求依次回答：有政策省份规模差距、无政策省份规模差距、两类省份差距之差（均以百分点表示，保留2位小数）。",
+  "guidelines": "依次回答有政策省份规模差距、无政策省份规模差距、两类省份差距之差。均以百分点表示，保留2位小数。如[5.46, 1.23, 4.23]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"非金属矿物制品业\"的地方政策记录，共找到12条。",
+    "从policy_resource.csv中读取这12条政策的全文内容，分析政策是否涉及建材行业（水泥、玻璃、陶瓷等非金属矿物制品）的碳达峰、节能减排或绿色低碳要求；匹配到的政策id为：42、116、163、489、106、36、7、477、268、376、512、161。经全文分析，确认这些政策对应省份共9个：湖南省、河南省、四川省、江西省、辽宁省、宁夏回族自治区、甘肃省、贵州省、安徽省。",
+    "从company_profile.csv筛选industry=\"非金属矿物制品业\"的企业，共找到125家有效企业（总资产、营业利润金额、营业收入金额均非空且营业收入非零）。",
+    "按总资产全行业三分位分层：Q33=2,805,969,330元，Q67=10,145,495,897元。大型企业（总资产>Q67）42家，小型企业（总资产<=Q33）42家。",
+    "按省份分组：有政策省份（9个）有效企业34家（大型13家、小型12家），无政策省份有效企业91家（大型29家、小型30家）。",
+    "从company_operation_status.csv提取营业利润金额和营业收入金额，计算各企业营业利润率=营业利润金额/营业收入金额×100%。",
+    "计算有政策省份规模差距=大型企业平均利润率-小型企业平均利润率=11.02%-3.30%=7.72个百分点。",
+    "计算无政策省份规模差距=7.37%-6.68%=0.69个百分点。",
+    "计算差距之差=7.72%-0.69%=7.03个百分点。"
+  ],
+  "steps_num": 9,
+  "evidence": [
+    "从policy_release_status.csv中找到12条非金属矿物制品业相关地方政策。",
+    "从policy_resource.csv中分析12条政策全文，确认9个省份的政策涉及建材行业碳达峰/节能减排要求。",
+    "从company_profile.csv中找到125家非金属矿物制品业有效企业。",
+    "从company_operation_status.csv中获取125家企业营业利润和营业收入数据，计算营业利润率。"
+  ],
+  "milestone": {
+    "碳达峰/节能减排政策条数（条）": 12,
+    "有政策省份数（个）": 9,
+    "非金属矿物制品业有效企业总数（家）": 125,
+    "全行业总资产Q33（元）": 2805969329.91,
+    "全行业总资产Q67（元）": 10145495897.25,
+    "有政策省份大型企业数（家）": 13,
+    "有政策省份小型企业数（家）": 12,
+    "有政策省份大型企业平均营业利润率（%）": 11.02,
+    "有政策省份小型企业平均营业利润率（%）": 3.3,
+    "有政策省份规模差距（百分点）": 7.72,
+    "无政策省份大型企业数（家）": 29,
+    "无政策省份小型企业数（家）": 30,
+    "无政策省份大型企业平均营业利润率（%）": 7.37,
+    "无政策省份小型企业平均营业利润率（%）": 6.68,
+    "无政策省份规模差距（百分点）": 0.69,
+    "两类省份差距之差（百分点）": 7.03
+  },
+  "answer": [
+    7.72,
+    0.69,
+    7.03
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/hard008_result.json b/assets/qa_raw/hypothesis_verification/hard008_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7800f827b0b95444665bed699a3648fe6b22b676
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard008_result.json
@@ -0,0 +1,42 @@
+{
+  "id": "hard008",
+  "question": "2022年，集成电路产业的地方政策竞争进入白热化阶段，各省在专项激励力度上差异显著。本题统计口径说明如下：①统计对象为营业利润与营业收入数据均完整且营业收入非零的内地企业，港澳台地区企业不纳入；②营业利润率 = 营业利润 ÷ 营业收入 × 100%；③认定为专项集成电路产业促进政策，须是专门针对集成电路或半导体产业的地方政策，且明确包含流片补贴、企业落户奖励、研发设计人才支持、产业规模发展目标等专项措施中的至少一项，仅泛提数字经济或科技创新的通用政策不符合要求。在此基础上，请计算2022年半导体业中出台了上述专项政策的省份与未出台省份的企业平均营业利润率，并给出差值（有政策省份均值减去无政策省份均值，以百分点计）。",
+  "guidelines": "依次回答有政策省份平均营业利润率、无政策省份平均营业利润率、两者差值（有政策-无政策）。数值均保留2位小数，以百分点表示。如[10.55, 13.87, -3.32]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"半导体\"的政策记录，找到44条半导体相关政策；进一步筛选policyClassification为\"地方政策\"的记录，共35条。",
+    "从policy_resource.csv中读取这35条地方政策的全文内容，分析哪些政策是专门针对集成电路或半导体产业的专项促进政策（而非泛制造业政策中附带覆盖半导体）。经深度阅读政策正文，筛选出4条包含明确IC专项扶持措施的政策，policy id为：80、125、196、290。对应政策分别为：广东省横琴粤澳深度合作区促进集成电路产业发展若干措施（id=80，含实缴资本奖励最高500万元、总部项目奖励最高2000万元等落户奖励）、浙江省杭州市促进集成电路产业高质量发展实施意见（id=125，含到2025年产业规模实现800亿元发展目标及集成电路企业研发费用>5%要求）、上海市新时期促进集成电路产业和软件产业高质量发展若干政策（id=196，含研发设计人员奖励最高50万元及企业核心团队分级奖励）、安徽省合肥市加快推进集成电路产业发展若干政策（id=290，含流片补贴最高1000万元、EDA工具补贴最高200万元、IP研发补贴）。",
+    "经政策内容分析确认，出台专项集成电路产业促进政策的省份为4个：广东省（id=80）、浙江省（id=125）、上海市（id=196）、安徽省（id=290）。",
+    "从company_profile.csv筛选industry=\"半导体业\"且province不属于港澳台地区的内地企业，共160家，其中有政策省份（广东省、浙江省、上海市、安徽省）97家，无政策省份63家。",
+    "从company_operation_status.csv获取这160家企业的营业利润金额和营业收入金额，全部数据完整且营业收入非零，保留全部160家为有效企业。计算每家企业营业利润率=营业利润金额/营业收入金额×100%。",
+    "计算有政策省份97家企业的平均营业利润率：（各企业营业利润率之和）÷97=7.02%。其中广东省54家均值3.72%、上海市27家均值18.85%、浙江省13家均值-4.42%、安徽省3家均值9.57%。",
+    "计算无政策省份63家企业的平均营业利润率：（各企业营业利润率之和）÷63=16.33%。两组差值=7.02-16.33=-9.31个百分点，有政策省份企业平均营业利润率低于无政策省份，假说未被支持。"
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "从policy_release_status.csv中找到44条半导体相关政策，其中35条地方政策。",
+    "从policy_resource.csv中分析35条地方政策全文，筛选出4条含IC专项扶持措施的政策，涉及广东省、浙江省、上海市、安徽省共4个省份。",
+    "从company_profile.csv中找到内地半导体业企业160家，有政策省份97家，无政策省份63家。",
+    "从company_operation_status.csv中获取160家企业的营业利润金额和营业收入金额，全部有效。"
+  ],
+  "milestone": {
+    "半导体相关政策总数（条）": 44,
+    "地方半导体政策数（条）": 35,
+    "专项IC促进政策数（条）": 4,
+    "有政策省份数（个）": 4,
+    "内地半导体有效企业总数（家）": 160,
+    "有政策省份企业数（家）": 97,
+    "无政策省份企业数（家）": 63,
+    "有政策省份平均营业利润率（%）": 7.02,
+    "无政策省份平均营业利润率（%）": 16.33
+  },
+  "answer": [
+    7.02,
+    16.33,
+    -9.31
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/hard009_result.json b/assets/qa_raw/hypothesis_verification/hard009_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0db98b02010481a34d5dec8f963b844a85c7cca
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard009_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "hard009",
+  "question": "在2022年汽车制造业上市企业中验证政策对企业营业利润有更好的促进作用的假设，计算出台了新能源汽车产业专项促进政策省份（有政策省份）的企业营业利润同比增减幅中位数，未出台此类政策省份（无政策省份）的企业营业利润同比增减幅中位数，以及两者的差值（有政策省份中位数减无政策省份中位数，以百分点计）。要求依次回答：有政策省份企业中位数、无政策省份企业中位数、两组中位数之差（均保留2位小数，单位为百分点）。",
+  "guidelines": "依次回答有政策省份企业中位数、无政策省份企业中位数和差值。数值均保留2位小数，单位为百分点（%）。差值=有政策省份中位数-无政策省份中位数。如[4.16, -3.80, 7.96]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选policyClassification为地方政策且industry字段包含汽车的政策，共找到53条汽车制造业相关地方政策。",
+    "从policy_resource.csv读取这53条政策的全文内容，筛选专门指向新能源汽车推广、换电模式应用、燃料电池汽车示范或智能网联汽车管理的政策（即题干定义的新能源汽车产业专项促进政策），共找到11条，政策id为：58、157、165、206、251、322、484、503、580、584、586。经内容分析涉及7个省份：广东省、上海市、海南省、重庆市、四川省、山东省、江苏省。",
+    "从company_profile.csv筛选industry为汽车制造业的企业，排除港澳台地区企业，得到226家内地汽车制造业有效企业。",
+    "从company_operation_status.csv获取226家企业的营业利润同比增减幅数据，所有企业数据均完整。",
+    "按省份分组：有新能源汽车产业专项促进政策省份（7个）共118家企业；无新能源汽车产业专项促进政策省份共108家企业。",
+    "计算两组企业营业利润同比增减幅的中位数：有政策省份中位数=2.73%，无政策省份中位数=-5.52%。差值=2.73% - (-5.52%) = 8.25个百分点。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从policy_release_status.csv中找到53条汽车制造业相关地方政策。",
+    "从policy_resource.csv中分析53条政策全文，筛选出11条符合新能源汽车产业专项促进政策定义的政策，涉及7个省份。",
+    "从company_profile.csv中找到226家内地汽车制造业企业。",
+    "从company_operation_status.csv中获取226家企业的营业利润同比增减幅数据，数据全部完整。"
+  ],
+  "milestone": {
+    "汽车相关地方政策总数（条）": 53,
+    "新能源汽车产业专项促进政策数（条）": 11,
+    "有政策省份数（个）": 7,
+    "内地汽车制造业有效企业总数（家）": 226,
+    "有政策省份企业数（家）": 118,
+    "无政策省份企业数（家）": 108,
+    "有政策省份营业利润同比增减幅中位数（%）": 2.73,
+    "无政策省份营业利润同比增减幅中位数（%）": -5.52,
+    "两组中位数差值（百分点）": 8.25
+  },
+  "answer": [
+    2.73,
+    -5.52,
+    8.25
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/hard010_result.json b/assets/qa_raw/hypothesis_verification/hard010_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b248070a3813b4cb4f6855be3c0ba71949ab2309
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard010_result.json
@@ -0,0 +1,48 @@
+{
+  "id": "hard010",
+  "question": "在2022年金属冶炼和压延加工业上市企业中验证政策对民因企业有更好的资产收益率假设，计算有专项推动有色金属冶炼产业高质量发展政策省份（有政策省份）的所有制效率差距（国有均值减民营均值）、无政策省份的所有制效率差距，以及两类省份差距之差（有政策省份差距减无政策省份差距，以百分点计）。要求依次回答：有政策省份的所有制效率差距、无政策省份的所有制效率差距、两类省份差距之差（均以百分点表示，保留2位小数）。",
+  "guidelines": "依次回答有政策省份的所有制效率差距、无政策省份的所有制效率差距、两类省份差距之差，均以百分点表示，保留2位小数。如[-7.52, -2.10, -5.42]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选涉及行业包含\"金属冶炼和压延加工业\"的政策，共得到41条；其中地方政策32条，涉及省份包括云南省、四川省、河南省、贵州省等17个省份。",
+    "从policy_resource.csv中读取上述32条地方政策全文，逐一分析政策目标和主要措施，区分两类：一是以节能减排、碳达峰为主要目标的通用环保政策；二是专项推动有色金属冶炼（绿色铝、钒钛、功能材料等）产业高质量发展的专项产业政策。",
+    "经过深度分析，识别出4个省份出台专项有色金属冶炼产业高质量发展政策，对应政策id为：云南省（id=281《绿色铝产业发展三年行动》、id=128《新材料产业发展三年行动》）、四川省（id=511《促进钒钛产业高质量发展的实施意见》）、河南省（id=87《加快材料产业优势再造换道领跑行动计划》）、贵州省（id=266《支持铜仁市打造国家级新型功能材料战略性新兴产业集群的若干政策措施》）。",
+    "从company_profile.csv筛选行业为\"金属冶炼和压延加工业\"且省份不属于港澳台地区的企业，共139家；与company_operation_status.csv合并后，筛选营业利润金额和总资产均非空且总资产>0的有效企业，得到139家；剔除外资企业（3家）和集体企业（1家），保留国有企业（61家）和民营企业（75家）共136家。",
+    "按省份分组：有政策省份（云南省、四川省、河南省、贵州省）共18家企业（国有10家、民营8家）；无政策省份共118家（国有51家、民营67家）。",
+    "计算各分组的平均资产收益率（资产收益率=营业利润金额/总资产×100%）：有政策省份国有均值=6.29%，民营均值=16.18%，所有制效率差距=6.29%-16.18%=-9.89%；无政策省份国有均值=2.88%，民营均值=6.17%，所有制效率差距=2.88%-6.17%=-3.28%。",
+    "计算两类省份差距之差=(-9.89%) - (-3.28%) = -6.61个百分点。"
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "从policy_release_status.csv中找到41条金属冶炼和压延加工业相关政策，其中地方政策32条，涉及17个省份。",
+    "从policy_resource.csv中分析32条地方政策全文，识别出4个省份共6条专项推动有色金属冶炼产业高质量发展的政策。",
+    "从company_profile.csv中筛选到139家金属冶炼和压延加工业内地企业，按国有/民营分类后得到136家有效企业（国有61家、民营75家）。",
+    "从company_operation_status.csv中获取136家有效企业的营业利润金额和总资产数据，计算各分组的资产收益率均值及所有制效率差距。"
+  ],
+  "milestone": {
+    "金属冶炼地方政策总条数（条）": 32,
+    "专项有色金属产业高质量发展政策数（条）": 6,
+    "涉及有政策省份数（个）": 4,
+    "有政策省份有效企业数（家）": 18,
+    "无政策省份有效企业数（家）": 118,
+    "有政策省份国有企业数（家）": 10,
+    "有政策省份民营企业数（家）": 8,
+    "无政策省份国有企业数（家）": 51,
+    "无政策省份民营企业数（家）": 67,
+    "有政策省份国有企业均值资产收益率（%）": 6.29,
+    "有政策省份民营企业均值资产收益率（%）": 16.18,
+    "无政策省份国有企业均值资产收益率（%）": 2.88,
+    "无政策省份民营企业均值资产收益率（%）": 6.17,
+    "有政策省份所有制效率差距（百分点）": -9.89,
+    "无政策省份所有制效率差距（百分点）": -3.28
+  },
+  "answer": [
+    -9.89,
+    -3.28,
+    -6.61
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/hard011_result.json b/assets/qa_raw/hypothesis_verification/hard011_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..38edaed150762f3ff7c9b0849056b6c53740d2c2
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard011_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "hard011",
+  "question": "在2022年专用设备制造业上市企业中验证政策对头头部企业影响更大的假设，计算有地方重大技术装备专项促进政策或先进制造业专项法规省份（有政策省份）的CR20%、无政策省份的CR20%，以及两者的差值（有政策省份CR20%减无政策省份CR20%，以百分点计）。要求依次回答：有政策省份CR20%、无政策省份CR20%、差值（均保留2位小数，单位为百分点）。",
+  "guidelines": "依次回答有政策省份CR20%、无政策省份CR20%、差值。数值保留2位小数，以百分点表示。如[82.35, 70.14, 12.21]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选涉及专用设备制造业的地方政策，共找到47条。",
+    "从policy_resource.csv中读取这47条政策的全文内容，深入分析政策核心目标与具体措施，筛选以'推动重大技术装备创新示范应用'或'专项促进先进制造业/装备产业高端化（含工程机械等专用设备子行业为核心集群目标）'为主要政策目标的地方专项政策。经逐条分析，共识别出2条符合条件的政策：天津市促进首台（套）重大技术装备示范应用若干措施（id=176）和湖南省先进制造业促进条例（id=189）。两条政策涉及天津市和湖南省共2个省份。",
+    "从company_profile.csv筛选行业='专用设备制造业'的企业，共447家；关联company_operation_status.csv获取营业收入金额，排除港澳台地区企业及营业收入为空或为零的企业，得到有效内地企业440家（天津市10家、湖南省11家、其他内地省份419家）。",
+    "将有效企业分为两组：有政策省份（天津市、湖南省，共21家）和无政策省份（其他22个内地省份，共419家）。",
+    "计算有政策省份CR20%：21家企业，按营业收入降序排列，前20%企业数=ceil(21×0.2)=5家；前5家营业收入合计=1326.87亿元，21家总营业收入=1661.09亿元；CR20%=1326.87/1661.09×100%=79.88%。",
+    "计算无政策省份CR20%：419家企业，按营业收入降序排列，前20%企业数=ceil(419×0.2)=84家；前84家营业收入合计=9836.00亿元，419家总营业收入=13344.68亿元；CR20%=9836.00/13344.68×100%=73.71%。",
+    "差值（有政策-无政策）=79.88%-73.71%=6.17个百分点。"
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "从policy_release_status.csv中找到47条专用设备制造业相关地方政策。",
+    "从policy_resource.csv中逐条分析47条政策全文，识别出2条以重大技术装备或先进制造业（工程机械等专用设备子行业）为核心目标的专项政策，涉及天津市和湖南省2个省份。",
+    "从company_profile.csv中筛选专用设备制造业内地有效企业共440家（天津市10家、湖南省11家、其他内地省份419家）。",
+    "从company_operation_status.csv获取440家企业的营业收入金额，用于计算各组CR20%。"
+  ],
+  "milestone": {
+    "专用设备制造业地方政策总数（条）": 47,
+    "符合条件的专项政策数（条）": 2,
+    "政策涉及省份数（个）": 2,
+    "有政策省份有效企业数（家）": 21,
+    "无政策省份有效企业数（家）": 419,
+    "有政策省份前20%企业数（家）": 5,
+    "无政策省份前20%企业数（家）": 84,
+    "有政策省份前20%营业收入合计（亿元）": 1326.87,
+    "有政策省份营业收入总计（亿元）": 1661.09,
+    "无政策省份前20%营业收入合计（亿元）": 9836.0,
+    "无政策省份营业收入总计（亿元）": 13344.68,
+    "有政策省份CR20%（%）": 79.88,
+    "无政策省份CR20%（%）": 73.71
+  },
+  "answer": [
+    79.88,
+    73.71,
+    6.17
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/hard012_result.json b/assets/qa_raw/hypothesis_verification/hard012_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ab47ef874fc23c5933cdf2ad4c68442601e8c2fc
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard012_result.json
@@ -0,0 +1,45 @@
+{
+  "id": "hard012",
+  "question": "一个行业内营业利润率的分布离散程度，在一定条件下可反映企业间的竞争分化态势或政策的结构性影响效应。以四分位距（IQR = Q3减去Q1，均基于企业个体营业利润率的分布计算）作为离散程度的测量工具，对2022年橡胶和塑料制品业（排除港澳台）数据进行分析，营业利润率=营业利润金额/营业收入金额×100%。若以出台了面向制造业企业、按发展阶段设定梯度化现金奖励或培育支持机制且将橡胶和塑料制品业列为受益行业的专项产业培育激励政策的省份为一组，其余省份为另一组，两组企业营业利润率的IQR分别是多少个百分点？两组IQR的差值（有政策省份减无政策省份）为多少个百分点？",
+  "guidelines": "依次回答有政策省份的IQR、无政策省份的IQR、两者差值。数值以百分点表示，保留2位小数。如[5.82, 9.45, -3.63]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"橡胶\"的地方政策，共找到17条橡胶和塑料制品业相关地方政策，涉及湖南省、山东省、上海市、四川省、云南省、辽宁省、安徽省、广西壮族自治区、海南省、陕西省、河北省、新疆维吾尔自治区等12个省份。",
+    "从policy_resource.csv中读取上述17条地方政策的全文内容，对每条政策进行深度分析，识别其政策类型：（1）安徽省（id=175，安徽省专精特新中小企业倍增行动方案）：对省专精特新冠军企业给予一次性奖补80万元，对国家级专精特新小巨人和单项冠军企业分别奖补100万元、200万元，并设置从创新型中小企业到单项冠军的五级梯度培育通道，明确将橡胶和塑料制品业列为覆盖行业；（2）河北省（id=409，推进规模以上工业企业培育工作若干措施）：建立企业培育库、梯次升级培育机制，涵盖橡胶和塑料制品业，通过资金倾斜、要素保障推动临规企业升规壮大；（3）海南省（id=303，激励企业上规模奖励资金实施细则）：对年产值首次突破3亿至50亿的橡胶和塑料制品业企业分别给予30万至500万元一次性奖励，但海南省无上市橡胶和塑料制品业企业。其余14条政策均为节能减排综合工作实施方案、先进制造业条例、智能制造标杆示范方案、新材料目录、落后产能退出公告、农业现代化规划等非企业梯度激励类政策，或仅将橡胶和塑料制品业列为附带受益行业而无专项企业培育奖励措施。",
+    "确定政策分组：出台了面向制造业企业梯度化现金奖励或培育激励政策（且覆盖橡胶和塑料制品业）的省份为安徽省和河北省（对应政策id：175、409）；海南省虽有同类型政策（id=303）但无上市橡胶和塑料制品业企业。其余13个有橡胶和塑料制品业企业的内地省份均无此类政策。",
+    "从company_profile.csv筛选industry字段为\"橡胶和塑料制品业\"且province不属于港澳台的内地企业，共得到105家有效企业。其中安徽省8家、河北省3家（有政策组共11家），其余省份合计94家（无政策组）。",
+    "从company_operation_status.csv获取105家企业的营业利润金额和营业收入金额，均为非空值且营业收入均不为零，全部105家企业满足有效条件。计算各企业营业利润率 = 营业利润金额 / 营业收入金额 × 100%。",
+    "计算有政策省份（安徽省+河北省，n=11）的营业利润率IQR：Q1 = 7.9124%，Q3 = 12.2784%，IQR = 12.2784 − 7.9124 = 4.3660，保留2位小数为4.37个百分点。",
+    "计算无政策省份（n=94）的营业利润率IQR：Q1 = 1.6678%，Q3 = 13.2604%，IQR = 13.2604 − 1.6678 = 11.5926，保留2位小数为11.59个百分点。",
+    "计算差值：精确值 = 4.3660 − 11.5926 = −7.2266，保留2位小数为−7.23个百分点。"
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "从policy_release_status.csv中找到17条橡胶和塑料制品业相关地方政策，涉及12个省份。",
+    "从policy_resource.csv中精读17条政策全文，识别出2条具有企业梯度化现金奖励或培育激励机制且覆盖橡胶和塑料制品业的政策（安徽省专精特新倍增、河北省规模以上企业培育），以及1条同类型政策但所在省无橡胶企业（海南省），其余14条均为节能减排、先进制造业法规或目录类政策。",
+    "从company_profile.csv中找到105家内地橡胶和塑料制品业企业，有政策省份（安徽省+河北省）11家，无政策省份94家。",
+    "从company_operation_status.csv中获取105家企业的营业利润金额和营业收入金额，计算各企业营业利润率，所有105家企业均满足有效条件（非空且营业收入非零）。"
+  ],
+  "milestone": {
+    "橡胶和塑料制品业相关地方政策总数（条）": 17,
+    "具有企业梯度激励机制的地方政策数（条）": 2,
+    "有政策省份数（个）": 2,
+    "有政策组企业数（家）": 11,
+    "无政策组企业数（家）": 94,
+    "有政策组Q1（%）": 7.91,
+    "有政策组Q3（%）": 12.28,
+    "无政策组Q1（%）": 1.67,
+    "无政策组Q3（%）": 13.26,
+    "有政策省份IQR（百分点）": 4.37,
+    "无政策省份IQR（百分点）": 11.59
+  },
+  "answer": [
+    4.37,
+    11.59,
+    -7.23
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/hard013_result.json b/assets/qa_raw/hypothesis_verification/hard013_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c344f2cbdb0ab21f0367ed54fbb9713b61dfb6e
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard013_result.json
@@ -0,0 +1,50 @@
+{
+  "id": "hard013",
+  "question": "国家与地方政策的协同效应，在产业经济学中通常以政策叠加框架加以讨论——双重政策覆盖的企业是否在劳动效率上具有系统性优势，是检验政策层级互补性的核心命题之一。人均营业收入作为劳动效率的代理变量，以2022年食品饮料业为例，将企业所在省份按政策覆盖状态分为三组：第一组，省份同时被国家消费品工业促进政策明确列为活动实施省份且已出台地方食品相关产业政策；第二组，省份仅有地方食品相关产业政策、未被前述国家政策明确覆盖；第三组，上述两类政策均无。有效企业须营业收入金额非空且雇员总数为正的内地企业（排除港澳台）。三组企业的平均人均营业收入分别为多少万元？第一组与第二组的均值差为多少万元？",
+  "guidelines": "依次回答双重覆盖组、仅地方覆盖组、无政策覆盖组的平均人均营业收入（万元/人），以及双重覆盖组与仅地方覆盖组的差值（万元/人）。所有数值保留2位小数。如[185.30, 152.75, 126.40, 32.55]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"食品\"的政策，共找到16条食品相关政策。其中国家级（部委政策）3条、地方政策13条。",
+    "从policy_resource.csv中读取3条国家级食品相关政策全文，政策id为：140、150、156。其中国家级政策包括：《关于推动轻工业高质量发展的指导意见》（id=140，覆盖食品等8个轻工业子行业）、《关于开展2022“三品”全国行活动的通知》（id=150，明确将上海、江苏、浙江、福建、山东、湖北、广东、重庆、四川等9个省份列为活动实施地区）、《数字化助力消费品工业“三品”行动方案》（id=156，全国性数字化转型指导方案）。其中id=150通过点名9省形成差异化国家级覆盖，其余2条为全国普适性指导文件。",
+    "从policy_resource.csv中读取13条地方食品相关政策全文，地方政策id为：43、111、131、134、253、276、303、317、375、377、399、409、512。经分析，有地方食品相关政策的省份共9个：甘肃省（id=43）、河南省（id=111、317）、云南省（id=131、377）、四川省（id=253、375）、湖南省（id=276）、海南省（id=303）、宁夏回族自治区（id=399）、河北省（id=409）、贵州省（id=512）。另有1条吉林省政策（id=134）在policy_release_status.csv中province字段标记为全国，不纳入地方政策省份统计。",
+    "将省份按政策覆盖层级分为三组：双重覆盖组（省份同时在国家三品全国行9省名单内且有地方食品相关政策）= 四川省；仅地方覆盖组（有地方食品相关政策但不在国家三品全国行9省名单内）= 甘肃省、河南省、云南省、湖南省、海南省、宁夏回族自治区、河北省、贵州省（8个省份）；无政策覆盖组（不在国家三品全国行9省名单内且无地方食品相关政策）= 天津市、安徽省、吉林省等13个省份。",
+    "从company_profile.csv筛选industry为食品饮料业且排除港澳台地区的内地企业，共234家。将各企业按所属省份分配到三组：双重覆盖组9家（四川省）、仅地方覆盖组37家（8省合计）、无政策覆盖组59家（13省合计）。",
+    "从company_operation_status.csv获取234家企业的营业收入金额和雇员总数，筛选有效企业（营业收入非空且雇员总数>0）：双重覆盖组有效9家、仅地方覆盖组有效37家、无政策覆盖组有效58家（1家因雇员总数缺失被排除）。",
+    "按组计算各企业人均营业收入（= 营业收入金额 / 雇员总数），再取组内均值：双重覆盖组均值 = 2019384.10元 / 10000 = 201.94万元/人；仅地方覆盖组均值 = 1681974.75元 / 10000 = 168.20万元/人；无政策覆盖组均值 = 1409640.38元 / 10000 = 140.96万元/人；双重覆盖组与仅地方覆盖组差值 = 201.94 - 168.20 = 33.74万元/人。"
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "从policy_release_status.csv中找到16条食品相关政策，其中国家级（部委）3条、地方13条。",
+    "从policy_resource.csv中分析3条国家级食品政策全文，识别出三品全国行明确列出9个活动实施省份：上海、江苏、浙江、福建、山东、湖北、广东、重庆、四川。",
+    "从policy_resource.csv中分析12条有明确省份的地方食品政策全文，确认覆盖9个省份：甘肃、河南、云南、四川、湖南、海南、宁夏、河北、贵州。",
+    "从company_profile.csv中筛选出234家食品饮料业内地企业，分配至三个省份政策覆盖组。",
+    "从company_operation_status.csv中获取各企业营业收入和雇员总数，共233家有效企业（排除1家雇员数据缺失）参与计算。"
+  ],
+  "milestone": {
+    "食品相关政策总数（条）": 16,
+    "国家级食品政策数（条）": 3,
+    "地方食品政策数（条，省份非全国）": 12,
+    "三品全国行明确列名省份数（个）": 9,
+    "有地方食品政策的省份数（个）": 9,
+    "双重覆盖省份数（个）": 1,
+    "仅地方覆盖省份数（个）": 8,
+    "无政策覆盖省份数（个）": 13,
+    "双重覆盖有效企业数（家）": 9,
+    "仅地方覆盖有效企业数（家）": 37,
+    "无政策覆盖有效企业数（家）": 58,
+    "双重覆盖组平均人均营业收入（万元/人）": 201.94,
+    "仅地方覆盖组平均人均营业收入（万元/人）": 168.2,
+    "无政策覆盖组平均人均营业收入（万元/人）": 140.96,
+    "双重覆盖组与仅地方覆盖组差值（万元/人）": 33.74
+  },
+  "answer": [
+    201.94,
+    168.20,
+    140.96,
+    33.74
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/hard014_result.json b/assets/qa_raw/hypothesis_verification/hard014_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..60c46bbf3db5c2d98870f70107ff0c2eabcf5ce8
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard014_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "hard014",
+  "question": "针对验证政府补贴能高效地转化为了企业的创新产出对于不同企业存在差异的假设。以通信传输设备业为例，这一转化效率在有无专项研发创新激励政策的省份之间是否存在差异。认定的专项扶持企业研发创新的地方促进政策须同时满足：①政策涵盖通信传输设备业；②政策明确包含金额确定的企业研发奖励、科技创新补贴或专项创新资金等直接企业激励措施。有效企业指政府补贴金额大于0且年度中国发明专利申请数有完整记录的内地企业。请以每百万元政府补贴所对应的年度发明专利申请数作为衡量补贴转化效率的指标，分别计算并给出2022年有政策省份与无政策省份中有效企业的该指标均值，并给出差值（有政策省份均值减去无政策省份均值）。",
+  "guidelines": "依次回答有政策省份的补贴转化效率、无政策省份的补贴转化效率、两者差值。均保留2位小数，单位为件/百万元。如[3.05, 2.47, 0.58]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv中筛选涉及行业包含「通信传输设备」且政策分类为「地方政策」的记录，共得到46条地方政策，涉及18个省份（含少数省份归属「全国」的政策）。",
+    "从policy_resource.csv中读取上述46条地方政策全文，逐条分析政策是否满足以下两个条件：（1）政策明确将通信传输设备业列为主要支持行业之一；（2）政策正文中包含直接面向企业的、金额明确的创新研发奖励或补贴措施（如专精特新企业研发奖励、科技型企业研发投入补助、创新平台建设补贴等）。经深度内容分析，筛选出5个省份的政策符合上述标准，对应政策id为：湖北省北斗产业高质量发展政策（id=69，对专精特新企业给予50万-100万元一次性奖励，对研发平台给予最高1000万元补助）、陕西省科技型企业创新发展倍增计划（id=238，对研发成果给予最高40%/200万元奖励，对专精特新企业研发新增投入补贴最高500万元）、广东省促进工业经济平稳增长政策（id=15，对通信传输设备等制造业企业实施研发投入奖补）、安徽省专精特新中小企业倍增行动方案（id=175，对专精特新冠军/小巨人企业给予80万-200万元创新奖补，含5G等新型信息技术应用奖补）、上海市推进高端制造业发展若干措施（id=386，对5G/工业互联网应用场景项目给予最高800万元奖励，对制造业创新平台给予最高2000万元支持）。",
+    "从company_profile.csv中筛选行业为「通信传输设备业」且省份不在港澳台的内陆企业，共得到118家，分属17个省份/直辖市。按政策分组：湖北、陕西、广东、安徽、上海5省共62家（有政策组），其余12省共56家（无政策组）。",
+    "从company_operation_status.csv中关联企业运营数据，筛选有效企业（政府奖励资金、补贴>0且年度中国发明专利申请数非空）：有政策组56家，无政策组43家，合计99家有效企业。",
+    "计算各企业的补贴转化效率 = 年度中国发明专利申请数 / (政府奖励资金、补贴 / 1,000,000)，单位为件/百万元。有政策组：56家企业效率值之和 / 56 = 2.38 件/百万元；无政策组：43家企业效率值之和 / 43 = 3.18 件/百万元。",
+    "计算差值：有政策省份均值 - 无政策省份均值 = 2.38 - 3.18 = -0.81 件/百万元，即有政策省份的补贴转化效率反而低于无政策省份，呈现出反常的「政策补贴挤出创新效率」现象。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从policy_release_status.csv中找到46条通信传输设备业相关地方政策，涉及18个省份。",
+    "从policy_resource.csv中深度分析46条政策全文，筛选出5个省份的政策明确包含面向通信传输设备业企业的直接研发创新奖补措施（湖北、陕西、广东、安徽、上海）。",
+    "从company_profile.csv中找到118家通信传输设备业内陆企业，有政策组62家（5省），无政策组56家（12省）。",
+    "从company_operation_status.csv中关联运营数据，筛选政府补贴>0且年度发明专利申请数非空的有效企业：有政策组56家，无政策组43家，合计99家。"
+  ],
+  "milestone": {
+    "通信传输设备业地方政策总条数（条）": 46,
+    "符合专项研发创新扶持条件的政策省份数（个）": 5,
+    "通信传输设备业内陆企业总数（家）": 118,
+    "有政策省份有效企业数（家）": 56,
+    "无政策省份有效企业数（家）": 43,
+    "有政策省份补贴转化效率均值（件/百万元）": 2.38,
+    "无政策省份补贴转化效率均值（件/百万元）": 3.18,
+    "差值（件/百万元）": -0.81
+  },
+  "answer": [
+    2.38,
+    3.18,
+    -0.81
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/hard015_result.json b/assets/qa_raw/hypothesis_verification/hard015_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..faaf3496921cd0dd8651b98aeb38c1951def43c2
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/hard015_result.json
@@ -0,0 +1,44 @@
+{
+  "id": "hard015",
+  "question": "政策文本的表述精度，即政策目标是以可量化的具体数值呈现，还是以方向性、原则性的定性语言为主，可能反映政府的政策执行意志，进而影响辖区内企业的研发行为。在仪器仪表制造业中，对已出台涉及本行业地方政策的省份进行内容分析：凡政策正文中含有具体产业发展数值目标（如产业规模达X亿元、增长X%、新建X家/X座等可核查数值）的省份，归为量化目标组；政策正文仅涵盖定性方向、原则性要求或门槛条件而不包含上述产业发展数值目标的省份，归为定性目标组。基于2022年数据，请分别计算两组省份内仪器仪表制造业有效企业（研发投入占比数据非空且数值在0%到100%之间）的研发投入占比均值，并给出量化目标组均值减去定性目标组均值的差值（单位：百分点）。",
+  "guidelines": "依次回答含量化目标省份的研发投入占比均值、仅含定性目标省份的研发投入占比均值、两者差值（量化组均值减去定性组均值）。均值和差值均保留2位小数，以百分点表示。如[12.53, 8.21, 4.32]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选涉及行业包含'仪器仪表制造业'且政策类型为地方政策的记录，共得到39条，涉及湖南省、山东省、四川省、上海市、福建省、江西省、重庆市、广东省、辽宁省、安徽省、天津市、黑龙江省、陕西省、广西壮族自治区、海南省、宁夏回族自治区、甘肃省、河北省、湖北省、新疆维吾尔自治区等省份。",
+    "筛选出与仪器仪表制造业企业分布重叠（即既有政策又有企业）的省份，共10个：湖南省（3家企业）、山东省（1家）、四川省（3家）、上海市（7家）、福建省（3家）、江西省（2家）、广东省（12家）、安徽省（4家）、河北省（2家）、湖北省（2家）。",
+    "从policy_resource.csv中读取上述10个省份共27条仪器仪表相关地方政策全文，对政策正文内容进行深度语义分析，区分量化目标与定性目标。27条政策对应id为：7、12、42、75、78、89、116、153、154、175、189、253、274、276、329、370、372、375、385、386、409、448、523、562、563、590、594。",
+    "经过对政策全文的逐条深度分析，含量化目标的省份共7个：上海市（id=75、385、386、448、562）、广东省（id=153、154、329、563、590）、安徽省（id=175、594）、湖北省（id=523）、江西省（id=89）、四川省（id=42、116、253、375）、湖南省（id=7、189、276、372）。",
+    "仅含定性目标的省份共3个：山东省（id=12、274、370）、福建省（id=78）、河北省（id=409）。",
+    "从company_profile.csv筛选industry='仪器仪表制造业'的企业，按省份分组，从company_operation_status.csv获取研发投入占比数据，筛选有效企业（研发投入占比非空且0%<占比<=100%）。含量化目标组7省共33家有效企业，研发投入占比之和=464.40，均值=464.40/33=14.07%；仅含定性目标组3省共6家有效企业，研发投入占比之和=58.01，均值=58.01/6=9.67%。",
+    "差值=量化目标组均值-定性目标组均值=14.07-9.67=4.40个百分点。"
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "从policy_release_status.csv中找到39条仪器仪表制造业相关地方政策，涉及10个有仪器仪表制造业企业的省份（27条与目标省份匹配）。",
+    "从policy_resource.csv中读取并分析27条政策全文，按量化目标与定性目标进行精细语义分类，得到7个量化目标省份和3个定性目标省份。",
+    "从company_profile.csv中找到仪器仪表制造业企业89家，分布于17个省份；目标10省共39家企业。",
+    "从company_operation_status.csv中获取39家企业的研发投入占比数据，筛选后量化组33家有效企业、定性组6家有效企业。"
+  ],
+  "milestone": {
+    "仪器仪表相关地方政策总数（条）": 39,
+    "有企业的目标省份数（个）": 10,
+    "含量化目标省份数（个）": 7,
+    "仅含定性目标省份数（个）": 3,
+    "量化目标组有效企业数（家）": 33,
+    "定性目标组有效企业数（家）": 6,
+    "量化目标组研发投入占比之和（%）": 464.4,
+    "定性目标组研发投入占比之和（%）": 58.01,
+    "量化目标组研发投入占比均值（%）": 14.07,
+    "定性目标组研发投入占比均值（%）": 9.67,
+    "差值（量化组-定性组，百分点）": 4.4
+  },
+  "answer": [
+    14.07,
+    9.67,
+    4.40
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/medium001_result.json b/assets/qa_raw/hypothesis_verification/medium001_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..318455b37e026f9bfbfe4a4ba518f7cfaf25a81e
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/medium001_result.json
@@ -0,0 +1,49 @@
+{
+  "id": "medium001",
+  "question": "Using 2022 data, to verify the hypothesis that policy has a positive effect on corporate R&D: among listed semiconductor firms, divide them by whether their registered province has ever issued a local industrial policy whose name or covered-industry field contains \"semiconductor\" or \"integrated circuit\" (either condition suffices). What is the difference in mean R&D investment ratio between the policy-covered group and the non-covered group (difference = policy-province group mean minus non-policy-province group mean), in percentage points?",
+  "guidelines": "Answer format: a numeric value (two decimal places, in percentage points). A positive value means the policy-province group is higher; a negative value means the non-policy-province group is higher. If relevant data cannot be found, answer \"No relevant data found\".",
+  "steps": [
+    "From policy_release_status.csv, filter policies with policyClassification=\"local policy\" and (policy name contains \"semiconductor\" or \"integrated circuit\", or industry field contains \"semiconductor\"), yielding 35 relevant policies. Extract unique province values (excluding \"nationwide\"), giving 15 policy provinces: Shanghai, Yunnan, Sichuan, Anhui, Shandong, Guangdong, Xinjiang Uygur Autonomous Region, Jiangxi, Henan, Zhejiang, Hainan, Hunan, Fujian, Chongqing, Shaanxi.",
+    "From company_profile.csv, filter firms whose industry is semiconductor, extract company name, bmCode, and province, yielding 172 semiconductor firms.",
+    "From company_operation_status.csv, join 2022 data by bmCode for these firms, extract R&D investment ratio; 172 firms successfully matched.",
+    "Exclude 3 firms with null R&D investment ratio and 0 firms with R&D ratio > 100% (outliers); 169 valid firms remain.",
+    "Split firms by whether their province is in the policy-province list: policy group (110 firms) and non-policy group (59 firms).",
+    "Policy group mean R&D investment ratio = Σ(firm R&D ratio) / firm count = 10.8880%; non-policy group mean = Σ(firm R&D ratio) / firm count = 11.6136%.",
+    "Difference = policy group mean − non-policy group mean = 10.8880 − 11.6136 = −0.73 percentage points."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "From policy_release_status.csv, 35 local policies related to semiconductor/integrated circuit were found, covering 15 provinces.",
+    "From company_profile.csv, 172 semiconductor firms were found.",
+    "From company_operation_status.csv, 2022 R&D investment ratio data were joined for 172 semiconductor firms; after filtering, 169 valid firms."
+  ],
+  "milestone": {
+    "Policy province list": [
+      "Shanghai",
+      "Yunnan",
+      "Sichuan",
+      "Anhui",
+      "Shandong",
+      "Guangdong",
+      "Xinjiang Uygur Autonomous Region",
+      "Jiangxi",
+      "Henan",
+      "Zhejiang",
+      "Hainan",
+      "Hunan",
+      "Fujian",
+      "Chongqing",
+      "Shaanxi"
+    ],
+    "Policy group firm count": 110,
+    "Non-policy group firm count": 59,
+    "Policy group mean R&D investment ratio (%)": 10.888,
+    "Non-policy group mean R&D investment ratio (%)": 11.6136
+  },
+  "answer": -0.73,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  }
+}
diff --git a/assets/qa_raw/hypothesis_verification/medium002_result.json b/assets/qa_raw/hypothesis_verification/medium002_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..95d545e866b9f418162a532b46ffe02f6f900e0d
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/medium002_result.json
@@ -0,0 +1,32 @@
+{
+  "id": "medium002",
+  "question": "In 2022, in the pharmaceutical manufacturing industry, what is the difference in Pearson correlation coefficient between private enterprises and state-owned enterprises(State-owned enterprises include central SOEs, local SOEs, SOEs under research institutes, and other SOEs.) with respect to R&D investment amount and cumulative number of granted Chinese invention patents? ",
+  "guidelines": "Answer format: the difference value (four decimal places). Difference = private enterprise correlation − state-owned enterprise correlation. A positive value means private enterprises have stronger correlation. If relevant data cannot be found, answer \"No relevant data found\".",
+  "steps": [
+    "From company_profile.csv, filter firms whose industry is pharmaceutical manufacturing, extract company name, bmCode, and ownership; 449 pharmaceutical manufacturing firms found.",
+    "Split firms by ownership into state-owned group (central SOEs, local SOEs, SOEs under research institutes, other SOEs), 67 firms, and private enterprise group, 346 firms.",
+    "From company_operation_status.csv, join 2022 data by bmCode for these firms, extract R&D investment amount and cumulative number of granted Chinese invention patents. State-owned group: 67 firms matched; private group: 346 firms matched.",
+    "State-owned group: exclude firms with null R&D amount or null cumulative invention patents; 67 valid firms remain. Pearson correlation between R&D amount and cumulative invention patents: r = 0.7170.",
+    "Private group: exclude firms with null R&D amount or null cumulative invention patents; 309 valid firms remain. Pearson correlation between R&D amount and cumulative invention patents: r = 0.5719.",
+    "Difference = private group correlation − state-owned group correlation = 0.5719 − 0.7170 = −0.1452.",
+    "Output the difference: −0.1452."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "From company_profile.csv, 449 pharmaceutical manufacturing firms were found: 67 state-owned and 346 private.",
+    "From company_operation_status.csv, 2022 R&D investment amount and cumulative granted Chinese invention patent data were joined for 67 state-owned and 346 private firms."
+  ],
+  "milestone": {
+    "State-owned valid count": 67,
+    "Private valid count": 309,
+    "State-owned Pearson correlation": 0.717,
+    "Private Pearson correlation": 0.5719,
+    "Difference": -0.1452
+  },
+  "answer": -0.1452,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  }
+}
diff --git a/assets/qa_raw/hypothesis_verification/medium003_result.json b/assets/qa_raw/hypothesis_verification/medium003_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..68b77c8f0ca9026d5fafc5eb83962c95f67d6036
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/medium003_result.json
@@ -0,0 +1,53 @@
+{
+  "id": "medium003",
+  "question": "To verify the hypothesis that policy density has a positive effect on average corporate profitability, using 2022 automotive manufacturing as an example, calculate the Spearman rank correlation coefficient between each province's policy density indicator (number of policies whose industry field contains \"automotive\") and average profitability (total operating profit amount / total operating revenue amount × 100%). Provinces without automotive manufacturing operating data or with zero operating revenue are excluded from the calculation; valid provinces with no policy records are assigned a policy count of 0.",
+  "guidelines": "Answer format: numeric value (rounded to 4 decimal places). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from policy_release_status.csv all policy records where the industry field contains \"automotive\", finding 69 related policies. After excluding national policies where province is null or \"National\", 53 remain. Group by province and count policy quantity per province, with 20 provinces having related policies.",
+    "Filter from regional_industry_status.csv all provincial records with industry=\"Automotive Manufacturing\", extract province, total operating profit amount, and total operating revenue amount fields, finding data for 34 provinces.",
+    "Filter out provinces where total operating revenue amount is null or zero, leaving 14 valid provinces. Calculate each province's average operating profit margin = total operating profit amount / total operating revenue amount × 100%, with profitability range from -4.4704% to 8.1755%.",
+    "Join policy count data with provincial profitability data by province; provinces without corresponding policies are assigned a policy count of 0. After joining, total valid provinces is 14, of which 11 have policies.",
+    "Calculate the Spearman rank correlation coefficient between policy count and average operating profit margin: rho = -0.0023, p-value = 0.993880.",
+    "Output the Spearman rank correlation coefficient value as -0.0023."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "Found 69 industrial policies related to \"automotive\" in policy_release_status.csv, covering 20 provinces.",
+    "Found operating profit and operating revenue data for 14 provinces in automotive manufacturing from regional_industry_status.csv."
+  ],
+  "milestone": {
+    "Total automotive-related policies": 69,
+    "Provinces with policies": 20,
+    "Total valid provinces": 14,
+    "Sample province data": {
+      "Guangdong Province": {
+        "Policy count": 10,
+        "Average operating profit margin (%)": 2.9496
+      },
+      "Shanghai": {
+        "Policy count": 7,
+        "Average operating profit margin (%)": 0.582
+      },
+      "Hunan Province": {
+        "Policy count": 5,
+        "Average operating profit margin (%)": 7.6412
+      },
+      "Sichuan Province": {
+        "Policy count": 4,
+        "Average operating profit margin (%)": 8.1755
+      },
+      "Shandong Province": {
+        "Policy count": 3,
+        "Average operating profit margin (%)": 3.402
+      }
+    },
+    "Spearman correlation coefficient": -0.0023,
+    "p-value": 0.99388
+  },
+  "answer": -0.0023,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  }
+}
diff --git a/assets/qa_raw/hypothesis_verification/medium004_result.json b/assets/qa_raw/hypothesis_verification/medium004_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b76db25d0f0f668aac8a8ca0d215fb13ce8003a2
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/medium004_result.json
@@ -0,0 +1,33 @@
+{
+  "id": "medium004",
+  "question": "In 2022, to verify the hypothesis on the effect of debt ratio on R&D investment ratio in Raw Chemical Materials and Chemical Products Manufacturing, calculate the difference in the mean R&D investment ratio (in percentage points) between the high-debt group (asset-liability ratio above the national industry median) and the low-debt group (asset-liability ratio below the national industry median).",
+  "guidelines": "Answer format: numeric value (rounded to 2 decimal places, unit: percentage points). A positive value indicates the high-debt group is higher. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from national_industry_status.csv for industry=\"Raw Chemical Materials and Chemical Products Manufacturing\", extract the asset-liability ratio median field. The national median asset-liability ratio for this industry is 36.815.",
+    "Filter from company_profile.csv all enterprises with industry=\"Raw Chemical Materials and Chemical Products Manufacturing\", extract enterprise name and bmCode fields, finding 364 enterprises.",
+    "Join with company_operation_status.csv by bmCode to obtain 2022 data for these enterprises, extract asset-liability ratio and R&D investment ratio fields, obtaining data for 364 enterprises.",
+    "Filter out enterprises where asset-liability ratio or R&D investment ratio is null; exclude 0 anomalous enterprises with R&D investment ratio above 100%, leaving 349 valid enterprises.",
+    "Group by comparing asset-liability ratio with the national median of 36.815: asset-liability ratio above median is the high-debt group (175 enterprises); below median is the low-debt group (174 enterprises); 0 enterprises equal to median are excluded from grouping.",
+    "Calculate high-debt group average R&D investment ratio = Σ(enterprise R&D investment ratio) / number of enterprises = 3.6595%; low-debt group average R&D investment ratio = Σ(enterprise R&D investment ratio) / number of enterprises = 4.2648%.",
+    "Calculate difference = high-debt group average R&D investment ratio - low-debt group average R&D investment ratio = 3.6595 - 4.2648 = -0.61 percentage points."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Found national industry aggregate data for Raw Chemical Materials and Chemical Products Manufacturing in national_industry_status.csv, with asset-liability ratio median of 36.815.",
+    "Found 364 enterprises in Raw Chemical Materials and Chemical Products Manufacturing in company_profile.csv.",
+    "Joined with company_operation_status.csv for 2022 asset-liability ratio and R&D investment ratio data of 364 enterprises; after excluding anomalies, 349 valid enterprises."
+  ],
+  "milestone": {
+    "National asset-liability ratio median": 36.815,
+    "High-debt group enterprises": 175,
+    "Low-debt group enterprises": 174,
+    "High-debt group average R&D investment ratio (%)": 3.6595,
+    "Low-debt group average R&D investment ratio (%)": 4.2648
+  },
+  "answer": -0.61,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  }
+}
diff --git a/assets/qa_raw/hypothesis_verification/medium005_result.json b/assets/qa_raw/hypothesis_verification/medium005_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..03851ddc77ff95423b1c731919e7df5b190057b9
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/medium005_result.json
@@ -0,0 +1,39 @@
+{
+  "id": "medium005",
+  "question": "In 2022, to verify the hypothesis that there is a clear relationship between enterprise total assets and invention patent count in the Consumer Electronics and Electrical Industry, we focus only on provinces with high R&D density (provinces where the mean R&D investment ratio in provincial industry aggregate data exceeds the corresponding national industry aggregate mean). Among all listed enterprises in this industry within these provinces, what is the Pearson correlation coefficient between enterprise total asset scale and annual China invention patent applications?",
+  "guidelines": "Answer format: numeric value (rounded to 4 decimal places). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from national_industry_status.csv for industry=\"Consumer Electronics and Electrical Industry\", extract the mean R&D investment ratio field. The national mean R&D investment ratio for Consumer Electronics and Electrical Industry is 7.77502994011976.",
+    "Filter from regional_industry_status.csv for industry=\"Consumer Electronics and Electrical Industry\", extract province and mean R&D investment ratio fields. 16 provinces have data.",
+    "Filter provinces where mean R&D investment ratio exceeds the national benchmark of 7.77502994011976, obtaining 4 R&D-intensive provinces: Guangdong Province, Beijing, Shanghai, Henan Province.",
+    "Filter from company_profile.csv enterprises with industry=\"Consumer Electronics and Electrical Industry\" and province in the R&D-intensive province list, finding 180 enterprises.",
+    "Join with company_operation_status.csv for 2022 data of these enterprises, extract total assets and annual China invention patent applications fields, obtaining data for 180 enterprises.",
+    "Filter out records where total assets or annual China invention patent applications is null, leaving 153 valid enterprises. Mean total assets is 24143218637.14 CNY, mean annual China invention patent applications is 197.82.",
+    "Calculate the Pearson correlation coefficient between total assets and annual China invention patent applications: r = 0.8294."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Retrieved national mean R&D investment ratio for Consumer Electronics and Electrical Industry of 7.77502994011976 from national_industry_status.csv.",
+    "Found mean R&D investment ratio data for 16 provinces in Consumer Electronics and Electrical Industry from regional_industry_status.csv; 4 provinces exceed the national benchmark.",
+    "Found 180 Consumer Electronics and Electrical Industry enterprises in R&D-intensive provinces from company_profile.csv.",
+    "Joined with company_operation_status.csv for total assets and annual China invention patent applications data of 153 valid enterprises."
+  ],
+  "milestone": {
+    "National mean R&D investment ratio": 7.77502994011976,
+    "R&D-intensive province list": [
+      "Guangdong Province",
+      "Beijing",
+      "Shanghai",
+      "Henan Province"
+    ],
+    "Valid enterprises": 153,
+    "Mean total assets (CNY)": 24143218637.14,
+    "Mean annual China invention patent applications": 197.82
+  },
+  "answer": 0.8294,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  }
+}
diff --git a/assets/qa_raw/hypothesis_verification/medium006_result.json b/assets/qa_raw/hypothesis_verification/medium006_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c5d3ab5fc6f1f003be1e1df18365689bf0d0d969
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/medium006_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium006",
+  "question": "In 2022, there was a view that government subsidies and revenue are clearly correlated among large-asset-scale enterprises in the food and beverage industry. Therefore, researchers sampled the top one-third of large enterprises ranked (rounded down) by total assets as the research subject. What is the Pearson correlation coefficient between government subsidy amount and year-over-year revenue growth rate?",
+  "guidelines": "Answer format: numerical value (retain 4 decimal places). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter companies with industry=\"food and beverage\" from company_profile.csv, extract company name and bmCode fields, and found 247 food and beverage companies.",
+    "From company_operation_status.csv, associate 2022 data for these companies based on bmCode, extract total assets, government incentive funds, subsidies, and year-over-year revenue growth rate fields, and associate data for 247 companies.",
+    "Filter out companies with empty total assets, leaving 247 companies, sorted in descending order by total assets.",
+    "Calculate total number of companies 247, take the top 82 companies (rounded down) as the large enterprise group.",
+    "In the large enterprise group, filter out companies with empty government incentive funds, subsidies, or year-over-year revenue growth rate, leaving 80 valid enterprises.",
+    "Calculate the Pearson correlation coefficient between government incentive funds and subsidies and year-over-year revenue growth rate = -0.1237."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "Found 247 food and beverage companies from company_profile.csv, including company name and bmCode fields.",
+    "Associated 2022 data on total assets, government incentive funds, subsidies, and year-over-year revenue growth rate for 247 food and beverage companies from company_operation_status.csv."
+  ],
+  "milestone": {
+    "Total number of food and beverage enterprises": 247,
+    "Number of large enterprise group": 82,
+    "Number of valid large enterprises": 80,
+    "Average government subsidy": 137397163.62,
+    "Average revenue growth rate": 14.8526
+  },
+  "answer": -0.1237,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/medium007_result.json b/assets/qa_raw/hypothesis_verification/medium007_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f316d791be7d425bb462817d045adf88262e9d00
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/medium007_result.json
@@ -0,0 +1,56 @@
+{
+  "id": "medium007",
+  "question": "In the 2022 data, to verify the hypothesis that private enterprises in the communications transmission equipment industry benefit more from policies than state-owned enterprises, we focus on communications transmission equipment enterprises in provinces that have issued local policies involving the \"communications\"-related industry. Enterprises are grouped by ownership type (private enterprises vs. state-owned enterprises, where state-owned enterprises include only central state-owned enterprises and local state-owned enterprises). Within each group, the per capita revenue (total revenue / total number of employees) is calculated using the weighted consolidation method. Finally, return the specific difference between per capita revenue of the private enterprise group and that of the state-owned enterprise group. ",
+  "guidelines": "Answer format: numerical value (retain 2 decimal places, unit: yuan/person). A positive value indicates private enterprises are higher. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter policy records from policy_release_status.csv where policyClassification=\"local policy\" and the industry field contains \"communications\", finding 48 relevant policies. Extract the deduplicated province field (excluding \"national\"), obtaining 18 provinces with policies: Shanghai, Yunnan, Beijing, Sichuan, Anhui, Shandong, Guangdong, Xinjiang Uygur Autonomous Region, Jiangxi, Henan, Hainan, Hubei, Hunan, Fujian, Guizhou, Liaoning, Chongqing, Shaanxi.",
+    "Filter from company_profile.csv companies whose industry contains \"communications transmission equipment\" and province is among the policy provinces. Extract company name, bmCode, and ownership fields, finding 88 companies.",
+    "Divide enterprises into the private enterprise group (ownership=\"private enterprise\", 61 companies) and the state-owned enterprise group (ownership is \"central state-owned enterprise\" or \"local state-owned enterprise\", 21 companies).",
+    "From company_operation_status.csv, associate 2022 data based on bmCode, extract revenue amount and total employee fields. Private enterprises: 61 associated; state-owned enterprises: 21 associated.",
+    "Filter out records with empty revenue amount or total employees, or zero total employees. Valid private enterprises: 61; valid state-owned enterprises: 20.",
+    "Calculate per capita revenue of private enterprise group = total revenue of private enterprises / total employees of private enterprises = 1009833229861.02 / 471176 = 2143218.73 yuan/person; calculate per capita revenue of state-owned enterprise group = total revenue of state-owned enterprises / total employees of state-owned enterprises = 227578562511.11 / 119917 = 1897800.67 yuan/person.",
+    "Calculate difference = per capita revenue of private enterprises - per capita revenue of state-owned enterprises = 2143218.73 - 1897800.67 = 245418.07 yuan/person."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Filtered 48 local policies involving industries containing \"communications\" from policy_release_status.csv, covering 18 provinces.",
+    "Found 88 communications transmission equipment companies in policy provinces from company_profile.csv, including 61 private enterprises and 21 state-owned enterprises.",
+    "Associated 2022 revenue and employee data from company_operation_status.csv. After filtering, 61 valid private enterprises and 20 valid state-owned enterprises."
+  ],
+  "milestone": {
+    "List of provinces with policies": [
+      "Shanghai",
+      "Yunnan",
+      "Beijing",
+      "Sichuan",
+      "Anhui",
+      "Shandong",
+      "Guangdong",
+      "Xinjiang Uygur Autonomous Region",
+      "Jiangxi",
+      "Henan",
+      "Hainan",
+      "Hubei",
+      "Hunan",
+      "Fujian",
+      "Guizhou",
+      "Liaoning",
+      "Chongqing",
+      "Shaanxi"
+    ],
+    "Number of private enterprises": 61,
+    "Number of state-owned enterprises": 20,
+    "Total revenue of private enterprises": 1009833229861.02,
+    "Total employees of private enterprises": 471176,
+    "Total revenue of state-owned enterprises": 227578562511.11,
+    "Total employees of state-owned enterprises": 119917,
+    "Per capita revenue of private enterprises": 2143218.73,
+    "Per capita revenue of state-owned enterprises": 1897800.67
+  },
+  "answer": 245418.07,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  }
+}
diff --git a/assets/qa_raw/hypothesis_verification/medium008_result.json b/assets/qa_raw/hypothesis_verification/medium008_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c7a2a5c57e841a167dcd527b7b6bee3c66d0a544
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/medium008_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium008",
+  "question": "Using the specialized equipment manufacturing industry in 2022 as the research subject, some researchers believe that enterprise size may confound the relationship between listing history length and cumulative patent accumulation. To test this hypothesis: First, among all valid enterprises (with non-null cumulative China patent applications), calculate the Pearson correlation coefficient r1 between listing years (derived by subtracting listing year from 2022) and cumulative China patent applications; second, restrict the sample to the large enterprise subset whose total assets exceed the industry median total assets in the national industry aggregate data, then calculate the Pearson correlation coefficient r2 for the same pair of variables; finally, report the specific value of the difference r2 − r1.",
+  "guidelines": "Answer format: the difference between the two correlation coefficients (retain 4 decimal places). Difference = large enterprise correlation coefficient − all enterprises correlation coefficient. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter industry=\"specialized equipment manufacturing\" from national_industry_status.csv, extract the median total assets field. National median total assets for specialized equipment manufacturing: 2619344581.0.",
+    "Filter all enterprises with industry=\"specialized equipment manufacturing\" from company_profile.csv, extract company name, bmCode, and listingDate fields, finding 447 companies.",
+    "Filter out 0 companies with empty listingDate, 447 companies remain. Calculate listing years = 2022 - listing year based on listingDate, listing years range from 0 to 34 years.",
+    "From company_operation_status.csv, associate 2022 data for these companies based on bmCode, extract total assets and cumulative China patent applications fields, successfully associating 447 companies.",
+    "Filter out 20 companies with empty cumulative China patent applications, 427 valid enterprises remain for the full sample. Calculate Pearson correlation coefficient r1 between listing years and cumulative China patent applications = 0.3377.",
+    "Further filter large enterprises with total assets exceeding the national industry median (2619344581.0), 217 companies in total.",
+    "Calculate Pearson correlation coefficient r2 between listing years and cumulative China patent applications for the large enterprise group = 0.3471.",
+    "Calculate difference = r2 - r1 = 0.3471 - 0.3377 = 0.0093."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "Obtained national median total assets for specialized equipment manufacturing as 2619344581.0 from national_industry_status.csv.",
+    "Found 447 specialized equipment manufacturing companies from company_profile.csv, 447 with listing dates after filtering.",
+    "Associated 2022 total assets and cumulative China patent applications data for 447 companies from company_operation_status.csv. After filtering: 427 valid enterprises for the full sample, 217 large enterprises."
+  ],
+  "milestone": {
+    "National median total assets": 2619344581.0,
+    "Number of all valid enterprises": 427,
+    "r1 value": 0.3377,
+    "Number of large enterprises": 217,
+    "r2 value": 0.3471
+  },
+  "answer": 0.0093,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  }
+}
diff --git a/assets/qa_raw/hypothesis_verification/medium009_result.json b/assets/qa_raw/hypothesis_verification/medium009_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..87f88ef60795cd2aa6eb214e1f663326934f8fd0
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/medium009_result.json
@@ -0,0 +1,30 @@
+{
+  "id": "medium009",
+  "question": "In the pharmaceutical manufacturing industry in 2022, verify whether the 'policy-innovation paradox' exists (i.e., the phenomenon where the strength of local pharmaceutical innovation support policies is negatively associated with innovation output). Among provinces that have issued pharmaceutical support policies, use the median number of policy entries as the threshold for support strength, and use the national average invention patent grants per province for pharmaceutical manufacturing as the innovation output benchmark. Count the number of provinces where policy support strength exceeds the median but total invention patent grants are below the national average.",
+  "guidelines": "Answer format: numerical value (integer). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from policy_release_status.csv: policyClassification=\"local policy\", publishDate in 2022, industry contains \"pharmaceutical manufacturing\", province is non-empty and not \"national\" (全国). Aggregate policy entry counts by province to obtain 21 provinces that have issued pharmaceutical support policies.",
+    "Take the median of policy entry counts across those 21 provinces, yielding 2.0; define \"support strength above the median\" as a strictly greater policy count than 2.",
+    "From regional_industry_status.csv, filter rows where industry=\"pharmaceutical manufacturing\". Treat empty values in 「年度中国发明专利授权数合计」 (annual total China invention patent grants) as undisclosed: compute the national average invention patent grants per province for pharmaceutical manufacturing using only provincial records with non-empty values, yielding 15 provinces with disclosed data, total 3243, provincial average = 3243 / 15 = 216.20.",
+    "Link policy provinces with patent data: a province participates in the paradox judgment only if it has a disclosed annual total for invention patent grants; provinces without disclosed patent data are excluded from the comparison.",
+    "Filter provinces that simultaneously satisfy: policy entry count > 2.0 and disclosed patent total < 216.20, yielding: Shanghai (policy count = 11, patents = 210), Henan (policy count = 3, patents = 102), Sichuan (policy count = 4, patents = 112).",
+    "The final number of qualifying provinces is 3."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "policy_release_status.csv: provincial records for 2022 local policies with industry containing \"pharmaceutical manufacturing\", covering 21 provinces (exact entry counts per file).",
+    "regional_industry_status.csv: for pharmaceutical manufacturing, 15 provincial records with non-empty 「年度中国发明专利授权数合计」 are used to compute the provincial average; empty values are excluded from the average and from paradox judgment."
+  ],
+  "milestone": {
+    "Number of provinces with valid disclosure": 15,
+    "Median policy entry count": 2.0,
+    "National average invention patent grants per province (pharmaceutical manufacturing)": 216.2,
+    "Number of paradox provinces": 3
+  },
+  "answer": 3,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/medium010_result.json b/assets/qa_raw/hypothesis_verification/medium010_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b3c747b7074efe5888199fd531ae402bfc062c1d
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/medium010_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium010",
+  "question": "In 2022, based on the 'high input low output' hypothesis in Romer's endogenous growth theory, verify whether state-owned enterprises in the semiconductor industry exhibit the dual anomaly of 'large asset scale but low operating profit margin' and 'high R&D investment but low patent conversion efficiency'. The hypothesis states: when state-owned enterprises rank high in the industry in both asset scale (total assets) and R&D investment (amount), their operating profit margin (operating profit / revenue) and patent conversion efficiency (cumulative China invention patent grants / R&D investment × 100 million) should rank low in the industry. Please count the number of enterprises that simultaneously satisfy the following conditions (state-owned enterprises include central state-owned enterprises, local state-owned enterprises, state-owned enterprises (research institutes), and state-owned enterprises (other)): ① total assets > median total assets of industry-wide state-owned enterprises; ② operating profit margin < median operating profit margin of industry-wide state-owned enterprises; ③ R&D investment amount > median R&D investment amount of industry-wide state-owned enterprises; ④ patent conversion efficiency < median patent conversion efficiency of industry-wide state-owned enterprises.",
+  "guidelines": "Answer format: numerical value (integer, unit: enterprises). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from company_profile.csv enterprises with industry=\"semiconductor\" and ownership as state-owned enterprise types (central state-owned enterprise, local state-owned enterprise, state-owned enterprise (research institute), state-owned enterprise (other)), finding 32 state-owned enterprises.",
+    "From company_operation_status.csv, associate 2022 data for these enterprises, extract total assets, operating profit amount, revenue amount, R&D investment amount, and cumulative China invention patent grants fields. Data associated for 32 enterprises.",
+    "Filter out enterprises with empty or zero revenue amount, 32 valid enterprises remain. Calculate operating profit margin = operating profit amount / revenue amount × 100%.",
+    "Filter out enterprises with empty or zero R&D investment amount, 32 valid enterprises remain. Calculate patent conversion efficiency = cumulative China invention patent grants / R&D investment amount × 100 million.",
+    "Calculate median for each indicator: median total assets = 5937801202.30 yuan, median operating profit margin = 11.9667%, median R&D investment amount = 245125759.25 yuan, median patent conversion efficiency = 50.7635 grants per 100 million yuan.",
+    "Filter enterprises satisfying all four conditions: total assets > median (16), operating profit margin < median (16), R&D investment amount > median (16), patent conversion efficiency < median (15). Enterprises satisfying all four conditions: 3.",
+    "Count dual anomaly enterprises: 3."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Found 32 semiconductor industry state-owned enterprises from company_profile.csv.",
+    "Associated 2022 operation data for 32 enterprises from company_operation_status.csv, with 32 enterprises having valid data for both operating profit margin and patent conversion efficiency."
+  ],
+  "milestone": {
+    "Total number of state-owned enterprises": 32,
+    "Number of valid enterprises": 32,
+    "Median total assets (yuan)": 5937801202.3,
+    "Median operating profit margin (%)": 11.9667,
+    "Median R&D investment amount (yuan)": 245125759.25,
+    "Median patent conversion efficiency (grants per 100 million yuan)": 50.7635,
+    "Number of enterprises satisfying conditions": 3
+  },
+  "answer": 3,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  }
+}
diff --git a/assets/qa_raw/hypothesis_verification/medium011_result.json b/assets/qa_raw/hypothesis_verification/medium011_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..42058433649e152a33c23be607604ae9bf382be9
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/medium011_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium011",
+  "question": "In the 2022 provincial-level data for the automotive manufacturing industry, some provinces exhibit a dual structural contradiction: first, the number of automotive manufacturing enterprises in the province exceeds the average number of enterprises across all provinces with automotive manufacturing, but the province's total automotive manufacturing revenue is below the average revenue of all provinces; second, the province's average R&D investment ratio is higher than the mean of this indicator across provinces, but the average profit margin (measured as total operating profit / total revenue × 100%) is lower than the mean of this profit margin indicator across all provinces. Please count the number of provinces in 2022 that simultaneously meet both of the above contradiction conditions.",
+  "guidelines": "Answer format: numerical value (integer). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from regional_industry_status.csv all provincial records with industry=\"automotive manufacturing\" and total number of enterprises > 0. Extract province, total number of enterprises, total revenue amount, total operating profit amount, and average R&D investment ratio fields. Found 14 valid provinces.",
+    "Calculate the mean total number of enterprises across all valid provinces = 14.0000, and mean total revenue amount = 297307507961.18 yuan.",
+    "Calculate each province's average profit margin = total operating profit amount / total revenue amount × 100%, and calculate the mean of average profit margin across all valid provinces = 3.2804%.",
+    "Calculate the mean of average R&D investment ratio across all valid provinces = 6.1612.",
+    "Filter provinces satisfying all four conditions: total number of enterprises > 14.0000, total revenue amount < 297307507961.18, average R&D investment ratio > 6.1612, and average profit margin < 3.2804%. Provinces satisfying conditions: none.",
+    "Count of provinces satisfying conditions: 0."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "Filtered 14 provinces with automotive manufacturing enterprises from regional_industry_status.csv, including total number of enterprises, total revenue amount, total operating profit amount, and average R&D investment ratio fields."
+  ],
+  "milestone": {
+    "Total number of valid provinces": 14,
+    "Mean total number of enterprises": 14.0,
+    "Mean revenue (yuan)": 297307507961.18,
+    "Mean profit margin (%)": 3.2804,
+    "Mean R&D ratio": 6.1612,
+    "List of provinces satisfying conditions": "none"
+  },
+  "answer": 0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  }
+}
diff --git a/assets/qa_raw/hypothesis_verification/medium012_result.json b/assets/qa_raw/hypothesis_verification/medium012_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0ac5e1f1b88f1b7ca63969c7dc06d884b4cf9b4f
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/medium012_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "medium012",
+  "question": "In 2022, based on the 'structural efficiency paradox' hypothesis in industrial economics (i.e., during enterprise technology transformation, the anomalous phenomenon may occur where revenue growth coexists with workforce reduction, and reduced R&D investment coexists with increased innovation output), verify whether the dual paradox exists in the chemical fiber manufacturing industry. The hypothesis states: when enterprises are in the technology upgrading stage, revenue grows but automation replaces labor leading to fewer employees; meanwhile, R&D investment ratio is below the industry median but patent grants are above the industry median. Please count the number of valid enterprises that simultaneously exhibit both paradox characteristics.",
+  "guidelines": "Answer format: numerical value (integer, unit: enterprises). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all enterprises with industry=\"chemical fiber manufacturing\" from company_profile.csv, extract company name and bmCode fields, finding 34 enterprises.",
+    "From company_operation_status.csv, associate 2022 data for these enterprises based on bmCode, extract four indicators: year-over-year revenue growth rate, year-over-year employee growth rate, R&D investment ratio, and annual China patent grants. Successfully associated 34 enterprises. Filter out 4 enterprises with any of the four indicators empty, obtaining 30 valid enterprises.",
+    "Calculate median R&D investment ratio of valid enterprises = 3.355, median annual China patent grants = 21.0.",
+    "Filter first paradox: enterprises with year-over-year revenue growth rate > 0 and year-over-year employee growth rate < 0, 3 enterprises in total.",
+    "Among first paradox enterprises, filter second paradox: enterprises with R&D investment ratio < industry median (3.355) and annual China patent grants > industry median (21.0), 2 enterprises in total.",
+    "Number of enterprises simultaneously satisfying both paradox conditions: 2."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "Found 34 chemical fiber manufacturing enterprises from company_profile.csv.",
+    "Associated 2022 data on year-over-year revenue growth rate, year-over-year employee growth rate, R&D investment ratio, and annual China patent grants for 34 enterprises from company_operation_status.csv. After filtering: 30 valid enterprises."
+  ],
+  "milestone": {
+    "Total number of valid enterprises": 30,
+    "Median R&D investment ratio": 3.355,
+    "Median annual patent grants": 21.0,
+    "Number of first paradox enterprises": 3,
+    "Number of dual paradox enterprises": 2
+  },
+  "answer": 2,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/medium013_result.json b/assets/qa_raw/hypothesis_verification/medium013_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..20406cb4c03e94cd3132be32038059de700e5557
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/medium013_result.json
@@ -0,0 +1,33 @@
+{
+  "id": "medium013",
+  "question": "In 2022, in the information transmission, software and information technology services industry, researchers aim to characterize a type of dual anomaly enterprises with high valuation and high R&D investment, yet weak profitability and insufficient patent influence. The specific criteria are: valid enterprises must have complete data for all four indicators—market cap, revenue (non-zero), R&D investment ratio, and cumulative total patent citations; net profit margin is calculated as net profit amount divided by revenue amount times 100%; among all valid enterprises, using each indicator's median as the threshold, filter enterprises where market cap exceeds the median and net profit margin is below the median, while R&D investment ratio exceeds the median and cumulative patent citations are below the median. What is the proportion of such enterprises among all valid enterprises?",
+  "guidelines": "Answer format: proportional value (retain 2 decimal places, unit: %). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter all enterprises with industry=\"information transmission, software and information technology services\" from company_profile.csv, finding 644 enterprises.",
+    "From company_operation_status.csv, associate 2022 data for these enterprises based on bmCode, extract market cap, net profit amount, revenue amount, R&D investment ratio, and cumulative total patent citations fields. Successfully associated 644 enterprises.",
+    "Filter out 100 enterprises with any of market cap, revenue amount, R&D investment ratio, or cumulative total patent citations empty, or with zero revenue. 544 valid enterprises remain.",
+    "Calculate net profit margin = net profit amount / revenue amount × 100%. Net profit margin range: -1857.0154% to 80.7125%.",
+    "Calculate median for each of the four indicators: median market cap = 59.5, median net profit margin = 3.2905%, median R&D investment ratio = 12.155000000000001, median cumulative total patent citations = 363.0.",
+    "Filter dual anomaly enterprises satisfying all four conditions (market cap > median, net profit margin < median, R&D investment ratio > median, cumulative patent citations < median), 19 enterprises in total.",
+    "Calculate proportion = 19 / 544 × 100% = 3.49%."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Found 644 information transmission, software and information technology services enterprises from company_profile.csv.",
+    "Associated 2022 data on market cap, net profit amount, revenue amount, R&D investment ratio, and cumulative total patent citations for 644 enterprises from company_operation_status.csv. After filtering: 544 valid enterprises."
+  ],
+  "milestone": {
+    "Total number of valid enterprises": 544,
+    "Median market cap": 59.5,
+    "Median net profit margin (%)": 3.2905,
+    "Median R&D investment ratio": 12.155,
+    "Median cumulative patent citations": 363.0,
+    "Number of anomaly enterprises": 19
+  },
+  "answer": 3.49,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/hypothesis_verification/medium014_result.json b/assets/qa_raw/hypothesis_verification/medium014_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d6dd9c7aacf4c31bcbe25b304f26ed7c357d8a3
--- /dev/null
+++ b/assets/qa_raw/hypothesis_verification/medium014_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium014",
+  "question": "In 2022, based on the 'leverage-growth coupling effect' hypothesis in the context of industrial upgrading (i.e., there is a non-linear relationship between corporate financial leverage and operating growth, potentially showing a reverse coupling pattern of high leverage with high growth or low leverage with negative growth), in provinces covered by local policies for the rubber and plastic products industry, verify whether listed enterprises in this industry exhibit structural leverage-growth association. Specifically: among valid enterprises, using the median asset-liability ratio as the financial leverage threshold and the sign of revenue growth rate as the growth direction indicator, count the number of coupling enterprises satisfying 'high leverage (asset-liability ratio > median) and high growth (growth rate > 0)' (A) and coupling enterprises satisfying 'low leverage (asset-liability ratio < median) and negative growth' (B), and calculate the difference A − B (integer). What is the specific value of this difference?",
+  "guidelines": "Answer format: numerical value (integer). A positive value indicates that 'high leverage high growth' coupling is more prominent; a negative value indicates that 'low leverage negative growth' coupling is more prominent. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from policy_release_status.csv policy records where policyClassification=\"local policy\" and industry contains \"rubber\" or \"plastic\", finding 17 relevant policies. Extract deduplicated province field (excluding \"national\"), obtaining 12 policy provinces: Shanghai, Yunnan, Sichuan, Anhui, Shandong, Guangxi Zhuang Autonomous Region, Xinjiang Uygur Autonomous Region, Hebei, Hainan, Hunan, Liaoning, Shaanxi.",
+    "Filter from company_profile.csv enterprises with industry=\"rubber and plastic products\" and province in policy provinces. Extract company name, bmCode, and province fields, finding 30 enterprises.",
+    "From company_operation_status.csv, associate 2022 data for these enterprises based on bmCode, extract asset-liability ratio and year-over-year revenue growth rate fields. Successfully associated 30 enterprises.",
+    "Filter out 0 enterprises with empty asset-liability ratio or year-over-year revenue growth rate. 30 valid enterprises remain (year-over-year revenue growth rate > 0 indicates high growth, < 0 indicates negative growth).",
+    "Calculate median asset-liability ratio of valid enterprises = 36.875.",
+    "Count 'high leverage high growth' coupling enterprises: asset-liability ratio > 36.875 and year-over-year revenue growth rate > 0, 13 enterprises in total.",
+    "Count 'low leverage negative growth' coupling enterprises: asset-liability ratio < 36.875 and year-over-year revenue growth rate < 0, 9 enterprises in total (those equal to median are not counted in either group).",
+    "Calculate coupling effect strength difference = 13 - 9 = 4."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "Filtered 17 rubber/plastic related local policies from policy_release_status.csv, involving 12 provinces.",
+    "Found 30 rubber and plastic products enterprises in policy provinces from company_profile.csv.",
+    "Associated 2022 asset-liability ratio and year-over-year revenue growth rate data for 30 enterprises from company_operation_status.csv. After filtering: 30 valid enterprises."
+  ],
+  "milestone": {
+    "Number of policy provinces": 12,
+    "Total number of valid enterprises": 30,
+    "Median asset-liability ratio": 36.875,
+    "Number of high leverage high growth coupling enterprises": 13,
+    "Number of low leverage negative growth coupling enterprises": 9
+  },
+  "answer": 4,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "hypothesis_verification"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/hard001_result.json b/assets/qa_raw/industry_planning/hard001_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce63ad00a02452d6ffffdfe39d87ecad7d8b3429
--- /dev/null
+++ b/assets/qa_raw/industry_planning/hard001_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "hard001",
+  "question": "基于2022年的数据，现对中国大陆半导体行业各省（相关企业数量>=5）进行政策分化情景推演：凡已落地集成电路产业发展促进类专项政策的省份，其辖内半导体企业可维持现有研发扩张节奏（以省内各企业研发投入同比增减幅的中位数为准）；而尚未出台上述专项政策的省份，受政策缺位影响，预计其研发增速将压缩至原有水平的一半。在此分化情景下，以3年复合增长方式推算至2025年，请问届时研发投入规模居于各省首位的是哪个省份？该省预测研发投入总额约为多少亿元？",
+  "guidelines": "依次回答省份名称和预计研发投入总额。金额以亿元为单位，保留2位小数。如[\"浙江省\", 312.75]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_resource.csv筛选政策使用行业包含\"集成电路\"或\"半导体\"的政策，找到44条相关政策，均为地方政策，涉及广东省（10条）、上海市（5条）、安徽省（4条）、四川省（3条）、山东省（2条）、重庆市（2条）、福建省（1条）、海南省（1条）、河南省（1条）、湖南省（1条）、江西省（1条）、陕西省（1条）、新疆维吾尔自治区（1条）、云南省（1条）、浙江省（1条）。",
+    "从policy_resource.csv中读取这44条政策的全文内容，逐一分析其是否包含集成电路产业发展促进的实质性内容。经分析确认，出台了集成电路产业发展促进专项政策：广东省（id=80、153、249、605）、上海市（id=201、398、461）、安徽省（id=4、301）、四川省（id=387）、山东省（id=284）、浙江省（id=125）。",
+    "从company_profile.csv筛选行业=\"半导体业\"的企业，共172家。排除港澳台地区后剩余160家，按省份统计企业数，筛选企业数>=5的省份共6个：广东省54家、上海市27家、江苏省23家、浙江省13家、北京市10家、湖北省5家。",
+    "从company_operation_status.csv提取上述6个省份半导体业企业的研发投入金额和研发投入同比增减幅，计算各省研发投入合计和增速中位数：上海市220.03亿元/32.19%、广东省140.27亿元/8.87%、江苏省56.12亿元/21.06%、北京市46.75亿元/20.13%、浙江省22.14亿元/18.33%、湖北省19.30亿元/39.54%。",
+    "根据假设条件，有专项政策的省份（上海市、广东省、浙江省）按原增速中位数增长，无专项政策的省份（江苏省、北京市、湖北省）增速减半。安徽省虽有政策但企业数不足5家，不参与计算。",
+    "以3年复合增长方式计算各省2025年研发投入预测：上海市=220.03×(1+32.19%)^3=508.26亿元、广东省=140.27×(1+8.87%)^3=180.98亿元、江苏省=56.12×(1+10.53%)^3=75.78亿元、北京市=46.75×(1+10.07%)^3=62.34亿元、浙江省=22.14×(1+18.33%)^3=36.69亿元、湖北省=19.30×(1+19.77%)^3=33.16亿元。",
+    "上海市以508.26亿元排名第一，预计2025年研发投入总额最高。"
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "从policy_release_status.csv中找到4条集成电路/半导体相关的地方政策，涉及4个省份。",
+    "从policy_resource.csv中分析4条政策全文，确认4个省份均出台了集成电路产业发展促进专项政策。",
+    "从company_profile.csv中筛选出172家半导体业企业，排除港澳台后160家，其中6个省份企业数>=5家，共132家。",
+    "从company_operation_status.csv中获取132家企业的研发投入金额和研发投入同比增减幅数据。"
+  ],
+  "milestone": {
+    "集成电路/半导体相关政策总数（条）": 4,
+    "经政策全文分析确认的省份数（个）": 4,
+    "半导体业企业总数（家）": 172,
+    "大陆半导体业企业数（家）": 160,
+    "企业数>=5的省份数（个）": 6,
+    "上海市半导体企业数（家）": 27,
+    "上海市2022研发投入合计（亿元）": 220.03,
+    "上海市研发增速中位数（%）": 32.19,
+    "上海市2025预测研发投入（亿元）": 508.26
+  },
+  "answer": [
+    "上海市",
+    508.26
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/hard002_result.json b/assets/qa_raw/industry_planning/hard002_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e5660b26169815c2058900a121bde628c2214875
--- /dev/null
+++ b/assets/qa_raw/industry_planning/hard002_result.json
@@ -0,0 +1,44 @@
+{
+  "id": "hard002",
+  "question": "以2022年化学原料和化学制品制造业的实际数据为起点，构建如下政策差异化情景：若某省（仅纳入中国大陆化学原料和化学制品制造业存续企业数量不低于8家的省份，港澳台不在统计范围内）已发布新材料领域的产业相关发展政策且涉及具体目标和量化指标，则该省化工企业得以按照各自当前研发增速（取省内企业研发投入同比增减幅的中位数）持续推进研发扩张；反之，凡无此类专项政策的省份，其研发增速将以现有水平的一半计算。在3年复合增长模型下展望至2025年，试问：哪个省份的研发投入省际排名跃升幅度最为显著？该省届时预估的研发投入总规模是多少亿元？",
+  "guidelines": "依次回答省份名称和预计研发投入总额。研发投入总额以亿元为单位，保留2位小数。如[\"湖北省\", 72.31]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_resource.csv筛选industry字段包含\"化学原料和化学制品制造业\"的政策记录，找到60条化工相关政策，其中地方政策48条。",
+    "从policy_resource.csv中读取这48条地方政策的全文内容，分析哪些政策属于新材料产业专项发展政策（包含具体的新材料产业发展目标和量化指标），筛选出3个省份出台了此类专项政策：湖南省（id=491）、安徽省（id=177）、广东省（id=605）、江西省（id=380）、四川省（id=526）、贵州省（id=276）、河南省（id=87）、云南省（id=128）。",
+    "从company_profile.csv筛选行业=\"化学原料和化学制品制造业\"的企业，排除港澳台地区后共357家，按省份统计企业数量，筛选企业数>=8的省份，得到13个符合条件的省份。",
+    "从company_operation_status.csv提取这13个省份化工企业的研发投入金额和研发投入同比增减幅，计算各省2022年研发投入总额和增速中位数。广东省36.66亿元/9.28%、江苏省55.06亿元/9.44%、上海市37.78亿元/-1.19%、浙江省74.99亿元/13.14%、山东省130.14亿元/24.79%、四川省30.06亿元/18.20%、安徽省18.53亿元/12.78%、湖南省34.64亿元/34.64%、河南省18.32亿元/-8.38%、河北省11.07亿元/52.95%、辽宁省96.99亿元/15.98%、湖北省没有相关数据。",
+    "按照假设条件设定各省有效增速：有新材料专项政策的省份（湖南省、广东省、四川省、河南省）维持当前增速中位数，无政策省份增速减半。广东省9.28%、江苏省4.72%、上海市-0.595%、浙江省6.57%、山东省12.395%、四川省9.1%、安徽省6.3875%、湖南省34.64%、河南省-8.38%、河北省26.4725%、辽宁省7.99%",
+    "使用3年复合增长公式计算2025年研发投入总额：2025年研发投入 = 2022年研发投入 × (1 + 有效增速/100)^3。湖南省：10.29 × (1+34.64/100)^3 = 25.13亿元（排名从第10升至第7）。",
+    "比较各省排名变化：湖南省上升3位（第10→第7），上海市下降3位（第4→第7），河北省上升2位（第10→第8）。排名上升幅度最大的省份为湖南省，上升3位，2025年预计研发投入总额25.13元。"
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "从policy_release_status.csv中找到60条涉及化学原料和化学制品制造业的政策，其中地方政策48条。",
+    "从policy_resource.csv中分析48条地方政策全文，筛选出4个省份（湖南省、广东省、四川省、河南省）出台了新材料产业专项发展政策，包含具体的产业发展目标和量化指标。",
+    "从company_profile.csv中筛选出中国大陆357家化学原料和化学制品制造业企业，分布在13个企业数>=8的省份。",
+    "从company_operation_status.csv中获取这些企业的研发投入金额和研发投入同比增减幅数据，用于计算各省研发投入总额和增速中位数。"
+  ],
+  "milestone": {
+    "化工相关政策总数（条）": 60,
+    "化工地方政策数（条）": 48,
+    "新材料专项政策省份数（个）": 3,
+    "中国大陆化工企业总数（家）": 357,
+    "符合条件省份数（企业>=8）": 13,
+    "湖南省化工企业数（家）": 19,
+    "湖南省2022年研发投入总额（亿元）": 18.32,
+    "湖南省研发增速中位数（%）": 24.64,
+    "湖南省2025年预计研发投入（亿元）": 25.13,
+    "四川省2022年排名": 10,
+    "四川省2025年排名": 7,
+    "四川省排名上升（位）": 3
+  },
+  "answer": [
+    "湖南省",
+    25.13
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/hard003_result.json b/assets/qa_raw/industry_planning/hard003_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..84c69b1011c2dd6cfe926332b6679f9c8634ed1b
--- /dev/null
+++ b/assets/qa_raw/industry_planning/hard003_result.json
@@ -0,0 +1,43 @@
+{
+  "id": "hard003",
+  "question": "针对中国大陆医药制造业，以2022年各省数据为基准，现设定如下政策效应假设：已颁布生物医药产业发展促进相关政策的省份（仅计入中国大陆医药制造业企业数量不少于8家的省级行政区，不含港澳台地区），其辖区内医药企业在未来三年内营业收入年增速可在原有基础上额外叠加5个百分点（原增速以该省全部医药企业营业收入同比增减幅的中位数衡量）；而那些尚未出台此类促进政策的省份，因缺乏政策催化，营业收入增速将较现有水平收缩20%。在上述差异化情景下，对各省营业收入以3年复合增长方式推算至2025年，请问：哪个省份在这轮重新洗牌后实现了最大幅度的营收排名晋升？对应的2025年预计营业收入总量为多少亿元？",
+  "guidelines": "依次回答省份名称和预计营业收入总额。营业收入总额以亿元为单位，保留2位小数。如[\"湖北省\", 425.18]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_resource.csv筛选正文包含\"生物\"，适用行业包含\"医药制造业\"的地方政策，找到55条生物医药产业发展相关地方政策，涉及9个省份：上海市（11条）、广东省（5条）、云南省（5条）、河南省（3条）、黑龙江省（3条）、山东省（2条）、天津市（2条）、浙江省（2条）、安徽省（1条）、甘肃省（1条）、广西壮族自治区（1条）、湖南省（1条）、吉林省（1条）、江苏省（1条）、江西省（1条）、四川省（1条）、新疆维吾尔族自治区（1条）。",
+    "从policy_resource.csv中读取这15条政策的全文内容，涉及生物医药产业发展促进相关政策省份有上海市（id=139、398、461、590、397、449、472、495）、广东省（id=92、153、605、325）、云南省（id=141、400）、河南省（id=559）、黑龙江（id=210、211、560）、山东省（id=181、284、446、天津市（id=447）、浙江省（id=393、445）、安徽省（id=444）、广西壮族自治区（id=273）、江苏省（id=443）、江西省（id=89）、四川省（id=387）、新疆维吾尔族自治区（id=556）",
+    "从company_profile.csv筛选industry=\"医药制造业\"的企业，关联company_operation_status.csv获取营业收入金额和营业收入同比增减幅，排除港澳台后得到420家有完整数据的企业。",
+    "按省份统计企业数，筛选企业数不少于8家的省份，得到14个符合条件的省份：北京市（55家）、广东省（51家）、江苏省（53家）、上海市（54家）、浙江省（45家）、山东省（22家）、四川省（15家）、湖北省（13家）、湖南省（11家）、吉林省（11家）、河南省（9家）、天津市（8家）、重庆市（8家）、福建省（8家）。",
+    "计算各省份2022年营业收入合计和营业收入同比增减幅中位数，确定2022年排名：北京市（4367.17亿，第1名）、广东省（3321.38亿，第2名）、上海市（1600.61亿，第3名）、浙江省（1423.10亿，第4名）、江苏省（1081.40亿，第5名）、山东省（960.48亿，第6名）、四川省（363.81亿，第7名）、湖南省（235.97亿，第8名）、吉林省（234.38亿，第9名）、河南省（188.79亿，第10名），其他省份缺失数据。",
+    "按假设条件调整增速：有政策省份增速额外加5个百分点（如河南省从17.90%调整为22.90%，广东省从9.93%调整为14.93%），无政策省份增速衰减20%（如北京市从3.64%调整为2.91%，山东省从7.53%调整为6.02%）。",
+    "以调整后增速进行3年复合增长预测各省2025年营收：广东省5042.83亿（第1名）、北京市4760.47亿（第2名）、上海市2637.73亿（第3名）、浙江省2137.93亿（第4名）、江苏省1802.98亿（第5名）、山东省1368.47亿（第6名）、四川市462.34亿（第7名）、河南省350.45亿（第8名）、湖南省296.34亿（第9名）、吉林省204.88亿（第10名）。",
+    "比较排名变化：河南省从第10名上升至第8名，排名上升2位，为上升幅度最大的省份"
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "从policy_release_status.csv中筛选出55条生物医药/生物经济相关地方政策，涉及17个省份。",
+    "从policy_resource.csv中分析55条政策全文，提取各省份生物医药产业发展的具体目标和支持措施，共14个省。",
+    "从company_profile.csv和company_operation_status.csv中找到420家有完整数据的医药制造业企业，涉及14个省份企业数>=8的省份。"
+  ],
+  "milestone": {
+    "生物医药/生物经济地方政策总数（条）": 55,
+    "已颁布生物医药产业发展促进相关政策的省份数（个）": 14,
+    "医药制造业有效数据企业数（家）": 420,
+    "企业数>=8的省份数（个）": 14,
+    "河南省2022年营收合计（亿元）": 188.79,
+    "河南省营收增速中位数（%）": 17.9,
+    "河南省调整后增速（%）": 22.9,
+    "河南省2025年预计营收（亿元）": 350.45,
+    "河南省2022年排名": 10,
+    "河南省2025年排名": 8,
+    "排名上升幅度（位）": 2
+  },
+  "answer": [
+    "河南省",
+    350.45
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/hard004_result.json b/assets/qa_raw/industry_planning/hard004_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..78a99731c9292e0c202f90ece19e3b5d79d04ddd
--- /dev/null
+++ b/assets/qa_raw/industry_planning/hard004_result.json
@@ -0,0 +1,43 @@
+{
+  "id": "hard004",
+  "question": "以2022年数据为起点，对中国大陆汽车制造业各省（统计范围限定为汽车制造业在册企业不少于5家的中国大陆省份，不含港澳台）进行如下情景模拟：已出台地方性新能源汽车及智能汽车产业发展专项政策的省份，其汽车制造业企业营业收入年增速将在当前中位增速基础上再叠加3个百分点（当前增速以各省企业营业收入同比增减幅中位数为准）；未出台此类专项政策的省份则呈现增长动力不足的局面，营业收入（增速取各省企业营业收入同比增减幅的中位数））增速将萎缩至原有水平的70%。按3年复合增长推算至2025年，请找出：在拥有政策支持的省份中，哪个省份将凭借政策加持超越其2022年时排名本高于自身的某个无政策省份，从而实现排名反超？该省届时的预计营业收入总量是多少亿元？",
+  "guidelines": "依次回答省份名称和预计营业收入总额。营业收入总额以亿元为单位，保留2位小数。如[\"广东省\", 5230.41]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_resource.csv筛选政策名称包含\"新能源汽车\"或\"燃料电池汽车\"或\"汽车\"的政策记录，找到13条新能源汽车相关地方政策。",
+    "从policy_resource.csv中读取这12条政策的全文内容，分析各地方政策的具体产业促进措施。涉及省份共8个，分别是北京市（id=71）、广东省（id=58、157、334、518）、海南省（id=212）、江苏省（id=601）、山东省（id=599）、上海市（id=499）、四川省（id=525）、重庆市（id=261、371）。",
+    "从company_profile.csv筛选行业=\"汽车制造业\"且省份不含港澳台的企业，按省份统计企业数量，筛选企业数>=5的省份，得到14个符合条件的省份，分别是浙江省、江苏省、广东省、上海市、山东省、北京市、湖北省、安徽省、河北省、河南省、四川省、重庆市、福建省、吉林省。",
+    "从company_operation_status.csv提取这14个省份汽车制造业企业的营业收入金额和营业收入同比增减幅，按省份计算营业收入合计和增速中位数。广东省（12407.19亿元，增速18.030%）、上海市（10315.04亿元，增速14.505%）、山东省（5249.64亿元，增速1.320%）、浙江省（3892.24亿元，增速10.120%）、北京市（3415.96亿元，增速9.400%）、河北省（3138.58亿元，增速3.295%）、江苏省（1032.08亿元，增速9.4%）、吉林省（765.19亿元，增速-2.79）、安徽省（504.95亿元，增速-5.735）、河南省（335.81亿元、增速-6.105），其余省份没有数据披露。",
+    "根据假设条件调整增速：有政策省份增速=原增速+3个百分点，无政策省份增速=原增速×0.7。",
+    "按3年复合增长预测2025年营收：北京市=3415.96×(1+12.400/100)^3=4850.79亿元，浙江省=3892.24×(1+7.084/100)^3=4779.40亿元。北京市2025年营收（4850.79亿）超过浙江省（4779.40亿），排名从第5升至第4，是唯一一个超越原排名更高的无政策省份的有政策省份。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从policy_release_status.csv中找到13条汽车相关的地方政策，涉及8个省份。",
+    "从policy_resource.csv中分析13条政策全文，共12条政策与新能源汽车及智能汽车产业发展专相关，涉及8个省份。",
+    "从company_profile.csv中找到14个符合条件的中国大陆省份共205家汽车制造业企业。",
+    "从company_operation_status.csv中获取这205家企业的营业收入金额和营业收入同比增减幅数据。"
+  ],
+  "milestone": {
+    "新能源汽车相关政策总数（条）": 12,
+    "地方性政策数（条）": 9,
+    "有政策省份数（个）": 7,
+    "符合条件省份数（企业>=5）": 14,
+    "北京市2022年营收合计（亿元）": 3415.96,
+    "北京市营收增速中位数（%）": 9.4,
+    "北京市调整后增速（%）": 12.4,
+    "浙江省2022年营收合计（亿元）": 3892.24,
+    "浙江省营收增速中位数（%）": 10.12,
+    "浙江省调整后增速（%）": 7.084,
+    "北京市2025年预计营收（亿元）": 4850.79,
+    "浙江省2025年预计营收（亿元）": 4779.4
+  },
+  "answer": [
+    "北京市",
+    4850.79
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/hard005_result.json b/assets/qa_raw/industry_planning/hard005_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7039fc5843c9ba0716b0475ee128f2cd4a18f073
--- /dev/null
+++ b/assets/qa_raw/industry_planning/hard005_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "hard005",
+  "question": "云南省于2022年正式发布了绿色铝产业发展三年行动计划，其中载明了2024年绿色铝全产业链产值的量化目标。若以该省金属冶炼和压延加工业所有上市企业（仅覆盖中国大陆范围内的金属冶炼和压延加工业企业，不含港澳台数据））2022年的实际营业收入为基数，并假设各企业按自身现有增速（取全部上市企业营业收入同比增减幅的中位数作为统一测算基准）持续保持增长，经2年复合增长推算至2024年，请问：届时这些上市企业的营收汇总值与政策文件明确设定的产业链产值目标之间还存在多大缺口（以亿元计）？若要在2年内完全弥合上述缺口，在现有增速基础上还需年均额外拉升多少个百分点？",
+  "guidelines": "依次回答缺口金额和额外增速。缺口金额以亿元为单位，保留2位小数；额外增速以百分点为单位，保留2位小数。如[280.50, 4.12]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_resource.csv筛选industry字段包含\"金属冶炼\"且province为\"云南省\"的政策记录，找到4条云南省金属冶炼和压延加工业相关政策。",
+    "从policy_resource.csv中读取这4条政策的全文内容，分析政策中关于产业发展的具体目标。在\"云南省绿色铝产业发展三年行动（2022—2024年）\"中明确提出行动目标：\"绿色铝产业链产值力争达到3500亿元左右\"，该目标年限为2024年。",
+    "从company_profile.csv筛选industry=\"金属冶炼和压延加工业\"且province=\"云南省\"的企业，找到7家上市企业。",
+    "从company_operation_status.csv提取这7家企业的营业收入金额和营业收入同比增减幅。7家企业2022年营业收入总额为279203336972.01元（约2792.03亿元），营业收入同比增减幅中位数为6.11%。",
+    "按2年复合增长预测2024年营业收入：2792.03×(1+6.11%)^2 = 3143.64亿元。与政策目标3500亿元的缺口 = 3500 - 3143.64 = 356.36亿元。",
+    "计算弥补缺口所需的年增速：设所需年增速为r，则2792.03×(1+r)^2 = 3500，解得r = √(3500/2792.03) - 1 = 11.96%。因此需要在当前6.11%的增速基础上额外提升 11.96 - 6.11 = 5.85个百分点。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从policy_release_status.csv中找到4条云南省金属冶炼和压延加工业相关政策。",
+    "从policy_resource.csv中分析4条政策全文，在绿色铝产业发展三年行动中提取到2024年产值目标3500亿元。",
+    "从company_profile.csv中找到云南省7家金属冶炼和压延加工业上市企业。",
+    "从company_operation_status.csv中获取这7家企业的营业收入金额和营业收入同比增减幅数据。"
+  ],
+  "milestone": {
+    "云南省金属冶炼相关政策数（条）": 4,
+    "政策产值目标（亿元）": 3500,
+    "云南省金属冶炼企业数（家）": 7,
+    "2022年营业收入总额（亿元）": 2792.03,
+    "营业收入同比增减幅中位数（%）": 6.11,
+    "2024年预测营业收入（亿元）": 3143.64,
+    "产值缺口（亿元）": 356.36,
+    "达标所需年增速（%）": 11.96,
+    "额外增速（百分点）": 5.85
+  },
+  "answer": [
+    356.36,
+    5.85
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/hard006_result.json b/assets/qa_raw/industry_planning/hard006_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa1a113b0c8bbd7b9e75cb17012ffac56bdf686c
--- /dev/null
+++ b/assets/qa_raw/industry_planning/hard006_result.json
@@ -0,0 +1,44 @@
+{
+  "id": "hard006",
+  "question": "聚焦2022年中国大陆非金属矿物制品业，考察以下政策差异化情景对各省（非金属矿物制品业企业数不少于5家的中国大陆省级行政区，不纳入港澳台数据）研发格局的重塑效果：凡已发布专门涉及建材行业绿色转型内容的省级碳达峰或节能减排专项实施方案的省份，其辖内非金属矿物制品业企业可保持现有研发投入增速不变（增速以该省各企业研发投入同比增减幅的中位数为准）；而尚未落地此类省级专项方案的省份，其研发扩张动能将打折，增速降至当前水平的50%。以上述差异化增速进行3年复合增长测算，预测各省2025年研发投入规模并进行重新排序，请问：从2022年到2025年，哪个省份实现了最大幅度的研发投入排名跃升？该省2025年研发投入总额约为多少亿元？",
+  "guidelines": "依次回答省份名称和预计研发投入总额。总额以亿元为单位，保留2位小数。如[\"湖北省\", 20.04]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_resource.csv筛选industry字段包含\"非金属矿物制品业\"的地方政策记录，找到30条相关政策。",
+    "从policy_resource.csv中读取这30条政策的全文内容，筛选其中属于地方政策且政策名称含\"实施方案\"或\"工作方案\"的30条地方政策，分析涉及碳达峰或节能减排，并对建材行业（水泥、玻璃、陶瓷等）有实质性内容的政策共条。",
+    "经过对政策内容的分析，筛选出7条符合条件的地方政策，涉及6个省份：湖南省（id=492）。四川省（id=116）、安徽省（id=504），贵州（id=388），江西省(id=106)，辽宁省（id=161）",
+    "从company_profile.csv筛选行业=\"非金属矿物制品业\"且不含港澳台的中国大陆企业，共125家。按省份统计企业数，筛选企业数不少于5家的省份，得到11个符合条件的省份。",
+    "从company_operation_status.csv提取这11个省份非金属矿物制品业企业的研发投入金额和研发投入同比增减幅，计算各省2022年研发投入总额和增速中位数。其中有政策的省份为安徽省（41.42亿元，52.64%）、湖南省（8.34亿元，55.19%）、四川省（1.12亿元，32.24%），无政策的省份包括广东省（31.93亿元，-1.73%）、北京市（24.94亿元，18.88%）、浙江省（23.51亿元，12.13%）等。",
+    "按假设条件调整增速：有政策省份维持当前增速，无政策省份增速减半。以3年复合增长计算2025年研发投入总额。例如湖南省：8.34×(1+55.19%)^3=8.34×3.7376=31.15亿元。",
+    "计算2022年和2025年各省排名变化：湖南省从第9名上升至第3名，排名上升6位，为上升幅度最大的省份。"
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "从policy_release_status.csv中找到39条涉及非金属矿物制品业的政策。",
+    "从policy_resource.csv中分析30条地方政策全文，筛选出7条涉及碳达峰或节能减排且对建材行业有实质性内容的省级专项方案，涉及6个省份。",
+    "从company_profile.csv中找到125家中国大陆非金属矿物制品业企业，分布在11个企业数不少于5家的省份。",
+    "从company_operation_status.csv中获取这些企业的研发投入金额和研发投入同比增减幅数据。"
+  ],
+  "milestone": {
+    "非金属矿物制品业相关政策总数（条）": 39,
+    "碳达峰/节能减排建材专项地方政策数（条）": 7,
+    "涉及省份数（个）": 6,
+    "非金属矿物制品业大陆企业数（家）": 125,
+    "符合条件省份数（企业>=5家）": 11,
+    "湖南省企业数（家）": 5,
+    "湖南省2022年研发投入总额（亿元）": 8.34,
+    "湖南省研发投入增速中位数（%）": 55.19,
+    "湖南省2022年研发投入排名": 9,
+    "湖南省2025年预计研发投入总额（亿元）": 31.15,
+    "湖南省2025年研发投入排名": 3,
+    "湖南省排名上升幅度（位）": 6
+  },
+  "answer": [
+    "湖南省",
+    31.15
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/hard007_result.json b/assets/qa_raw/industry_planning/hard007_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0728671ff245a43ad6eadf6d79ce3c83ee7ec6ee
--- /dev/null
+++ b/assets/qa_raw/industry_planning/hard007_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "hard007",
+  "question": "就2022年中国大陆软件和信息技术服务业的人才规模格局而言，若设定如下政策导向假设（仅含软件和信息技术服务业存续企业数量达到5家及以上门槛的中国大陆省份，港澳台数据不纳入计算）：凡已正式出台软件产业高质量发展专项政策的省份，其企业员工总量可沿现有轨道持续扩张，年增速以该省各企业雇员同比增减幅的中位数+13%为准；而对于尚未出台此类专项政策的省份，因政策引领欠缺导致人才吸附力不足，员工扩张速度将缩减至现有增速的一半。在这一情景设定下以3年复合增长推算至2025年，哪个省份在员工总量省际排名中实现了最大幅度的正向位次变动？请一并报告该省届时预计的从业人员总规模。",
+  "guidelines": "依次回答省份名称和预计员工总人数。员工总人数保留2位小数，单位为万人。如[\"四川省\", 2.58]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_resource.csv筛选industry字段包含'软件和信息技术服务业'的政策，共找到206条相关政策，其中地方政策142条。",
+    "从policy_resource.csv中读取全文与\"软件\"行业相关的5条地方政策全文，经过对政策内容的分析，提取出台了软件产业高质量发展专项政策且在本题统计范围内的省份：上海市（id=201、360、397）、广东省（id=153、249、516、540、589、605）、安徽省（id=4、177、67）、山东省（id=175、435、489、600）、海南省（id=315）、福建省（id=104）、湖北省（id=260、598）、云南省（id=252）、重庆市（id=96）、河南省（id=331）、山西省（id=460）",
+    "从company_profile.csv筛选industry为信息传输、软件和信息技术服务业且省份不含港澳台的企业，共620家，按省份统计企业数，筛选企业数不少于5家的省份共12个：北京市（197家）、广东省（125家）、上海市（73家）、浙江省（59家）、江苏省（43家）、福建省（27家）、四川省（17家）、山东省（17家）、湖北省（9家）、湖南省（7家）、吉林省（6家）、安徽省（5家）。",
+    "从company_operation_status.csv获取上述12个省份各企业的雇员总数和雇员同比增减幅，按省份汇总雇员总数合计与雇员同比增减幅中位数。各省雇员总数（人）和中位数增速（%）如下：广东省（361555人、-0.385%）、北京市（3836316人、-0.15%）、江苏省（96880人、1.45%）、上海市（28077人、-1.3%）、浙江省（647855人、-4.21%）、山东省（36446人、6.17%）、四川省（31651人、3.68%）、安徽省（28505人、-6.02%）、湖南省（9413人、0.75%）、吉林省（12133人、3.745）、其他省份缺失数据。",
+    "确定各省适用增速：有政策省份（上海市、广东省、山东省、安徽省）共4个，使用原始中位数增速+13%，无政策省份增速减半。以3年复合增长计算2025年预测雇员总数：2025年雇员 = 2022年雇员 × (1 + 适用增速)^3。有政策省份：广东省361555×(1-((-0.385)+13)/100)^3≈516372人；上海市280077×(1-(0.012+13)/100)^3≈391384人；山东省280077×(1-(6.17+13)/100)^3≈61681人；安徽省28505×(1-(-6.02+13)/100)^3≈34900。无政策省份以减半增速计算。",
+    "对比各省2022年排名和2025年预测排名：2022年安徽省雇员总数28505人，排名第8位；2025年预测34900人，超越四川省（32533人），上升至第7位，排名提升1位，为所有省份中排名上升幅度最大的省份。安徽省2025年预计员工总人数34900人人，约合3.49万人。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从policy_release_status.csv中找到206条涉及软件和信息技术服务业的政策，其中地方政策142条。",
+    "从policy_resource.csv中分析142条以软件命名的地方政策全文，确定上海市、广东省、安徽省、山东省、海南省、福建省、湖北省、云南省、重庆市、河南省、山西省出台了软件产业高质量发展专项政策。",
+    "从company_profile.csv中找到企业数不少于5家的12个大陆省份，共620家软件和信息技术服务业企业。",
+    "从company_operation_status.csv中获取12个省份的企业雇员总数合计与雇员同比增减幅中位数，完成3年复合增长预测和排名对比。"
+  ],
+  "milestone": {
+    "软件相关地方政策总数（条）": 142,
+    "地方政策涉及软件高质量发展的省份数量": 11,
+    "有政策且企业数>=5的省份数（个）": 4,
+    "企业数>=5的省份总数（个）": 12,
+    "安徽省2022年雇员总数（人）": 28505,
+    "安徽省2022年雇员同比增减幅中位数（%）": -6.02,
+    "安徽省2025年预测雇员（人）": 34900,
+    "安徽省2022年排名（位）": 8,
+    "安徽省2025年排名（位）": 7,
+    "四川省2025年预测雇员（人）": 12827.41
+  },
+  "answer": [
+    "安徽省",
+    3.49
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/hard008_result.json b/assets/qa_raw/industry_planning/hard008_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2053ae6828e6bdafc621014e03d38010b6b1bf58
--- /dev/null
+++ b/assets/qa_raw/industry_planning/hard008_result.json
@@ -0,0 +1,39 @@
+{
+  "id": "hard008",
+  "question": "以2022年为基期，针对中国大陆通用设备制造业设计双情景对比测算框架（通用设备制造业在册企业数量不少于5家的中国大陆省份，港澳台不纳入）。情景一（政策分化情景）：已出台制造业高质量发展省级专项文件的省份，其通用设备制造业企业总资产按各自当前增速+6%（以省内各企业营业收入同比增减幅的中位数作为总资产增速的替代指标）持续扩张，无政策省份则以减半增速计算；情景二（全量减半基准情景）：所有省份不论有无政策，一律按当前增速的一半推算总资产增长。以3年复合增长分别推算两种情景下各省2025年总资产，并以两情景之差作为\"政策带来的额外总资产增量\"，请问：在出台了上述专项政策且符合最低企业数量门槛的省份中，哪个省份从该政策中撬动的额外总资产增量（额外增量=政策分化情景下2025年总资产 - 全量减半情景下2025年总资产）最为可观？该增量具体为多少亿元？",
+  "guidelines": "依次回答省份名称和额外总资产增量。额外增量保留2位小数，单位为亿元。如[\"湖南省\", 520.38]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_resource.csv筛选政策类型为地方政策且政策名称包含制造业的政策，共找到7条候选政策。",
+    "从policy_resource.csv中读取上述7条政策全文内容，深入分析各政策的涉及制造业高质量发展的省份有：福建省（id=78、）、上海市（id=75、398）、广东省（id=154）、湖南省（id=184）。",
+    "从company_profile.csv筛选行业为通用设备制造业的企业，排除港澳台，共得到212家中国大陆通用设备制造业上市企业；按省份统计企业数量，筛选企业数不少于5家的省份，得到11个合规省份：浙江省（62家）、江苏省（42家）、广东省（20家）、山东省（17家）、上海市（12家）、四川省（10家）、辽宁省（7家）、安徽省（6家）、湖北省（6家）、福建省（5家）、湖南省（5家）。",
+    "从company_operation_status.csv提取11个合规省份通用设备制造业企业的总资产和营业收入同比增减幅，按省份计算总资产合计（亿元）和营业收入同比增减幅中位数（%），分别为广东省总资产合计1536.03亿元、中位数增速-5.745%；江苏省总资产合计1557.78亿元、中位数增速0.635%；上海市总资产合计7001.17亿元、中位数增速-10.94%；浙江省3177.50亿元、中位数增速1.18%；山东省633.65亿元、中位数增速9.19%；四川省2658.05亿元、中位数增速15.255%；安徽省167.97亿元、中位数增速-2.095%；湖南省222.13亿元、-10.57%；河北省57.30亿元、中位数增速66.665%；辽宁省374.25亿元、中位数增速10.12%；其他省份缺失数据。",
+    "对3个有政策省份分别计算两种情景下2025年总资产：情景A（有政策，原速增长）= 2022年总资产×(1+原增速/100)^3；情景B（全部减半）= 2022年总资产×(1+原增速/200)^3；额外增量=情景A-情景B。广东省：1547.81亿元-1407.43亿元=140.38亿元",
+    "三个有政策省份中，四川省以额外总资产增量755.66亿元排名第一，为从该政策中获得边际贡献最大的省份。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从policy_release_status.csv中找到5条制造业高质量发展相关地方政策，确认涉及湖南省、云南省、上海市、四川省4个省份。",
+    "从policy_resource.csv中读取4条政策全文，分析发布主体和政策性质，确认湖南省、上海市、四川省（成都市）为出台制造业高质量发展专项政策的合规省份。",
+    "从company_profile.csv筛选出通用设备制造业212家中国大陆上市企业，确定11个企业数不少于5家的合规省份。",
+    "从company_operation_status.csv中获取11个合规省份企业的总资产合计和营业收入同比增减幅数据，用于两种情景下的总资产预测计算。"
+  ],
+  "milestone": {
+    "制造业高质量发展专项政策涉及省份数（个）": 4,
+    "通用设备制造业合规省份总数（个）": 11,
+    "满足企业数不少于5家条件的有政策省份数（个）": 3,
+    "广东省2022年总资产合计（亿元）": 1547.81,
+    "广东省营业收入增速中位数（%）": -5.745,
+    "广东省2025年总资产（有政策情景，亿元）": 1547.81,
+    "广东省2025年总资产（全部减半情景，亿元）": 1407.43,
+    "广东省额外总资产增量（亿元）": 140.38
+  },
+  "answer": [
+    "广东省",
+    140.38
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/hard009_result.json b/assets/qa_raw/industry_planning/hard009_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..dca56cd4964cb85799087d7ec94b02d122dec405
--- /dev/null
+++ b/assets/qa_raw/industry_planning/hard009_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "hard009",
+  "question": "以2022年铁路、船舶、航空航天和其他运输设备制造业的省级（限定为铁路、船舶、航空航天和其他运输设备制造业辖内企业数量达到3家以上的中国大陆省级行政区，港澳台数据不纳入）数据为基准，模拟如下政策分化对行业格局的冲击：已出台船舶与海洋工程装备产业专项发展政策的省份，其企业营业收入增速在未来三年将在现有中位水平上额外叠加5个百分点；而没有落地此类专项政策的省份，受制于政策真空，营业收入增速将萎缩至现有水平的70%（增速以各省企业营业收入同比增减幅中位数为准）。按各省调整后的增速进行3年复合增长推算，对比2022年与2025年的省际营收排名变动，请从无政策省份中找出：哪个省份享受政策红利而获得最大幅度的排名上升？它一共上升了几位？该省2025年的预测营业收入总量为多少亿元？",
+  "guidelines": "依次回答省份名称、排名下降名数（整数）、该省2025年预计营业收入总额。营业收入总额保留2位小数，单位为亿元。如[\"广东省\", 2, 512.34]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_resource.csv筛选涉及行业包含「船舶」或「海洋工程」关键词的政策，共找到46条相关政策，其中地方政策36条。",
+    "从policy_resource.csv中读取46条相关政策的全文内容，对每条政策进行深度分析，判断其是否专门针对船舶与海洋工程装备制造产业。经分析，绝大多数政策属于泛行业或节能减排政策，仅有少数政策以船舶与海洋工程装备为核心支持方向：上海市（id=75、139、397）、河南省（id=87）、广东省（id=153、303）、山东省（id=181、284）。",
+    "从company_profile.csv中筛选行业为铁路、船舶、航空航天和其他运输设备制造业筛选有效省份，共得到8个有效省份：北京市（22家）、江苏省（14家）、广东省（9家）、浙江省（6家）、四川省（6家）、上海市（3家）、山东省（3家）、湖南省（3家）。",
+    "从regional_industry_status.csv筛选行业为铁路、船舶、航空航天和其他运输设备制造业、省份为中国大陆、上述省份营业收入金额合计及营业收入同比增减幅中位数数据为：广东省（307.73亿元，-1.44%）、北京市（7049.18亿元，-1.055%）、江苏省（561.84亿元，7.775%），上海市（620.95亿元，-0.3%），浙江省（264.27亿元，27.62%），山东省（431.65亿元，22.08%）、四川省（87.02亿元，10.23%）、湖南省（377.54亿元，19.23）。",
+    "按政策情景计算各省实际增速：有政策省份（上海市、广东省、山东省）在原增速基础上额外提升5个百分点；无政策省份增速衰减至原增速的70%。以2022年营业收入为基数，按3年复合增长计算各省2025年预计营业收入：广东省（330.03亿元）、北京市（6930.69亿元）、江苏省（633.91亿元），上海市（680.69亿元），浙江省（393.95亿元），山东省（697.09亿元）、四川省（101.85亿元）、湖南省（502.64亿元）",
+    "对比各省排名变化：山东省从第4名上升到第2名，上升2位，2025年预计营收697.09亿元。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从policy_release_status.csv中找到46条涉及船舶或海洋工程行业的政策，其中地方政策36条。",
+    "从policy_resource.csv中读取并深度分析46条政策全文，识别出2个省份共3条政策专门针对船舶与海洋工程装备产业：上海市（《上海打造未来产业创新高地发展壮大未来产业集群行动方案》专项部署深远海船舶与海洋工程装备产业集群；《聚焦临港核心区打造上海全球动力之城实施方案》专项支持船舶动力研发制造）、山东省（《山东省新旧动能转换重大产业攻关项目管理实施细则》将海洋工程装备及高技术船舶列为重点攻关方向）。",
+    "从regional_industry_status.csv中筛选出8个有效省份（铁路、船舶、航空航天和其他运输设备制造业企业不少于3家且有营收数据的中国大陆省份）：北京市、上海市、江苏省、山东省、湖南省、广东省、浙江省、四川省，获取其2022年营业收入金额合计和营业收入同比增减幅中位数。"
+  ],
+  "milestone": {
+    "涉及船舶或海洋工程行业的政策总数（条）": 46,
+    "地方政策数（条）": 36,
+    "出台专项政策的省份数（个）": 4,
+    "有效省份总数（个）": 8,
+    "无政策省份总数（个）": 5,
+    "山东省2022年预计营收（亿元）": 431.65,
+    "山东省2025年预计营收（亿元）": 697.09,
+    "山东省2022年排名（名）": 4,
+    "山东省2025年排名（名）": 2,
+    "山东省排名上市幅度（名）": 2
+  },
+  "answer": [
+    "山东省",
+    2,
+    697.09
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/hard010_result.json b/assets/qa_raw/industry_planning/hard010_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5ba56c0d9055874e47c3cd4f7dceebd0f4f4c36
--- /dev/null
+++ b/assets/qa_raw/industry_planning/hard010_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "hard010",
+  "question": "在铁路、船舶、航空航天和其他运输设备制造业（即广义轨道交通装备制造业）领域，以2022年为基期，构建以下轨道交通政策激励传导模型（轨道交通装备制造相关企业存续数量不低于3家的中国大陆省级行政区，港澳台不计入）：若某省已明确将轨道交通装备产业列入重点打造产业集群目录并配套专项支持措施，则该省企业营业收入的年增速可在现有中位水平上再叠加3个百分点；而未落地此类产业集群专项支持政策的省份，其营业收入增速将在原有基础上收缩30%（增速取省内各企业营业收入同比增减幅中位数）。以上述情景增速进行3年复合增长预测，并计算各有政策省份从2022年到2025年的营业收入绝对增量（定义为：2025年预测营业收入 − 2022年实际营业收入），请问：哪个获得政策支持的省份营业收入的绝对增量最为可观？该省这一增量数值为多少亿元？",
+  "guidelines": "依次回答省份名称和营业收入绝对增量。绝对增量保留2位小数。如[\"上海市\", 185.42]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选涉及行业包含「铁路、船舶、航空航天和其他运输设备制造业」的政策，共找到46条，其中地方政策36条、部委政策10条。",
+    "从policy_resource.csv中读取上述36条地方政策的全文内容，逐一分析政策正文是否明确将轨道交通装备列为重点打造的产业集群并给予专项支持。经过深度内容分析，共6个省份有相关内容：上海市（id=75、139、397）、云南省（id=152）、广东省（id=605）、湖南省（id=194、417）、山东省（id=284）、四川省（id=387）。",
+    "从company_profile.csv筛选行业为「铁路、船舶、航空航天和其他运输设备制造业」的大陆企业（排除港澳台），共96家，按省份统计企业数量，筛选企业数不少于3家的省份，共12个：北京市（22家）、江苏省（14家）、广东省（9家）、陕西省（6家）、四川省（6家）、浙江省（6家）、黑龙江省（4家）、山东省（3家）、湖北省（3家）、湖南省（3家）、贵州省（3家）、上海市（3家）。",
+    "从company_operation_status.csv获取上述12个省份的铁路运输设备制造业企业营业收入金额和营业收入同比增减幅，计算各省营业收入合计（2022年基准值）和增速中位数。广东省：营业收入合计307.73亿元，增速中位数-1.44%；北京市：营业收入合计7049.18亿元，增速中位数-1.055%；江苏省：营业收入合计561.84亿元，增速中位数7.775%；上海市：营业收入合计620.95亿元，增速中位数-1.055%；浙江省：营业收入合计264.27亿元，增速中位数-0.3%；山东市：营业收入合计431.65亿元，增速中位数-22.08%；四川省：营业收入合计87.02亿元，增速中位数10.23%；湖南省：营业收入合计377.54亿元，增速中位数19.23%。",
+    "对有政策省份，按有效增速 = 增速中位数 + 3% 计算，3年复合增长预测2025年营业收入。山东省＝431.65*（1+（22.08+3）/100）=844.69亿元",
+    "比较有政策省份的营业收入绝对增量：山东省844.69亿元 - 431.65亿元=413.38亿元，山东省的营业收入绝对增量最大。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从policy_release_status.csv中找到46条涉及铁路、船舶、航空航天和其他运输设备制造业的相关政策，其中地方政策36条。",
+    "从policy_resource.csv中分析36条地方政策全文，筛选出3条明确将轨道交通装备列为重点打造产业集群并给予专项支持的政策，涉及湖南省（2条）和上海市（1条）。",
+    "从company_profile.csv中找到96家大陆铁路运输设备制造业企业，筛选出企业数不少于3家的12个省份。",
+    "从company_operation_status.csv中获取12个省份铁路运输设备制造业企业的营业收入金额和同比增减幅数据，计算各省基准营业收入和增速中位数。"
+  ],
+  "milestone": {
+    "铁路运输设备制造业相关政策总数（条）": 46,
+    "地方政策数量（条）": 36,
+    "有政策省份数（个）": 6,
+    "满足企业数>=3条件的省份数（个）": 12,
+    "山东省2022年营业收入合计（亿元）": 431.65,
+    "山东省增速中位数（%）": 22.08,
+    "山东省调整后增速（%）": 25.08,
+    "山东省2025年预测营业收入（亿元）": 844.69,
+    "山东省营业收入绝对增量（亿元）": 413.38
+  },
+  "answer": [
+    "山东省",
+    413.38
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/hard011_result.json b/assets/qa_raw/industry_planning/hard011_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..63709a56f5c0b8d5b31c82fd56023091a72a2702
--- /dev/null
+++ b/assets/qa_raw/industry_planning/hard011_result.json
@@ -0,0 +1,37 @@
+{
+  "id": "hard011",
+  "question": "在铁路、船舶、航空航天和其他运输设备制造业中，以2022年各省（港澳台地区数据不计入本题统计范围）实际数据为起点，设定如下情景假设：已颁布航空航天产业发展专项支持政策的省份，其行业内企业能够维持现有营业收入增速（以各省企业营业收入同比增减幅的中位数衡量）持续扩张；而未出台此类专项支持政策的省份，由于缺乏政策引导，营业收入增速将被动压缩至当前水平衰减40%。在以上差异化条件下按3年复合增长推算至2025年，请重点关注2022年营业收入总量尚未跨越100亿元门槛的有政策省份——在这一子集中，哪个省份将在政策扶持下率先实现百亿营收的历史性突破？该省届时的预测营业收入总额为多少亿元？",
+  "guidelines": "依次回答省份名称和2025年预计营业收入总额。营业收入总额保留2位小数，单位为亿元。如[\"江西省\", 89.37]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选policyClassification为地方政策且industry包含铁路、船舶、航空航天和其他运输设备制造业的记录，得到36条地方政策。",
+    "从policy_resource.csv中读取上述36条政策的全文内容，逐条分析政策正文中是否含有针对航空航天产业的专项支持措施。经分析共有个6省份有相关政策，分别是：上海市（id=75、139）、广东省（id=153）、湖北省（id=172）、湖南省（id=194）、江西省（id=89）、四川省（id=387）。",
+    "从company_profile.csv筛选industry为铁路、船舶、航空航天和其他运输设备制造业且省份不含港澳台的企业，共得到96家企业，分布在21个省份。北京市（22家）、江苏省（14家）、广东省（9家）、陕西省（6家）、四川省（6家）、浙江省（6家）、黑龙江省（4家）、山东省（3家）、湖北省（3家）、湖南省（3家）、贵州省（3家）、上海市（3家）、安徽省（2家）、河北省（2家）、河南省（2家）、江西省（2家）、重庆市（2家）、吉林省（1家）、内蒙古自治区（1家）、山西省（1家）、天津市（1家）",
+    "从company_operation_status.csv提取上述96家企业的营业收入金额和营业收入同比增减幅，按省份汇总，计算各省营业收入总额和增速中位数。有政策省份中，2022年营业收入总额低于100亿元的共2个：四川省（87.02亿元，中位增速10.23%）、安徽省（12.16亿，中位增速18.66）、吉林省（4.23亿元，中位增速-3.50%）。无政策省份低于100亿元的有：安徽省（14.16亿元）、山西省（12.45亿元）、吉林省（4.23亿元）。",
+    "对2022年营业收入低于100亿元的有政策省份按3年复合增长计算2025年预计营业收入：四川省：87.024112亿元 × (1 + 10.23/100)^3 = 116.5572亿元，超过100亿元门槛（2024年即已突破）;安徽省：14.1649亿元 × (1 + 18.66*0.6/100)^3 = 19.4752亿元，未超过100亿元门槛;吉林省：4.2264亿元 × (1 + 3.42*0.6/100)^3 = 4.4919亿元，未超过100亿元门槛。",
+    "结论：在2022年营业收入低于100亿元的有政策省份中，四川省将首次突破100亿元门槛，2025年预计营业收入总额为116.56亿元。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从policy_release_status.csv中筛选到36条涵盖铁路、船舶、航空航天和其他运输设备制造业的地方政策，涉及湖南省、江西省、四川省、湖北省等省份。",
+    "从policy_resource.csv中读取并分析36条政策全文，发现湖南省、江西省、四川省、湖北省共4个省份出台了对航空航天产业具有专项发展支持的地方政策。",
+    "从company_profile.csv中找到中国大陆21个省份共96家铁路、船舶、航空航天和其他运输设备制造业企业（不含港澳台）。",
+    "从company_operation_status.csv中获取96家企业的营业收入金额和同比增减幅，按省份汇总后，确认有政策且2022年营业收入低于100亿元的省份为四川省（87.02亿元，中位增速10.23%）和江西省（73.65亿元，中位增速-3.50%）。"
+  ],
+  "milestone": {
+    "分析地方政策数量（条）": 36,
+    "认定为有航空航天专项政策的省份数（个）": 4,
+    "有政策且2022年营业收入低于100亿元的省份数（个）": 3,
+    "四川省2022年营业收入合计（亿元）": 87.02,
+    "四川省营业收入中位增速（%）": 10.23,
+    "四川省2025年预计营业收入（亿元）": 116.56
+  },
+  "answer": [
+    "四川省",
+    116.56
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/hard012_result.json b/assets/qa_raw/industry_planning/hard012_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..726ca30f4429aac2d2623ea43fee446160d1fd7d
--- /dev/null
+++ b/assets/qa_raw/industry_planning/hard012_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "hard012",
+  "question": "2022年，在中国大陆医疗仪器设备及器械制造业中（仅统计仪器仪表制造业企业不少于3家的中国大陆省份，不含港澳台），假设出台了医疗器械产业高端化（适用于医药制造和仪表仪器）发展专项政策的省份，其企业净利润增速在未来3年额外提升5个百分点，而未出台此类政策的省份净利润增速衰减20%，到2025年净利润总额排名上升幅度最大的省份是哪个（增速使用各省份企业净利润同比增减幅的中位数；增长方式为3年复合增长）？其预计净利润总额是多少亿元？",
+  "guidelines": "依次回答省份名称和预计净利润总额。净利润总额保留2位小数。如[\"安徽省\", 6.92]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从regional_industry_status.csv筛选行业为仪器仪表制造业、省份为中国大陆（排除台湾省、香港特别行政区、澳门特别行政区）且企业总数不少于3家、净利润金额合计和净利润同比增减幅中位数均非空的省份，共得到9个符合条件的省份：浙江省（20家）、江苏省（14家）、广东省（12家）、上海市（7家）、北京市（8家）、河南省（4家）、安徽省（4家）、湖南省（3家）、四川省（3家）。",
+    "从policy_release_status.csv筛选上述9个省份的仪器仪表制造业及医药制造业相关地方政策，共38条；",
+    "从policy_resource.csv中读取上述重点政策的全文内容，共4个省份出台了医疗器械产业高端化相关政策，分别是：上海市（id=75、590）、广东省（id=92、153、341）、江苏省（id=443）、安徽省（id=444）",
+    "根据政策情景计算各省份调整后净利润增速：有政策省份的调整增速 = 净利润同比增减幅中位数 + 5个百分点；无政策省份的调整增速 = 净利润同比增减幅中位数 × 0.8（衰减20%）。各省调整后增速：浙江省 -0.155%×0.8 = -0.124%；江苏省 3.615%+5 = 8.615%；广东省 -9.585%+5 = -4.585%；湖南省 7.670%×0.8 = 6.136%；河南省 -37.150%×0.8 = -29.720%；安徽省 10.675%+5 = 15.675%；上海市 -45.100%+5 = -40.100%；北京市 20.745%×0.8 = 16.596%；四川省 -3.500%×0.8 = -2.800%。",
+    "以2022年净利润金额合计为基数，按各省调整后增速进行3年复合增长，计算2025年预计净利润总额：浙江省 33.5927亿×(1-0.00124)³ = 33.47亿；江苏省 15.6321亿×(1+0.08615)³ = 20.03亿；广东省 8.5411亿×(1-0.04585)³ = 7.42亿；湖南省 6.6674亿×(1+0.06136)³ = 7.97亿；河南省 6.0233亿×(1-0.2972)³ = 2.09亿；安徽省 4.4203亿×(1+0.15675)³ = 6.84亿；上海市 4.0677亿×(1-0.401)³ = 0.87亿；北京市 2.7353亿×(1+0.16596)³ = 4.34亿；四川省 1.2727亿×(1-0.028)³ = 1.17亿。",
+    "按2025年预计净利润总额排名：第1浙江省（33.47亿）、第2江苏省（20.03亿）、第3湖南省（7.97亿）、第4广东省（7.42亿）、第5安徽省（6.84亿）、第6北京市（4.34亿）、第7河南省（2.09亿）、第8四川省（1.17亿）、第9上海市（0.87亿）。与2022年排名对比，北京市从第8名升至第6名，上升2位，为排名上升幅度最大的省份；其2025年预计净利润总额为4.34亿元。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从regional_industry_status.csv中筛选仪器仪表制造业中国大陆各省数据，企业总数≥3且净利润数据完整的省份共9个。",
+    "从policy_release_status.csv中找到9个省份的仪器仪表制造业相关地方政策共20条，另筛选涉及医疗器械和生物医药高端化发展的地方政策20余条。",
+    "从policy_resource.csv中读取并深度分析重点政策全文，识别出4个出台医疗器械产业高端化发展专项政策的省份：上海市（id=75、590）、广东省（id=92、153、341）、江苏省（id=443）、安徽省（id=444）。",
+    "从regional_industry_status.csv中获取9个省份的净利润金额合计和净利润同比增减幅中位数，作为基础数据进行2025年预测计算。"
+  ],
+  "milestone": {
+    "符合条件省份数（个）": 9,
+    "出台专项政策省份数（个）": 4,
+    "北京市2022净利润合计（亿元）": 2.74,
+    "北京市净利润增速中位数（%）": 20.745,
+    "北京市调整后增速（%）": 16.596,
+    "北京市2022排名（名）": 8,
+    "北京市2025预计净利润（亿元）": 4.34,
+    "北京市2025排名（名）": 6,
+    "北京市排名上升幅度（位）": 2
+  },
+  "answer": [
+    "北京市",
+    4.34
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/hard013_result.json b/assets/qa_raw/industry_planning/hard013_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..86e745b5da224e78dc6421761cef262f37d2f51a
--- /dev/null
+++ b/assets/qa_raw/industry_planning/hard013_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "hard013",
+  "question": "假设针对金属冶炼和压延加工业出台了新材料产业发展专项扶持政策的省份（仅统计2022年净利润数据完整的中国大陆省份，不含港澳台），其金属冶炼和压延加工业上市企业净利润按2022年的增速（各省企业净利润同比增减幅中位数）持续增长，而未出台此类政策的省份净利润增速减半，在3年复合增长模型下，到2025年有政策省份净利润总和与无政策省份净利润总和的比值是多少？相比2022年该比值变化了多少（提升或下降）？",
+  "guidelines": "依次回答2025年的比值和比值变化量。比值保留2位小数，变化量保留2位小数并注明提升或下降。如[1.85, \"提升0.93\"]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_resource.csv筛选industry字段包含\"金属冶炼和压延加工业\"的政策记录，共找到21条相关政策，其中地方政策19条。",
+    "从policy_resource.csv中读取上述19条政策的全文内容，进行深度分析，出台推进新材料产业扶持相关政策的省份有10个：上海市（id=75）、河南省（id=87）、内蒙古自治区（id=101）、云南省（id=128、152）、湖南省（id=194、492）、山东省（id=284）、广西壮族自治区（id=273）、贵州省（id=276）、四川省（id=526、387）、新疆维吾尔自治区（id=556）。",
+    "从regional_industry_status.csv筛选行业=\"金属冶炼和压延加工业\"的记录，剔除港澳台，进一步筛选净利润金额合计数据完整且企业总数大于0的省份，共得到14个有效省份：广东省、北京市、江苏省、上海市、浙江省、山东省、四川省、安徽省、湖南省、河南省、河北省、辽宁省、吉林省、新疆维吾尔自治区、山西省。注：云南省和贵州省虽有专项政策但净利润数据缺失，不纳入计算。有效数据范围内的政策省份为四川省和河南省，无政策省份为其余13个省份。",
+    "读取2022年各省净利润金额合计和净利润同比增减幅中位数。2022年有政策/无政策比值 = 1108.93 / 610.96 = 1.82。",
+    "按情景假设计算各省2025年预计净利润（3年复合增长）：有政策省份保持原增速，四川省预计净利润 = 650.92 × (1 + 46.61%)³ = 2051.25亿元；河南省预计净利润 = 91.79 × (1 + (-8.17%))³ = 71.08亿元；湖南省预计净利润 = 76.12 × (1 + （-56.83%)）³ = 61.23亿元；上海市预计净利润 = 142.58 × (1 + （-31.27）%)³ = 46.29亿元；山东省预计净利润 = 14.75 × (1 + 9.10%)³ = 19.16亿元；四川省预计净利润 = 64.61 × (1 + 46.61%)³ = 205.13亿元；；有政策省份2025年合计 = 2366.32亿元。",
+    "无政策省份各省增速减半后计算2025年预计净利润，13省合计 = 654.81亿元。",
+    "计算2025年比值 = 2366.32 / 654.81 = 3.61；比值变化量 = 3.61 - 1.82 = 1.79，比值提升1.51。"
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "从policy_release_status.csv中找到41条涉及金属冶炼和压延加工业的政策，其中地方政策32条，筛选出4条新材料产业发展专项扶持政策候选。",
+    "从policy_resource.csv中读取4条候选政策全文，经分析确认云南省、河南省、四川省、贵州省共4个省份出台了新材料产业发展专项扶持政策。",
+    "从regional_industry_status.csv中筛选金属冶炼和压延加工业数据，在净利润数据完整的中国大陆省份中确定15个有效省份，其中有效数据范围内政策省份2个（四川省、河南省），无政策省份13个。"
+  ],
+  "milestone": {
+    "有效数据省份总数（个）": 14,
+    "有效数据范围内政策省份数（个）": 5,
+    "有政策省份2022年净利润合计（亿元）": 1108.93,
+    "无政策省份2022年净利润合计（亿元）": 610.96,
+    "2022年比值（有政策/无政策）": 1.82,
+    "有政策省份2025年净利润预计合计（亿元）": 2366.32,
+    "无政策省份2025年净利润预计合计（亿元）": 654.81,
+    "2025年比值（有政策/无政策）": 3.61,
+    "比值变化量": 1.79
+  },
+  "answer": [
+    3.61,
+    "提升1.79"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/hard014_result.json b/assets/qa_raw/industry_planning/hard014_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3edd546d1a23bed2c853d5b8c471b54014cb6586
--- /dev/null
+++ b/assets/qa_raw/industry_planning/hard014_result.json
@@ -0,0 +1,38 @@
+{
+  "id": "hard014",
+  "question": "2022年，在中国大陆消费电子及电气业中，假设出台了电子信息产业集群培育专项政策的省份（仅统计消费电子及电气业企业不少于5家的中国大陆省份，不含港澳台），其企业净利润增速在未来3年额外提升8个百分点，而未出台此类专项政策的省份净利润增速衰减至当前水平的50%（净利润增速取各省份企业净利润同比增速的中位数；增长方式为3年复合增长），到2025年，在2022年净利润总额排名处于后半段（排名靠后一半）的有政策省份中，排名提升幅度最大的是哪个省份？其预计净利润总额是多少亿元？",
+  "guidelines": "依次回答省份名称和预计净利润总额。净利润总额保留2位小数。如[\"江西省\", 45.82]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选policyClassification为'地方政策'、industry字段包含'消费电子'或'通信传输'或'半导体'的政策记录，得到54条涉及电子类行业的地方政策。",
+    "从policy_resource.csv中读取上述54条地方政策全文，分析涉及电子信息产业集群培育专项地方政策的省份有4个：安徽省（id=301、609）、广东省（id=23、153、249、605）、四川省（id=387）、浙江省（id=125）。",
+    "从company_profile.csv筛选industry='消费电子及电气业'且province不含港澳台的企业，共344家。统计各省企业数，筛选企业数不少于5家的省份，得到12个有效省份：广东省、山东省、浙江省、北京市、湖南省、江苏省、四川省、湖北省、安徽省、江西省、上海市、福建省。",
+    "从company_operation_status.csv提取这12个省份消费电子及电气业企业的净利润金额和净利润同比增减幅数据。计算各省净利润合计及增速中位数，按2022年净利润合计降序排名：第1广东省（1223.93亿），第2山东省（364.09亿），第3浙江省（341.43亿），第4北京市（160.74亿），第5湖南省（47.10亿），第6江苏省（46.71亿），第7四川省（31.29亿），第8湖北省（19.13亿），第9安徽省（15.80亿），第10江西省（4.83亿），第11上海市（1.09亿），第12福建省（-10.34亿）。",
+    "确定后半段（排名第7至第12名）且有政策的省份：四川省（第7名，增速中位数18.57%）和江西省（第10名，增速中位数-25.48%）。按政策情景计算：四川省调整后增速=18.57%+8%=26.57%，2025年预测净利润=31.29×(1+(18.57 + 8)/100)³=63.45亿元，2025排名升至第5名（提升2位）。"
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "从policy_release_status.csv中找到54条涉及消费电子及电气业、通信传输设备业或半导体业的地方政策。",
+    "从policy_resource.csv中分析55条的地方政策全文，最终认定安徽省、广东省、四川省、浙江省共4个省份出台了电子信息产业集群培育专项政策。",
+    "从company_profile.csv中找到344家消费电子及电气业大陆企业，筛选出企业数不少于5家的12个省份。",
+    "从company_operation_status.csv中获取这12个省份企业的净利润金额和净利润同比增减幅数据，计算各省净利润合计与增速中位数。"
+  ],
+  "milestone": {
+    "涉及电子类行业的地方政策总数（条）": 54,
+    "认定有政策的省份数（个）": 3,
+    "符合条件的省份总数（个）": 12,
+    "四川省2022净利润合计（亿元）": 31.29,
+    "四川省净利润增速中位数（%）": 18.57,
+    "四川省政策加持后增速（%）": 26.57,
+    "四川省2025净利润预测（亿元）": 63.45,
+    "四川省2025排名提升幅度（位）": 2
+  },
+  "answer": [
+    "四川省",
+    63.45
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/medium001_result.json b/assets/qa_raw/industry_planning/medium001_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..9b35e26ff9865ca7d1a718a9248b873402f5acc7
--- /dev/null
+++ b/assets/qa_raw/industry_planning/medium001_result.json
@@ -0,0 +1,38 @@
+{
+  "id": "medium001",
+  "question": "Based on 2022 data, the following policy effect scenario is simulated: For provinces that have promulgated industrial policies containing the keywords \"semiconductor\" or \"integrated circuit\", policy empowerment accelerates the R&D investment expansion pace of their semiconductor enterprises to 2 times the current growth rate over the next 3 years; for provinces that have not yet issued such policies, R&D growth rate remains unchanged. Using the median year-on-year change in enterprise R&D investment as the baseline growth rate for each province, and projecting with 3-year compound growth, which province will have the highest total semiconductor industry R&D investment nationwide by 2025? What is the corresponding estimated amount?",
+  "guidelines": "Answer format: [province name, value (2 decimal places, unit: yuan)]. If relevant data cannot be found, please answer \"No relevant data found\".",
+  "steps": [
+    "Filter from policy_release_status.csv policy records whose name contains \"semiconductor\" or \"integrated circuit\", obtaining 4 records. Extract the list of involved provinces (deduplicated, excluding national policies), obtaining 4 provinces: Guangdong, Zhejiang, Shanghai, Anhui.",
+    "Filter from company_profile.csv all enterprise records with industry=\"semiconductor industry\", extract enterprise name, bmCode, and province fields, finding 172 enterprises.",
+    "From company_operation_status.csv, link the above enterprises by bmCode, extract R&D investment amount and year-on-year R&D investment change rate fields, filter out enterprises with either field empty, obtaining 168 valid enterprises.",
+    "Group by province, calculate total R&D investment amount and median year-on-year R&D investment change rate for each province, totaling 22 provinces.",
+    "For each province, determine whether it belongs to provinces with policy: adjusted growth rate for provinces with policy = median growth rate × 2; adjusted growth rate for provinces without policy = median growth rate. The top-ranked province Shanghai has policy support, with adjusted growth rate of 64.38%.",
+    "Calculate estimated total R&D investment for 2025 for each province = total R&D investment amount × (1 + adjusted growth rate/100)^3. Shanghai's estimated 2025 R&D investment = 22003461800.09 × (1+64.38/100)^3 = 97732260069.03 (yuan).",
+    "Sort all provinces by estimated total R&D investment for 2025 in descending order.",
+    "The top-ranked province is Shanghai, with estimated total R&D investment for 2025 of 97732260069.03 yuan."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "Found 4 policy records containing \"semiconductor\" or \"integrated circuit\" keywords in policy_release_status.csv, involving 4 provinces.",
+    "Found 172 semiconductor industry enterprises in company_profile.csv.",
+    "Found R&D investment amounts and year-on-year R&D investment change rates for 168 semiconductor industry enterprises in company_operation_status.csv."
+  ],
+  "milestone": {
+    "Provinces with policy": 4,
+    "Valid semiconductor industry enterprises": 168,
+    "Provinces involved": 22,
+    "Shanghai total R&D investment 2022 (yuan)": 22003461800.09,
+    "Shanghai adjusted growth rate (%)": 64.38,
+    "Shanghai estimated R&D investment 2025 (yuan)": 97732260069.03
+  },
+  "answer": [
+    "Shanghai",
+    97732260069.03
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/medium002_result.json b/assets/qa_raw/industry_planning/medium002_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..92c1ded24ff214c42ccfa1de59abf4d65f27da4c
--- /dev/null
+++ b/assets/qa_raw/industry_planning/medium002_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium002",
+  "question": "In 2022, assume that in the pharmaceutical manufacturing industry, the annual operating revenue growth rate of private enterprises is 5 percentage points higher than state-owned enterprises (including central and local state-owned enterprises) in the same province, while state-owned enterprises maintain their current growth rate unchanged (growth rate measured by the median year-on-year change in operating revenue of enterprises in the same province). By 2025, how many provinces will have private enterprise total revenue exceeding state-owned enterprise total revenue for the first time?",
+  "guidelines": "Answer format: integer (unit: count). If relevant data cannot be found, please answer \"No relevant data found\".",
+  "steps": [
+    "Filter from company_profile.csv all enterprise records with industry=\"pharmaceutical manufacturing industry\", extract enterprise name, bmCode, province, and ownership fields, finding 449 enterprises.",
+    "Divide enterprises into two groups by ownership: private enterprise group (ownership=\"private enterprise\") 346 enterprises, state-owned enterprise group (ownership=\"central state-owned enterprise\" or \"local state-owned enterprise\") 65 enterprises.",
+    "From company_operation_status.csv, link the above enterprises by bmCode, extract operating revenue amount and year-on-year operating revenue change rate fields, filter out enterprises with either field empty. Private enterprises: 331 valid records; state-owned enterprises: 64 valid records.",
+    "Group by province, calculate total operating revenue amount and median year-on-year operating revenue change rate for private and state-owned enterprises in each province respectively. Private enterprises cover 30 provinces, state-owned enterprises cover 24 provinces.",
+    "Filter provinces that have both private and state-owned enterprise data, totaling 24. Calculate adjusted growth rates: private enterprise growth = private median growth + 5; state-owned enterprise growth = state-owned median growth (unchanged).",
+    "For each valid province, project 2025 revenue: private 2025 revenue = private 2022 total × (1 + private adjusted growth/100)^3; state-owned 2025 revenue = state-owned 2022 total × (1 + state-owned growth/100)^3.",
+    "Filter provinces meeting the condition: private revenue lower than state-owned revenue in 2022, but estimated private revenue higher than state-owned revenue in 2025. Provinces meeting the condition: Heilongjiang.",
+    "Count provinces meeting the condition: 1."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "Found 449 pharmaceutical manufacturing enterprises in company_profile.csv, including 346 private enterprises and 65 state-owned enterprises.",
+    "Found 331 valid operating revenue records for private enterprises and 64 for state-owned enterprises in company_operation_status.csv.",
+    "24 provinces have both private and state-owned enterprise data."
+  ],
+  "milestone": {
+    "Pharmaceutical manufacturing private enterprises": 346,
+    "Pharmaceutical manufacturing state-owned enterprises": 65,
+    "Provinces with both enterprise types": 24,
+    "Provinces where private exceeded state-owned in 2025 after being lower in 2022": 1,
+    "Provinces meeting the condition": [
+      "Heilongjiang Province"
+    ]
+  },
+  "answer": 1,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/medium003_result.json b/assets/qa_raw/industry_planning/medium003_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8642e6a8f5d6be151d6cf7486d85c38de5a3f43
--- /dev/null
+++ b/assets/qa_raw/industry_planning/medium003_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium003",
+  "question": "Using 2022 as the base period, a differentiated policy incentive is proposed for the automobile manufacturing industry: For provinces that have implemented automobile industry policies containing the keywords \"new energy\" or \"electric\", the annual operating profit growth rate of automobile manufacturing enterprises within their jurisdiction will add 10 percentage points on top of the current median growth rate, creating a policy acceleration effect; other provinces are unaffected, and operating profit growth rate continues along the current trajectory. Under this differentiated scenario, using the median year-on-year change in operating profit as the baseline growth rate for each province, and projecting to 2025 via 3-year compound growth, which province has the most prominent increase in total operating profit compared to actual 2022 levels? What is the specific increase (increase = (2025 estimated value - 2022 actual value) / 2022 actual value × 100%)?",
+  "guidelines": "Answer format: value (2 decimal places, unit: %). If relevant data cannot be found, please answer \"No relevant data found\".",
+  "steps": [
+    "Filter from policy_release_status.csv policy records where industry field contains \"automobile\" and policy name contains \"new energy\" or \"electric\", obtaining 10 records. Extract the list of involved provinces (deduplicated, excluding national policies), obtaining 7 provinces: Hainan, Chongqing, Guangdong, Jiangxi, Sichuan, Shandong, Jiangsu.",
+    "Filter from company_profile.csv all enterprise records with industry=\"automobile manufacturing industry\", extract enterprise name, bmCode, and province fields, finding 230 enterprises.",
+    "From company_operation_status.csv, link the above enterprises by bmCode, extract operating profit amount and year-on-year operating profit change rate fields, filter out enterprises with either field empty, obtaining 230 valid enterprises.",
+    "Group by province, calculate total operating profit amount and median year-on-year operating profit change rate for each province, filter provinces with total operating profit amount greater than 0, totaling 22 provinces.",
+    "For each valid province, determine whether it belongs to provinces with policy: adjusted growth rate for provinces with policy = median growth rate + 10; adjusted growth rate for provinces without policy = median growth rate.",
+    "Calculate estimated total operating profit for 2025 for each province = total operating profit amount × (1 + adjusted growth rate/100)^3. Guangxi Zhuang Autonomous Region's estimated 2025 operating profit = 117492412.97 × (1+112.51/100)^3 = 1127581487.29 (yuan).",
+    "Calculate growth rate for each province = (2025 estimated value - 2022 actual value) / 2022 actual value × 100. Guangxi Zhuang Autonomous Region's growth rate = (1127581487.29 - 117492412.97) / 117492412.97 × 100 = 859.71%.",
+    "Sort by growth rate in descending order. The province with the largest growth rate is Guangxi Zhuang Autonomous Region, at 859.71%."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "Found 10 automobile-related policy records containing \"new energy\" or \"electric\" keywords in policy_release_status.csv, involving 7 provinces.",
+    "Found 230 automobile manufacturing enterprises in company_profile.csv.",
+    "Found operating profit amounts and year-on-year change rates for 230 automobile manufacturing enterprises in company_operation_status.csv."
+  ],
+  "milestone": {
+    "Provinces with policy": 7,
+    "Valid automobile manufacturing enterprises": 230,
+    "Provinces with total operating profit > 0": 22,
+    "Guangxi total operating profit 2022 (yuan)": 117492412.97,
+    "Guangxi adjusted growth rate (%)": 112.51,
+    "Guangxi estimated operating profit 2025 (yuan)": 1127581487.29,
+    "Guangxi growth rate (%)": 859.71
+  },
+  "answer": 859.71,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/medium004_result.json b/assets/qa_raw/industry_planning/medium004_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f7d1a28a10a3aff67a6da3549a8e79ec05bd3454
--- /dev/null
+++ b/assets/qa_raw/industry_planning/medium004_result.json
@@ -0,0 +1,33 @@
+{
+  "id": "medium004",
+  "question": "In 2022, in the communication transmission equipment industry, assuming that enterprises with R&D investment ratio below the national industry median will be eliminated from the market within 3 years, and only enterprises with R&D investment ratio not lower than the national median will be retained, what is the proportion for the province with the highest ratio of remaining enterprises' operating revenue after elimination to operating revenue before elimination?",
+  "guidelines": "Answer format: Value (2 decimal places, unit: %). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from national_industry_status.csv where industry=\"communication transmission equipment industry\", extract the R&D investment ratio median of 9.92 as the national benchmark.",
+    "Filter from company_profile.csv all enterprise records with industry=\"communication transmission equipment industry\", extract enterprise name, bmCode, and province fields, finding 120 enterprises.",
+    "From company_operation_status.csv, link the above enterprises by bmCode, extract operating revenue amount and R&D investment ratio fields, linking 120 enterprises.",
+    "Group by province, calculate total operating revenue amount for each province before elimination (all enterprises, including those with empty R&D investment ratio), totaling 19 provinces.",
+    "Filter enterprises with non-empty R&D investment ratio and not lower than the national median of 9.92 (enterprises with empty R&D investment ratio are deemed below median and automatically eliminated), group by province to calculate total operating revenue of remaining enterprises after elimination. 13 provinces have surviving enterprises after elimination.",
+    "For each valid province, calculate revenue retention ratio = total revenue after elimination / total revenue before elimination × 100. Anhui Province's revenue retention ratio = 2727186878.01 / 2727186878.01 × 100 = 100.00%.",
+    "Sort by revenue retention ratio in descending order. Anhui Province ranks highest with a ratio of 100.00%."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Obtained national R&D investment ratio median of 9.92 for communication transmission equipment industry from national_industry_status.csv.",
+    "Found 120 communication transmission equipment industry enterprises in company_profile.csv.",
+    "Linked operating revenue and R&D investment ratio data for 120 enterprises from company_operation_status.csv."
+  ],
+  "milestone": {
+    "National R&D investment ratio median": 9.92,
+    "Total communication transmission equipment industry enterprises": 120,
+    "Anhui Province total revenue before elimination (yuan)": 2727186878.01,
+    "Anhui Province total revenue after elimination (yuan)": 2727186878.01,
+    "Anhui Province revenue retention ratio (%)": 100.0
+  },
+  "answer": 100.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/medium005_result.json b/assets/qa_raw/industry_planning/medium005_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae3a0fd9eac8366e44cc064e8346b4dc3bb93bab
--- /dev/null
+++ b/assets/qa_raw/industry_planning/medium005_result.json
@@ -0,0 +1,38 @@
+{
+  "id": "medium005",
+  "question": "In 2022, regarding the strategic choice for Guangdong Province's consumer electronics and electrical industry, a policy consulting agency proposed two competing development paths: the first is the \"high-end transformation route\", evaluated by the average R&D investment ratio of private enterprises, invention patent density (= total annual Chinese invention patent grants ÷ total number of enterprises), and the number of relevant industrial policies in that province; the second is the \"export-oriented route\", evaluated by per capita revenue (= total operating revenue ÷ total number of employees), average asset turnover rate (= mean operating revenue ÷ mean total assets), and total number of enterprises. Both routes use inter-provincial peer comparison ranking scores (score = (N - ranking) / (N - 1) × 100), with equal weight across dimensions to calculate the route total score. What is the difference between Guangdong Province's total score on the \"high-end transformation route\" and the \"export-oriented route\" (former minus latter)?",
+  "guidelines": "Answer format: Value (2 decimal places). A positive number indicates the high-end transformation route has a higher score, a negative number indicates the export-oriented route has a higher score. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "From company_profile.csv, filter industry=\"consumer electronics and electrical industry\", join with company_operation_status.csv for year=2022 on bmCode; keep only provinces with enterprise count > 0 to obtain N=23 provinces for peer comparison within the same industry.",
+    "Within each province, filter ownership=\"private enterprise\" and take the arithmetic mean of non-missing R&D investment ratios; Guangdong Province is approximately 9.2610.",
+    "From regional_industry_status.csv, filter industry=\"consumer electronics and electrical industry\", and use for each province: total annual Chinese invention patent grants ÷ enterprise count, total operating revenue ÷ total employees, mean operating revenue ÷ mean total assets, and enterprise count; Guangdong: invention patent density 96.4800, per capita revenue about 1501516.49, asset turnover about 0.8845, enterprise count 150.",
+    "From policy_release_status.csv, filter policyClassification=\"local policy\", publishDate in calendar year 2022, and industry field containing the full industry name \"consumer electronics and electrical industry\"; records with empty province are excluded from province-level counts.",
+    "For all six indicators, rank provinces in descending order among N=23 (higher is better), use minimum rank for ties, and compute score = (N - ranking) / (N - 1) × 100 for each province. Guangdong: private R&D ratio about rank 5 → about 81.82; patent density about rank 3 → about 90.91; policy count about rank 2 → about 95.45; per capita revenue about rank 6 → about 77.27; asset turnover about rank 6 → about 77.27; enterprise count rank 1 → 100.00.",
+    "High-end transformation score = (81.82 + 90.91 + 95.45) / 3 ≈ 89.39; export-oriented score = (77.27 + 77.27 + 100.00) / 3 ≈ 84.85.",
+    "Score difference = high-end transformation score - export-oriented score ≈ 4.55 (rounded to two decimal places)."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Filtered 269 consumer electronics and electrical industry private enterprises from company_profile.csv.",
+    "Linked R&D investment ratio data for 252 private enterprises from company_operation_status.csv.",
+    "Found regional data for the consumer electronics and electrical industry in 23 provinces from regional_industry_status.csv.",
+    "Filtered 6 consumer electronics and electrical industry policy records from policy_release_status.csv, involving 5 provinces."
+  ],
+  "milestone": {
+    "Private enterprise average R&D investment ratio score": 81.82,
+    "Invention patent density score": 90.91,
+    "Policy count score": 95.45,
+    "Per capita revenue score": 77.27,
+    "Asset turnover rate score": 77.27,
+    "Enterprise count score": 100.0,
+    "High-end transformation score": 89.39,
+    "Export-oriented score": 84.85,
+    "Score difference": 4.55
+  },
+  "answer": 4.55,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/medium006_result.json b/assets/qa_raw/industry_planning/medium006_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..de3b70ac002a7c8f87098c6f13b3c4df5aba99c9
--- /dev/null
+++ b/assets/qa_raw/industry_planning/medium006_result.json
@@ -0,0 +1,37 @@
+{
+  "id": "medium006",
+  "question": "In 2022, regarding the industrial development direction of Hebei Province's metal smelting and rolling processing industry, researchers intend to compare two alternative transformation paths through a multi-dimensional scoring method. Path one \"green and low-carbon route\" covers three evaluation indicators: average enterprise R&D investment amount (reflecting technology upgrade willingness), invention patent density (= total cumulative Chinese invention patent grants ÷ total number of enterprises, weight 0.3), and count of provincial green-related policies (i.e., policy records whose name contains \"green\", \"low-carbon\", or \"energy-saving\", weight 0.4), with average R&D investment amount weight 0.3; Path two \"traditional capacity expansion route\" also includes three indicators: total assets (weight 0.4), total operating revenue (weight 0.3), and total number of enterprises (weight 0.3). Each province's score for each indicator is calculated by inter-provincial peer ranking (score = (N - ranking) / (N - 1) × 100). Please calculate the score difference between Hebei Province on the above two routes (green and low-carbon route score minus traditional capacity expansion route score).",
+  "guidelines": "Answer format: Value (2 decimal places). A positive number indicates the green and low-carbon route has a higher score, a negative number indicates the traditional capacity expansion route has a higher score. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "From regional_industry_status.csv, filter industry = \"金属冶炼和压延加工业\" (metal smelting and rolling processing industry), and extract for each province: total number of enterprises, mean R&D investment amount (研发投入金额均值), total cumulative Chinese invention patent grants, total assets, and total operating revenue. After dropping nulls for mean R&D, patent density, total assets, and operating revenue, there are 16 valid provinces for each of these four indicators.",
+    "Average enterprise R&D investment uses the table field mean R&D investment amount (non-null values participate in ranking). Hebei Province: 3062942709.59 yuan, rank 2 among 16 provinces (ties take the best rank), score = (16 - 2) / (16 - 1) × 100 = 93.33.",
+    "Invention patent density = total cumulative Chinese invention patent grants ÷ total number of enterprises (enterprises > 0 and both fields non-null). Hebei Province: 793 ÷ 1 = 793.0000, rank 2 among 16, score = 93.33.",
+    "From policy_release_status.csv, filter records with publishDate in calendar year 2022 (parsed as day/month/year), policyClassification = \"地方政策\", province non-empty and not \"全国\", industry field containing \"金属冶炼和压延加工业\", and policy name containing at least one of \"绿色\", \"低碳\", or \"节能\"; count by province. Eight records matched, across six provinces; Hebei Province: 0 records.",
+    "Provincial green-related policy count: count on the 34 provinces that appear in the industry table (provinces with no match are 0), rank within this indicator; Hebei has 0, tied with 27 other provinces with competitive rank 7, score = (34 - 7) / (34 - 1) × 100 = 81.82.",
+    "Traditional capacity expansion: total assets and operating revenue are ranked among the 16 provinces with non-null values; Hebei ranks 6th and 11th, scores 66.67 and 33.33 respectively. Total number of enterprises is ranked among all 34 provinces; Hebei has 1 enterprise, rank 24 (tied with six provinces), score = 30.30.",
+    "Green and low-carbon route score = 93.33 × 0.3 + 93.33 × 0.3 + 81.82 × 0.4 = 88.73; traditional capacity expansion route score = 66.67 × 0.4 + 33.33 × 0.3 + 30.30 × 0.3 = 45.76.",
+    "Score difference (green and low-carbon − traditional capacity expansion) = 88.73 − 45.76 = 42.97."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "Filtered regional_industry_status.csv for industry \"metal smelting and rolling processing industry\" and obtained valid provincial-level data for 16 regions.",
+    "Filtered policy_release_status.csv and found 8 green-related policy records involving six provinces."
+  ],
+  "milestone": {
+    "Average enterprise R&D investment score": 93.33,
+    "Invention patent density score": 93.33,
+    "Green policy count score": 81.82,
+    "Total assets score": 66.67,
+    "Operating revenue score": 33.33,
+    "Enterprise count score": 30.3,
+    "Green and low-carbon route score": 88.73,
+    "Traditional capacity expansion route score": 45.76,
+    "Score difference": 42.97
+  },
+  "answer": 42.97,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/medium007_result.json b/assets/qa_raw/industry_planning/medium007_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..971acbdfd5c0f520432d1e9ac026a15456207d6a
--- /dev/null
+++ b/assets/qa_raw/industry_planning/medium007_result.json
@@ -0,0 +1,42 @@
+{
+  "id": "medium007",
+  "question": "In the 2022 data for the chemical raw materials and chemical products manufacturing industry, first identify the enterprise with the largest total assets in the entire industry and determine its registration province; then take that province as the research object, and use the four-indicator equal-weight scoring method (each indicator score = (N - inter-provincial ranking) / (N - 1) × 100) to compare two industrial strategy routes: the \"R&D-driven route\" comprehensively evaluates four indicators: total private enterprise R&D investment, total state-owned enterprise R&D investment (central state-owned + local state-owned + other state-owned enterprises + state-owned enterprises (research institutes)), regional R&D intensity (mean R&D investment ratio), and count of relevant R&D policies (policy name contains \"R&D\", \"innovation\", or \"technology/science\"); the \"scale expansion route\" comprehensively evaluates four indicators: total assets, total operating revenue, total number of enterprises, and total government subsidies. Which route has the higher comprehensive score for that province?",
+  "guidelines": "Answer format: \"R&D-driven route\" or \"Scale expansion route\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from company_profile.csv enterprises with industry=\"chemical raw materials and chemical products manufacturing industry\", extract enterprise name, bmCode, province, and ownership fields, finding 364 enterprises.",
+    "From company_operation_status.csv, link enterprises by bmCode and restrict year=2022, extract the total assets field, filter out empty values and sort in descending order. The enterprise with the highest total assets is Hengyi Changhua Fine Chemical Company (total assets 200843104094.19 yuan), located in Shandong Province.",
+    "From company_operation_status.csv, extract the R&D investment amount field for enterprises in this industry, filter out empty values, then aggregate by province and ownership; Shandong Province private enterprise total R&D investment is 6239314301.75 yuan, and state-owned enterprises (central state-owned + local state-owned + other state-owned enterprises + state-owned enterprises (research institutes)) total R&D investment is 6670389432.99 yuan.",
+    "Filter from regional_industry_status.csv where industry=\"chemical raw materials and chemical products manufacturing industry\" and province is non-empty and not \"national aggregate\", yielding 34 province rows; extract mean R&D investment ratio, total assets, total operating revenue, enterprise count, and total government reward and subsidy per province. For each of the eight indicators, ranking uses only provinces with non-missing values for that field; effective province count N is computed separately per indicator (e.g. R&D intensity N=16, enterprise count N=34).",
+    "Filter from policy_release_status.csv records whose policy name contains \"R&D\" or \"innovation\" or \"technology/science\" and whose province is non-empty and not \"national aggregate\", group by province to count; Shandong has 7 records (93 policy records in the full database satisfying this rule).",
+    "Rank each of the eight indicators in descending order by value (ties share the same rank and subsequent ranks skip accordingly), and compute Shandong's score per indicator = (N - rank) / (N - 1) × 100 (if N < 2, that indicator score is 100). Private R&D score = 96.30, state-owned R&D score = 100.00, R&D intensity score = 40.00, policy count score = 91.67, total assets score = 100.00, operating revenue score = 100.00, enterprise count score = 93.94, total subsidy score = 93.33.",
+    "R&D-driven score = (96.30 + 100.00 + 40.00 + 91.67) × 0.25 = 81.99; scale expansion score = (100.00 + 100.00 + 93.94 + 93.33) × 0.25 = 96.82.",
+    "R&D-driven score 81.99 < scale expansion score 96.82, therefore the scale expansion route has the higher score."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "Filtered 364 chemical raw materials and chemical products manufacturing industry enterprises from company_profile.csv.",
+    "Found in company_operation_status.csv that the enterprise with the highest total assets is Hengyi Changhua Fine Chemical Company, located in Shandong Province.",
+    "From regional_industry_status.csv, obtained 34 province rows under the industry and province rules; each regional indicator independently determines N from non-empty provinces and is ranked separately.",
+    "From policy_release_status.csv, counted by policy-name keywords and province rules: Shandong has 7 R&D-related policies; 93 qualifying records in the full database."
+  ],
+  "milestone": {
+    "Enterprise with highest total assets": "Hengyi Changhua Fine Chemical Company",
+    "Target province": "Shandong Province",
+    "Private R&D investment score": 96.3,
+    "State-owned R&D investment score": 100.0,
+    "R&D intensity score": 40.0,
+    "R&D policy score": 91.67,
+    "Total assets score": 100.0,
+    "Operating revenue score": 100.0,
+    "Enterprise count score": 93.94,
+    "Total subsidy score": 93.33,
+    "R&D-driven route score": 81.99,
+    "Scale expansion route score": 96.82
+  },
+  "answer": "Scale expansion route",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/medium008_result.json b/assets/qa_raw/industry_planning/medium008_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..12c18cad30cd144c92c818179ef7d3041675c77d
--- /dev/null
+++ b/assets/qa_raw/industry_planning/medium008_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium008",
+  "question": "In 2022, in the food and beverage industry, for the province where the enterprise with the most cumulative Chinese invention patent grants is located, if that province chooses the \"brand upgrade route\" (evaluating market-cap-to-revenue ratio, profit margin, per capita market cap, with weights of 35%, 35%, 30% respectively) versus the \"industrial chain extension route\" (evaluating total number of enterprises, revenue scale, upstream-downstream enterprise diversity, with weights of 40%, 30%, 30% respectively), which route has the higher score?",
+  "guidelines": "Answer format: \"Brand upgrade route\" or \"Industrial chain extension route\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Step 1: Filter from company_profile.csv enterprises with industry=\"food and beverage industry\", extract enterprise name, bmCode, and province fields, finding 247 enterprises.",
+    "Step 2: From company_operation_status.csv, link enterprises by bmCode, extract cumulative Chinese invention patent grants field, filter out empty values, sort in descending order. The enterprise with the most cumulative Chinese invention patent grants is Qingqing Jinyin Food Company (644 grants), located in Beijing.",
+    "Step 3: Filter from regional_industry_status.csv where industry=\"food and beverage industry\", extract total company market cap, total operating revenue, total operating profit, total number of employees, and total number of enterprises for each province, totaling 30 provinces.",
+    "Step 4: Calculate each province's market-cap-to-revenue ratio = total company market cap / total operating revenue, Beijing: 0.0000; profit margin = total operating profit / total operating revenue, Beijing: 0.0748; per capita market cap = total company market cap / total number of employees, Beijing: 0.02.",
+    "Step 5: From company_profile.csv, count the number of ownership types (distinct count) for food and beverage industry enterprises by province. Beijing: 4 types.",
+    "Step 6: Sort the six indicators in descending order respectively, calculate Beijing's ranking score for each indicator = (N - ranking) / (N - 1) × 100. Market-cap-to-revenue ratio score = 26.67, profit margin score = 33.33, per capita market cap score = 13.33, enterprise count score = 75.86, revenue scale score = 82.76, diversity score = 93.10.",
+    "Step 7: Calculate brand upgrade score = 26.67 × 0.35 + 33.33 × 0.35 + 13.33 × 0.3 = 25.00; industrial chain extension score = 75.86 × 0.4 + 82.76 × 0.3 + 93.10 × 0.3 = 83.10.",
+    "Step 8: Brand upgrade score 25.00 < industrial chain extension score 83.10, therefore the industrial chain extension route has the higher score."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "Filtered 247 food and beverage industry enterprises from company_profile.csv.",
+    "Found the enterprise with the most cumulative Chinese invention patent grants in company_operation_status.csv: Qingqing Jinyin Food Company, located in Beijing.",
+    "Found regional data for food and beverage industry in 30 provinces from regional_industry_status.csv.",
+    "Counted ownership type diversity for food and beverage industry enterprises by province from company_profile.csv."
+  ],
+  "milestone": {
+    "Enterprise with most patents": "Qingqing Jinyin Food Company",
+    "Target province": "Beijing",
+    "Market-cap-to-revenue ratio score": 26.67,
+    "Profit margin score": 33.33,
+    "Per capita market cap score": 13.33,
+    "Enterprise count score": 75.86,
+    "Revenue scale score": 82.76,
+    "Diversity score": 93.1,
+    "Brand upgrade route score": 25.0,
+    "Industrial chain extension route score": 83.1
+  },
+  "answer": "Industrial chain extension route",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/medium009_result.json b/assets/qa_raw/industry_planning/medium009_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..90ede54ee49ec0caa5634655c52fb5f23b300183
--- /dev/null
+++ b/assets/qa_raw/industry_planning/medium009_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium009",
+  "question": "Among the enterprises in the textile, footwear and apparel industry in 2022, find the enterprise with the largest R&D personnel scale; its province is the analysis object. For that province's textile, footwear and apparel industry, use the following two route scoring systems to determine which development route has the advantage—the \"automation upgrade route\" scores by three indicators: R&D investment intensity (= total R&D investment amount / total operating revenue, weight 0.4), capitalized R&D investment ratio (= total capitalized R&D investment / total R&D investment amount, weight 0.3), and count of equipment manufacturing policies (policy name contains \"equipment\" or \"intelligent manufacturing\", weight 0.3); the \"brand overseas expansion route\" scores by cumulative PCT patent applications (weight 0.4), per capita revenue (weight 0.3), and count of export-related policies (policy name contains \"export\", \"foreign trade\", or \"international\", weight 0.3); both routes use inter-provincial ranking scores for each indicator (score = (N - ranking) / (N - 1) × 100). Which route has the higher score for that province?",
+  "guidelines": "Answer format: \"Automation upgrade route\" or \"Brand overseas expansion route\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from company_profile.csv enterprises with industry=\"textile, footwear and apparel industry\", extract enterprise name, bmCode, and province fields, finding 177 enterprises.",
+    "From company_operation_status.csv, link enterprises by bmCode, filter year=2022, extract R&D personnel count, filter out empty values, sort in descending order. The enterprise with the most R&D personnel is Anbu Shangchang Brand Company (2,826 people), located in Guangdong Province.",
+    "Filter from regional_industry_status.csv where industry=\"textile, footwear and apparel industry\", extract for each province total R&D investment amount, total operating revenue, total capitalized R&D investment, total cumulative PCT patent applications, and total number of employees, for 34 provincial-level regions.",
+    "Calculate each province's R&D investment intensity = total R&D investment amount / total operating revenue, Guangdong: 0.017511; capitalized R&D investment ratio = total capitalized R&D investment / total R&D investment amount, Guangdong: 0.004225; per capita revenue = total operating revenue / total number of employees, Guangdong: 269886.81. (Provinces with null values or a zero denominator are excluded from ranking for that indicator.)",
+    "From policy_release_status.csv, count by policy name: equipment manufacturing policies whose name contains \"equipment\" or \"intelligent manufacturing\"; export-related policies whose name contains \"export\", \"foreign trade\", or \"international\". Policies with province=\"全国\" count once for each of the 34 provincial samples; other policies are assigned to the corresponding province.",
+    "For each of the six indicators, rank provinces in descending order within the valid set for that indicator, with ties taking the best rank; Guangdong's ranking score = (N - ranking) / (N - 1) × 100. R&D intensity score = 30.77, capitalized R&D ratio score = 91.67, equipment policy score = 90.91, PCT patent score = 86.67, per capita revenue score = 0.00, export policy score = 90.91.",
+    "Calculate automation upgrade score = 30.77 × 0.4 + 91.67 × 0.3 + 90.91 × 0.3 = 67.08; brand overseas expansion score = 86.67 × 0.4 + 0.00 × 0.3 + 90.91 × 0.3 = 61.94.",
+    "Automation upgrade score 67.08 > brand overseas expansion score 61.94, therefore the automation upgrade route has the higher score."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "Filtered 177 textile, footwear and apparel industry enterprises from company_profile.csv.",
+    "Found the enterprise with the most R&D personnel in company_operation_status.csv: Anbu Shangchang Brand Company (2,826 people), located in Guangdong Province.",
+    "Filtered regional data for textile, footwear and apparel industry across 34 provincial-level regions from regional_industry_status.csv.",
+    "Counted equipment manufacturing and export-related policies from policy_release_status.csv by policy name keywords; nationwide policies are applied to all 34 sample provinces."
+  ],
+  "milestone": {
+    "Enterprise with most R&D personnel": "Anbu Shangchang Brand Company",
+    "Target province": "Guangdong Province",
+    "R&D investment intensity score": 30.77,
+    "Capitalized R&D ratio score": 91.67,
+    "Equipment policy score": 90.91,
+    "PCT patent score": 86.67,
+    "Per capita revenue score": 0.0,
+    "Export policy score": 90.91,
+    "Automation upgrade route score": 67.08,
+    "Brand overseas expansion route score": 61.94
+  },
+  "answer": "Automation upgrade route",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/medium010_result.json b/assets/qa_raw/industry_planning/medium010_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..72a1108cccffd564f422d7accf8810194708d9ca
--- /dev/null
+++ b/assets/qa_raw/industry_planning/medium010_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium010",
+  "question": "Sort the semiconductor industry in 2022 by operating revenue from high to low, identify the top 10 enterprises by revenue scale, and count the provinces to which these leading enterprises belong; the province with the highest frequency is the research target. Next, conduct a national horizontal comparison of that province based on four industrial competitiveness dimensions—dimension one is enterprise agglomeration (total number of semiconductor industry enterprises per province), dimension two is innovation activity (= total annual Chinese invention patent grants ÷ total number of enterprises), dimension three is policy support (= count of relevant policies whose name contains \"semiconductor\" or \"integrated circuit\"), dimension four is industry scale (= total operating revenue). Among the national provincial rankings for each dimension, in how many dimensions did that province rank in the top 3?",
+  "guidelines": "Answer format: Integer. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from company_profile.csv enterprises with industry=\"semiconductor industry\", extract enterprise name, bmCode, and province fields, finding 172 enterprises.",
+    "From company_operation_status.csv, link enterprises by bmCode, extract operating revenue amount field; 172 enterprises after filtering empty values. Sort in descending order and take the top 10 enterprises.",
+    "Count provincial distribution of top 10 enterprises: Shanghai (4), Taiwan (3), Guangdong (2), Jiangsu (1). Shanghai has the highest frequency with 4 enterprises.",
+    "Filter from regional_industry_status.csv where industry=\"semiconductor industry\", extract total number of enterprises, total annual Chinese invention patent grants, and total operating revenue for each province, totaling 22 provinces.",
+    "Calculate each province's innovation activity = total annual Chinese invention patent grants / total number of enterprises (exclude provinces with 0 enterprises). Shanghai's innovation activity = 35.1852.",
+    "Filter from policy_release_status.csv policies whose name contains \"semiconductor\" or \"integrated circuit\", 4 records total. Group by province to count relevant policy count. Shanghai has 1 relevant policy.",
+    "Sort the four indicators in descending order respectively. Shanghai's rankings: enterprise agglomeration 2nd (27 enterprises), innovation activity 4th (35.1852), policy support 2nd (1 record), industry scale 2nd (247438863786.89 yuan).",
+    "Count the number of dimensions in which Shanghai ranks ≤ 3 among the four dimensions. 3 dimensions rank in the national top 3."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "Found 172 semiconductor industry enterprises in company_profile.csv.",
+    "Found operating revenue data for 172 semiconductor industry enterprises in company_operation_status.csv.",
+    "Found semiconductor industry statistics for 22 provinces in regional_industry_status.csv.",
+    "Found 4 policy records containing \"semiconductor\" or \"integrated circuit\" in policy_release_status.csv."
+  ],
+  "milestone": {
+    "Province with most top 10 enterprises by revenue": "Shanghai",
+    "Shanghai enterprise agglomeration (enterprise count)": 27,
+    "Shanghai enterprise agglomeration ranking": 2,
+    "Shanghai innovation activity": 35.1852,
+    "Shanghai innovation activity ranking": 4,
+    "Shanghai policy support (policy count)": 1,
+    "Shanghai policy support ranking": 2,
+    "Shanghai industry scale (total operating revenue)": 247438863786.89,
+    "Shanghai industry scale ranking": 2,
+    "Number of dimensions in top 3": 3
+  },
+  "answer": 3,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/medium011_result.json b/assets/qa_raw/industry_planning/medium011_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2689b19ba60b77954ed24283cfed95a99c81d0ad
--- /dev/null
+++ b/assets/qa_raw/industry_planning/medium011_result.json
@@ -0,0 +1,42 @@
+{
+  "id": "medium011",
+  "question": "In the 2022 pharmaceutical manufacturing industry data, find the listed enterprise with the highest total Chinese patent grants in that year; after identifying that enterprise's province, construct a four-dimensional evaluation framework around \"innovation ecosystem\": R&D intensity (= total R&D investment amount ÷ total operating revenue), patent conversion efficiency (= total cumulative Chinese invention patent grants ÷ total cumulative Chinese invention patent applications), industry scale (= total operating revenue), and policy support (= count of relevant policies whose name contains \"pharmaceutical\" or \"biotechnology\"). For each dimension, identify the top 5 provinces nationwide and calculate their mean. Is that province's overall innovation ecosystem level above or below the average of the national top 5 across these four dimensions?",
+  "guidelines": "Answer format: \"Above average\" or \"Below average\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from company_profile.csv enterprises with industry=\"pharmaceutical manufacturing industry\", extract enterprise name, bmCode, and province fields, finding 449 enterprises.",
+    "From company_operation_status.csv, link enterprises by bmCode, extract annual Chinese patent grants field; after filtering empty values, sort in descending order. The enterprise with the most annual Chinese patent grants is Puge Ruijian Biopharmaceutical Company (337 grants), located in Henan Province.",
+    "Filter from regional_industry_status.csv where industry=\"pharmaceutical manufacturing industry\", extract total R&D investment amount, total operating revenue, total cumulative Chinese invention patent grants, and total cumulative Chinese invention patent applications for each province, totaling 34 provinces.",
+    "Calculate each province's R&D intensity = total R&D investment amount / total operating revenue, patent conversion efficiency = total cumulative Chinese invention patent grants / total cumulative Chinese invention patent applications (exclude provinces with zero denominator), 15 valid provinces.",
+    "Filter from policy_release_status.csv policies whose name contains \"pharmaceutical\" or \"biotechnology\", 21 records total. Group by province to count relevant policy count.",
+    "Sort the four indicators in descending order respectively, take the top 5 provinces for each, and calculate averages: R&D intensity top 5 average = 0.182465, patent conversion efficiency top 5 average = 0.522743, industry scale top 5 average = 237558091257.40, policy support top 5 average = 1.80.",
+    "Henan Province's values for the four indicators: R&D intensity = 0.075724, patent conversion efficiency = 0.311526, industry scale = 18878671803.01, policy support = 2. Compared with top 5 averages, 1 dimension is above average.",
+    "Comprehensive score 1 < 2, output \"Below average\"."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "Found 449 pharmaceutical manufacturing industry enterprises in company_profile.csv.",
+    "Found annual Chinese patent grants data for 368 pharmaceutical manufacturing industry enterprises in company_operation_status.csv.",
+    "Found pharmaceutical manufacturing industry statistics for 15 provinces in regional_industry_status.csv.",
+    "Found 21 policy records containing \"pharmaceutical\" or \"biotechnology\" in policy_release_status.csv."
+  ],
+  "milestone": {
+    "Enterprise with most annual Chinese patent grants": "Puge Ruijian Biopharmaceutical Company",
+    "Province of that enterprise": "Henan Province",
+    "Annual Chinese patent grants": 337,
+    "Henan Province R&D intensity": 0.075724,
+    "R&D intensity top 5 average": 0.182465,
+    "Henan Province patent conversion efficiency": 0.311526,
+    "Patent conversion efficiency top 5 average": 0.522743,
+    "Henan Province industry scale": 18878671803.01,
+    "Industry scale top 5 average": 237558091257.4,
+    "Henan Province policy support": 2,
+    "Policy support top 5 average": 1.8,
+    "Number of dimensions above average": 1
+  },
+  "answer": "Below average",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/medium012_result.json b/assets/qa_raw/industry_planning/medium012_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..1c02544d7e7139f8c1ac843ec79d355b41998db4
--- /dev/null
+++ b/assets/qa_raw/industry_planning/medium012_result.json
@@ -0,0 +1,39 @@
+{
+  "id": "medium012",
+  "question": "In 2022, in the railway, ship, aerospace and other transport equipment manufacturing industry, take the province with the highest concentration of the top 5 enterprises by total assets as the research object. What is that province's comprehensive ranking (ranking based on the arithmetic mean of the three dimension rankings) across three high-end manufacturing dimensions: technology intensity (technology intensity = total R&D personnel / total employees), capital intensity (capital intensity = total assets / total employees), and policy concentration (policy concentration = count of relevant policies)?",
+  "guidelines": "Answer format: Integer (unit: rank). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Filter from company_profile.csv enterprises with industry=\"railway, ship, aerospace and other transport equipment manufacturing industry\", extract enterprise name, bmCode, and province fields, finding 99 enterprises.",
+    "From company_operation_status.csv, link enterprises by bmCode, extract total assets field; 99 enterprises after filtering empty values. Sort in descending order and take the top 5 enterprises.",
+    "Count provincial distribution of top 5 enterprises: Beijing (4), Shanghai (1). Beijing has the highest concentration with 4 enterprises.",
+    "Filter from regional_industry_status.csv where industry=\"railway, ship, aerospace and other transport equipment manufacturing industry\", extract total R&D personnel, total employees, and total assets for each province, totaling 34 provinces.",
+    "Calculate each province's technology intensity = total R&D personnel / total employees, capital intensity = total assets / total employees (exclude provinces with 0 employees), 14 valid provinces.",
+    "Filter from policy_release_status.csv policies whose involved industry contains \"aviation\" or \"aerospace\" or \"ship\", 46 records total. Group by province to count relevant policy count (excluding \"national\" level).",
+    "Sort the three indicators in descending order respectively. Beijing's rankings: technology intensity 4th, capital intensity 4th, policy concentration 9th. Average ranking = 5.67.",
+    "Sort all provinces by average ranking in ascending order. Beijing's comprehensive ranking is 4th."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "Found 99 railway, ship, aerospace and other transport equipment manufacturing industry enterprises in company_profile.csv.",
+    "Found total assets data for 99 enterprises in company_operation_status.csv.",
+    "Found industry statistics for 14 provinces in regional_industry_status.csv.",
+    "Found 46 policy records whose involved industry contains \"aviation\" or \"aerospace\" or \"ship\" in policy_release_status.csv."
+  ],
+  "milestone": {
+    "Province with most top 5 enterprises by total assets": "Beijing",
+    "Beijing technology intensity": 0.210283,
+    "Beijing technology intensity ranking": 4,
+    "Beijing capital intensity": 3488225.06,
+    "Beijing capital intensity ranking": 4,
+    "Beijing policy concentration (policy count)": 0,
+    "Beijing policy concentration ranking": 9,
+    "Beijing average ranking": 5.67,
+    "Comprehensive ranking": 4
+  },
+  "answer": 4,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/medium013_result.json b/assets/qa_raw/industry_planning/medium013_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca0b33a36b7f2b640ae62c6e5e46397af7260580
--- /dev/null
+++ b/assets/qa_raw/industry_planning/medium013_result.json
@@ -0,0 +1,46 @@
+{
+  "id": "medium013",
+  "question": "In the 2022 automotive manufacturing industry, take the top 10 provinces by profitability (measured by net profit) as the candidate set, and construct a three-dimensional comprehensive scoring model to identify the province with the best industrial development quality: Dimension one \"industrial chain completeness\" (weight 0.3) comprehensively assesses the number of enterprise ownership types (ownership diversity) and the interquartile range of total assets of enterprises within the province (scale diversity); dimension two \"technological capability\" (weight 0.4) comprehensively assesses mean R&D investment ratio (R&D intensity) and total annual Chinese invention patent grants divided by total number of enterprises (patent density); dimension three \"market performance\" (weight 0.3) comprehensively assesses total operating revenue (revenue scale) and total operating profit divided by total operating revenue (profit margin). Each dimension's score is represented by the average ranking of its sub-indicators among the candidate provinces (comprehensive score = weighted average of each dimension's mean ranking). Which province has the highest comprehensive score (i.e., the best overall performance)?",
+  "guidelines": "Answer format: Province name. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "From regional_industry_status.csv, filter records with industry=\"automotive manufacturing industry\"; for each province, after excluding empty values in the total net profit field, sort in descending order by that field and take the top 10 provinces industry-wide as the candidate set.",
+    "From company_profile.csv, filter enterprises with industry=\"automotive manufacturing industry\", extract bmCode and province; join company_operation_status.csv on bmCode with year=2022 and extract total assets; for each province, after excluding empty total assets, compute the interquartile range Q3-Q1 as scale diversity.",
+    "For candidate provinces, on the corresponding rows in regional_industry_status.csv compute ownership diversity: among the eight enterprise-count fields for Sino-foreign joint ventures, central state-owned, state-owned (other), state-owned (research institutes), local state-owned, foreign-funded, private, and collective, count how many types have a value greater than zero.",
+    "From regional_industry_status.csv candidate rows, extract mean R&D investment ratio (R&D intensity), total annual Chinese invention patent grants, total number of enterprises, total operating revenue, and total operating profit; patent density = total annual Chinese invention patent grants / total number of enterprises, profit margin = total operating profit / total operating revenue.",
+    "For the six sub-indicators (ownership diversity, total-assets IQR, R&D intensity, patent density, total operating revenue, profit margin), rank candidate provinces separately for each indicator using only provinces with valid values for that indicator, in descending order by value; tied values share the same rank and the next rank is skipped; larger values rank higher (better).",
+    "Industrial chain completeness dimension score = (ownership diversity rank + scale diversity rank) / 2, technological capability dimension score = (R&D intensity rank + patent density rank) / 2, market performance dimension score = (revenue scale rank + profit margin rank) / 2 (if only one sub-indicator is valid in a dimension, use only that rank in the average).",
+    "Comprehensive score = industrial chain completeness dimension score × 0.3 + technological capability dimension score × 0.4 + market performance dimension score × 0.3; this is a weighted average of mean ranks, so a smaller value indicates better overall performance.",
+    "Guangdong Province: industrial chain completeness dimension score 2.5, technological capability 2.0, market performance 4.0; comprehensive score = 0.3 × 2.5 + 0.4 × 2 + 0.3 × 4 = 2.75, the smallest among the 10 candidate provinces, hence the best overall performance (rank 1)."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "From regional_industry_status.csv, 34 province-level rows for the automotive manufacturing industry; sorted by valid total net profit descending, the top 10 provinces form the candidate set.",
+    "From company_profile.csv, 230 enterprises filtered as automotive manufacturing industry, for joining with 2022 operating data.",
+    "From company_operation_status.csv, 230 records with year=2022 and bmCode among the above automotive manufacturing enterprises, used to compute each province's total-assets interquartile range."
+  ],
+  "milestone": {
+    "Number of candidate provinces": 10,
+    "Candidate province list": [
+      "Guangdong",
+      "Hebei",
+      "Shandong",
+      "Zhejiang",
+      "Beijing",
+      "Jiangsu",
+      "Shanghai",
+      "Jilin",
+      "Henan",
+      "Sichuan"
+    ],
+    "Guangdong industrial chain completeness ranking": 2.5,
+    "Guangdong technological capability ranking": 2.0,
+    "Guangdong market performance ranking": 4.0,
+    "Guangdong comprehensive score": 2.75
+  },
+  "answer": "Guangdong Province",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/industry_planning/medium014_result.json b/assets/qa_raw/industry_planning/medium014_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..1dc2352c3f4839a0579204c0cb043eeca4340898
--- /dev/null
+++ b/assets/qa_raw/industry_planning/medium014_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium014",
+  "question": "In the 2022 general equipment manufacturing industry provincial data, find the province with the highest total government reward and subsidy amount; after identifying that province, conduct a comprehensive rating of its industrial competitiveness across two strategic dimensions: the \"industrial upgrade capability\" dimension combines the inter-provincial mean ranking of three sub-indicators—mean year-on-year change in R&D investment (R&D investment growth rate), year-on-year growth rate of annual Chinese patent applications (patent application growth rate), and mean R&D personnel ratio (high-end talent ratio); the \"industrial foundation\" dimension combines the inter-provincial mean ranking of three sub-indicators—total number of enterprises (enterprise scale), total operating revenue (revenue scale), and the ratio of total operating revenue to total government reward and subsidy (subsidy efficiency). The overall comprehensive performance ranking is determined by the average of the two dimension rankings. Can that province's comprehensive industrial performance rank among the national top 5?",
+  "guidelines": "Answer format: \"Yes\" or \"No\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "From regional_industry_status.csv, filter industry=\"general equipment manufacturing industry\", extract each province's government reward funds and subsidy total fields; after dropping nulls and keeping values >0, 14 provinces remain.",
+    "Sort by subsidy total descending; the province with the highest total government subsidy is Shanghai (2466523519.29 yuan).",
+    "From the same table, extract each province's mean year-on-year change in R&D investment, mean R&D personnel share, total number of enterprises, and total operating revenue;",
+    "Compute each province's subsidy efficiency = total operating revenue / combined government reward and subsidy (exclude zero or missing denominators); 14 provinces have computable subsidy efficiency.",
+    "For the six sub-indicators (R&D investment growth rate, patent application YoY growth rate, R&D personnel share, enterprise count, revenue scale, subsidy efficiency), rank provinces separately within each indicator's valid sample from high to low; rank 1 is best; nulls are excluded from that indicator's ranking; ties share the same rank and subsequent ranks are skipped.",
+    "Industrial upgrade capability mean of three sub-indicator ranks = (R&D growth rank + patent application growth rank + R&D personnel share rank) / 3; for Shanghai (15+1+11)/3 = 9.",
+    "Industrial foundation mean of three sub-indicator ranks = (enterprise scale rank + revenue scale rank + subsidy efficiency rank) / 3; for Shanghai (5+1+7)/3 = 4.33 (two decimal places).",
+    "Rank each province's \"industrial upgrade capability three-rank mean\" and \"industrial foundation three-rank mean\" separately in ascending order (smaller mean is better) to obtain the two dimension ranks; overall comprehensive score = (industrial upgrade dimension rank + industrial foundation dimension rank) / 2; for Shanghai (13+4)/2 = 8.5.",
+    "Among provinces that have both dimension ranks (all six sub-indicators yield both dimensional composites), sort by overall score ascending; 13 provinces qualify; Shanghai ranks 10th (tied with Henan on score), which is >5, output \"No\"."
+  ],
+  "steps_num": 9,
+  "evidence": [
+    "From regional_industry_status.csv, 14 provinces have valid general equipment manufacturing government subsidy data (non-null and total >0).",
+    "From regional_industry_status.csv, 14 provinces have computable subsidy efficiency (denominator >0).",
+    "Year-on-year growth of patent applications is aggregated from company_profile.csv and company_operation_status.csv for 2021 and 2022;"
+  ],
+  "milestone": {
+    "Province with highest government subsidy": "Shanghai",
+    "Shanghai total government subsidy (yuan)": 2466523519.29,
+    "Shanghai industrial upgrade capability ranking": 9,
+    "Shanghai industrial foundation ranking": 4.33,
+    "Shanghai comprehensive performance ranking value": 8.5,
+    "Shanghai comprehensive ranking": 10
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "industry_planning"
+  }
+}
\ No newline at end of file
diff --git a/assets/qa_raw/international_comparison/hard001_result.json b/assets/qa_raw/international_comparison/hard001_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a07e89a939bcada401872457a4af205daf7d1c3a
--- /dev/null
+++ b/assets/qa_raw/international_comparison/hard001_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "hard001",
+  "question": "Against the backdrop of China's strong push for semiconductor industry self-sufficiency and many regions issuing special support policies, what are the R&D expenses and operating revenue disclosed in United Microelectronics Corporation (UMC)'s 2022 annual report (in New Taiwan Dollars)? What is the R&D investment ratio calculated therefrom? In the domestic semiconductor industry, the top 10% by operating revenue are classified as leading enterprises. By how many percentage points does UMC's R&D investment ratio differ from the median R&D investment ratio of these leading enterprises?",
+  "guidelines": "Please answer in order: (1) UMC 2022 R&D expenses (hundred million NTD, 2 decimal places); (2) UMC 2022 operating revenue (hundred million NTD, 2 decimal places); (3) UMC R&D investment ratio (%, 2 decimal places); (4) Difference between UMC's R&D investment ratio and the median R&D investment ratio of domestic semiconductor industry leading enterprises (top 10% by revenue) (percentage points, 2 decimal places, negative if lower；return as an array). If relevant data cannot be found, please answer \"No relevant data found\".",
+  "steps": [
+    "Extract financial data from United Microelectronics Corporation (UMC) 2022 Annual Report (Form 20-F) PDF: Research and development expenses NT$12,953 million (i.e., 12.953 billion NTD), Operating revenues NT$278,705 million (i.e., 278.705 billion NTD). The annual report pages 46-47 explicitly state R&D as percentage of revenue is 4.6%.",
+    "Calculate UMC's R&D investment ratio precisely: 12953 / 278705 × 100 = 4.65%.",
+    "Filter from company_profile.csv enterprises with industry='semiconductor industry', 172 enterprises total. Merge with company_operation_status.csv, filter records with non-empty R&D investment ratio, 169 valid records.",
+    "Sort by operating revenue in descending order, take top 10% (operating revenue >= 12.169 billion yuan) as leading enterprises, 17 enterprises; the remaining 152 are non-leading enterprises.",
+    "Calculate the median R&D investment ratio of leading enterprises: 5.81%.",
+    "Calculate the difference between UMC and the leading enterprise median: 4.65 - 5.81 = -1.16 percentage points, indicating UMC's R&D investment intensity is below the median of domestic semiconductor industry leading enterprises."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "United Microelectronics Corporation (UMC) 2022 Annual Report (Form 20-F), extracted R&D expenses NT$12,953 million and operating revenue NT$278,705 million",
+    "company_profile.csv (filtered 172 semiconductor industry enterprises) merged with company_operation_status.csv (with R&D investment ratio), 169 valid enterprise records after merge"
+  ],
+  "milestone": {
+    "UMC 2022 R&D expenses (hundred million NTD)": 129.53,
+    "UMC 2022 operating revenue (hundred million NTD)": 2787.05,
+    "UMC R&D investment ratio (%)": 4.65,
+    "Valid semiconductor industry enterprises": 169,
+    "Leading enterprise count (top 10%)": 17,
+    "Non-leading enterprise count": 152,
+    "Top 10% revenue threshold (hundred million yuan)": 121.69,
+    "Leading enterprise median R&D investment ratio (%)": 5.81,
+    "Non-leading enterprise median R&D investment ratio (%)": 7.98,
+    "UMC vs. leading median difference (percentage points)": -1.16
+  },
+  "answer": [
+    129.53,
+    2787.05,
+    4.65,
+    -1.16
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/hard002_result.json b/assets/qa_raw/international_comparison/hard002_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e3bfa9d406367b68c455161005bf6c6f1520aa1d
--- /dev/null
+++ b/assets/qa_raw/international_comparison/hard002_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "hard002",
+  "question": "In 2022, against the backdrop of the expiration of new energy vehicle purchase subsidy policies and the transition of industry support from direct subsidies to indirect incentives such as purchase tax exemption, what is the ratio of government subsidies and related income to operating revenue in Li Auto's annual report? Compared with the median government subsidy-to-revenue ratio of private enterprises and state-owned enterprises (including central and local state-owned enterprises) in the domestic automotive manufacturing industry, by how many percentage points is it higher for each?",
+  "guidelines": "Answer in order: Li Auto government subsidy-to-revenue ratio (%), percentage points above private enterprise median, percentage points above state-owned enterprise median. Retain 2 decimal places and return as an array. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Extract from Li Auto (LI) 2022 Annual Report PDF (20-F): Consolidated Statements of Comprehensive Loss - Total revenues = RMB 45,286,816 thousand; Others, net = RMB 625,633 thousand. According to the accounting policy notes in the annual report, non-designated-purpose government subsidies (Other subsidies) are recorded as income under Others, net; this line item also includes VAT refunds of RMB 234,531 thousand and other government-related income.",
+    "Calculate Li Auto's 2022 government subsidy and related income as percentage of revenue = 625,633 / 45,286,816 × 100% = 1.38%.",
+    "Filter from company_profile.csv enterprises with industry=\"automotive manufacturing industry\", 230 enterprises total. Group by ownership: 161 private enterprises, 31 local state-owned enterprises, 17 central state-owned enterprises, 17 foreign enterprises, 4 Sino-foreign joint ventures.",
+    "Merge with company_operation_status.csv, read \"government reward and subsidy\" and \"operating revenue amount\" fields for each enterprise. After excluding records with zero revenue or missing subsidy data, 224 valid samples. Calculate each enterprise's government subsidy-to-revenue ratio = government reward and subsidy / operating revenue amount × 100%.",
+    "Calculate median by ownership: 156 private enterprises, median = 0.72%; state-owned enterprises (16 central + 31 local) = 47 total, median = 0.58%.",
+    "Calculate comparison gap: Li Auto is 1.38% - 0.72% = 0.66 percentage points above private enterprise median; 1.38% - 0.58% = 0.80 percentage points above state-owned enterprise median."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "Li Auto (LI) 2022 Annual Report PDF (20-F), total revenues RMB 45,286,816 thousand, Others, net RMB 625,633 thousand, including non-designated-purpose government subsidies and VAT refunds.",
+    "Found 230 automotive manufacturing industry enterprises in company_profile.csv, including 161 private enterprises and 48 state-owned enterprises (central + local).",
+    "Obtained government subsidy and operating revenue data for 224 valid enterprises from company_operation_status.csv."
+  ],
+  "milestone": {
+    "Li Auto total revenue (thousand RMB)": 45286816,
+    "Li Auto Others, net (thousand RMB)": 625633,
+    "Li Auto government subsidy-to-revenue ratio (%)": 1.38,
+    "Valid automotive manufacturing private enterprise samples": 156,
+    "Valid automotive manufacturing state-owned enterprise samples": 47,
+    "Private enterprise median subsidy-to-revenue ratio (%)": 0.72,
+    "State-owned enterprise median subsidy-to-revenue ratio (%)": 0.58,
+    "Li Auto above private median (percentage points)": 0.66,
+    "Li Auto above state-owned median (percentage points)": 0.8
+  },
+  "answer": [
+    1.38,
+    0.66,
+    0.8
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/hard003_result.json b/assets/qa_raw/international_comparison/hard003_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a38ba85b488c7e67ecbaa8936209ea590803482
--- /dev/null
+++ b/assets/qa_raw/international_comparison/hard003_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "hard003",
+  "question": "2022年，在碳达峰政策驱动光伏装机需求激增、多晶硅阶段性供不应求导致价格大幅上涨的背景下，大全新能源（Daqo New Energy）年报中的净利润率（净利润÷营业收入×100%）是多少？分别与国内化学原料和化学制品制造业中民营企业和国有企业（含中央及地方国有企业）的净利润率中位数相差多少个百分点？",
+  "guidelines": "依次回答：大全新能源2022年净利润率（%）、与民营企业中位数之差（百分点）、与国有企业中位数之差（百分点）。数值保留2位小数，并返回数组。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从大全新能源（DQ）2022年度报告PDF（20-F年报）的合并损益表（Consolidated Statements of Operations）中提取：Revenues = $4,608,350千美元；Net income = $2,479,642千美元。",
+    "计算大全新能源2022年净利润率 = 2,479,642 / 4,608,350 × 100% = 53.81%。由于计算比率，美元计价与人民币计价结果一致。",
+    "从company_profile.csv中筛选industry为\"化学原料和化学制品制造业\"的企业，共364家。按ownership分组：民营企业263家、地方国有企业62家、中央国有企业21家、国有企业（其他）1家、国有企业（院所）2家、外资企业13家、集体企业1家、中外合资经营企业1家。",
+    "关联company_operation_status.csv，读取各企业的\"净利润金额\"和\"营业收入金额\"字段，剔除营收为0或数据缺失的记录后，有效样本364家。计算每家企业的净利润率 = 净利润金额 / 营业收入金额 × 100%。",
+    "分所有制计算净利润率中位数：民营企业263家，中位数 = 8.33%；国有企业（中央国有企业21家 + 地方国有企业62家 + 国有企业（其他）1家 + 国有企业（院所）2家）共86家，中位数 = 9.05%。",
+    "综合分析：大全新能源2022年净利润率53.81%，远高于国内化工行业民营企业中位数8.33%（高45.48个百分点）和国有企业中位数9.05%（高44.76个百分点）；2022年多晶硅价格阶段性大幅上涨，大全作为多晶硅龙头企业净利润率远超传统化工行业水平。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "大全新能源（DQ）2022年度报告PDF（20-F年报），合并损益表：Revenues = $4,608,350千美元，Net income = $2,479,642千美元。",
+    "从company_profile.csv中找到化学原料和化学制品制造业364家企业；国有样本按完整口径为中央+地方+其他+院所共86家。",
+    "从company_operation_status.csv中获取364家有效企业的净利润和营业收入数据。"
+  ],
+  "milestone": {
+    "大全新能源Revenue(千美元)": 4608350,
+    "大全新能源Net income(千美元)": 2479642,
+    "大全新能源净利润率(%)": 53.81,
+    "化工业民营企业有效样本(家)": 263,
+    "化工业国有企业有效样本(家)": 86,
+    "民营企业净利润率中位数(%)": 8.33,
+    "国有企业净利润率中位数(%)": 9.05,
+    "大全高于民营中位数(百分点)": 45.48,
+    "大全高于国有中位数(百分点)": 44.76
+  },
+  "answer": [
+    53.81,
+    45.48,
+    44.76
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/hard004_result.json b/assets/qa_raw/international_comparison/hard004_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..908db26eff0961e8d6cd864ad2df2d1e15df43c6
--- /dev/null
+++ b/assets/qa_raw/international_comparison/hard004_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "hard004",
+  "question": "In 2022, against the backdrop of intensively issued policies promoting biopharmaceutical industry cluster development and encouraging innovative drug R&D across regions, what is Zai Lab's price-to-sales ratio (market cap ÷ annual operating revenue) in multiples? How many times is its price-to-sales ratio relative to the median price-to-sales ratio of domestic pharmaceutical manufacturing industry leading enterprises (top 10% by revenue)?",
+  "guidelines": "Answer in order: (1) Zai Lab 2022 price-to-sales ratio (multiples); (2) Zai Lab's price-to-sales ratio as a multiple of the domestic pharmaceutical manufacturing industry top 10% by revenue leading enterprises' median price-to-sales ratio. Retain 2 decimal places for all values and return as an array. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Extract from Zai Lab (ZLAB) 2022 Annual Report (10-K): Consolidated Statements of Operations - Total revenues = $215,040 thousand, including Product revenue, net $212,672 thousand and Collaboration revenue $2,368 thousand.",
+    "Extract from annual report Consolidated Balance Sheets: As of December 31, 2022, issued and outstanding common shares = 960,219,570. Each ADS represents 10 common shares, hence ADS equivalent = 96,021,957. ZLAB ADS closing price on December 30, 2022 (last trading day of year) = $33.97. Calculate market cap = 96,021,957 × $33.97 = $3,261,865,879.",
+    "Calculate Zai Lab price-to-sales ratio = market cap / operating revenue = $3,261,865,879 / $215,040,000 = 15.17x.",
+    "Filter from company_profile.csv enterprises with industry=\"pharmaceutical manufacturing industry\", 449 enterprises total. Merge with company_operation_status.csv; after excluding records with zero operating revenue or missing/zero company market cap, 436 valid samples.",
+    "Calculate each enterprise's price-to-sales ratio = company market cap (hundred million yuan) × 10^8 / operating revenue amount (yuan). Rank by operating revenue, take top 10% (revenue ≥ 9.266 billion yuan) as leading enterprises, 44 enterprises. Leading enterprise median price-to-sales ratio = 1.96x.",
+    "Calculate Zai Lab's price-to-sales ratio as a multiple of domestic pharmaceutical manufacturing industry top 10% leading enterprise median: 15.17 / 1.96 = 7.73x."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "Zai Lab (ZLAB) 2022 Annual Report (10-K), Total revenues $215,040 thousand, outstanding common shares 960,219,570, each ADS represents 10 shares, year-end ADS closing price $33.97.",
+    "Filtered 449 pharmaceutical manufacturing industry enterprises from company_profile.csv.",
+    "Obtained revenue and market cap data for 436 valid samples from company_operation_status.csv."
+  ],
+  "milestone": {
+    "Zai Lab annual revenue (thousand USD)": 215040,
+    "Zai Lab outstanding common shares": 960219570,
+    "ADS year-end closing price (USD)": 33.97,
+    "Zai Lab market cap (USD)": 3261865879,
+    "Zai Lab price-to-sales ratio (x)": 15.17,
+    "Valid pharmaceutical manufacturing samples": 436,
+    "Leading enterprise revenue threshold (yuan)": 9266378824,
+    "Leading enterprise count": 44,
+    "Leading enterprise median price-to-sales ratio (x)": 1.96,
+    "Industry-wide median price-to-sales ratio (x)": 4.21,
+    "Zai Lab to leading median ratio (x)": 7.73
+  },
+  "answer": [
+    15.17,
+    7.73
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/hard005_result.json b/assets/qa_raw/international_comparison/hard005_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..dde009a92b8884966e1d7be24e3cccac9394842b
--- /dev/null
+++ b/assets/qa_raw/international_comparison/hard005_result.json
@@ -0,0 +1,42 @@
+{
+  "id": "hard005",
+  "question": "In 2022, against the backdrop of intensively issued policies in Shanghai, Hefei, Hangzhou and elsewhere promoting high-quality development of the integrated circuit industry, what is Silicon Motion Technology's per capita net profit in the annual report converted to RMB in ten thousand yuan? Compared with the median per capita net profit of private enterprises and state-owned enterprises in the domestic semiconductor industry respectively, by how many times is it higher?",
+  "guidelines": "Answer in order: Silicon Motion 2022 per capita net profit (ten thousand yuan, converted at 2022 average exchange rate 1 USD ≈ 6.73 RMB), ratio to domestic semiconductor industry private enterprise median per capita net profit (times), ratio to state-owned enterprise median per capita net profit (times). Retain 2 decimal places and return as an array. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "Extract from Silicon Motion (SIMO) 2022 Annual Report PDF (20-F): Consolidated Statements of Income - Net Income = US$172,510 thousand (i.e., $172.51 million). From Item 6 Employees section: As of December 31, 2022, the company had 1,643 employees.",
+    "Calculate Silicon Motion 2022 per capita net profit: 172,510,000 / 1,643 = 104,996.96 USD per person. Convert to RMB at 2022 average rate 6.73: 104,996.96 × 6.73 = 706,629.52 yuan = 70.66 ten thousand yuan.",
+    "Filter from company_profile.csv enterprises with industry=\"semiconductor industry\", 172 enterprises total. Group by ownership: 111 private enterprises, 20 foreign enterprises, 16 local state-owned enterprises, 12 central state-owned enterprises, 9 Sino-foreign joint ventures, 2 other state-owned, 2 research-institute state-owned.",
+    "Merge with company_operation_status.csv, read \"net profit amount\" and \"total employees\" fields, calculate each enterprise's per capita net profit = net profit amount / total employees, convert to ten thousand yuan. Exclude records with zero employees or missing data.",
+    "Calculate median per capita net profit by ownership: 111 private enterprises, median = 8.58 ten thousand yuan; state-owned enterprises (12 central + 16 local + 2 other + 2 research-institute) = 32 total, median = 14.75 ten thousand yuan.",
+    "Calculate Silicon Motion per capita net profit as multiple of domestic semiconductor industry private enterprise median: 70.66 / 8.58 = 8.23x; as multiple of state-owned enterprise median: 70.66 / 14.75 = 4.79x."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "Silicon Motion (SIMO) 2022 Annual Report PDF (20-F), Net Income = US$172,510 thousand, total employees 1,643.",
+    "Found 172 semiconductor industry enterprises in company_profile.csv.",
+    "Obtained net profit and employee data for each enterprise from company_operation_status.csv."
+  ],
+  "milestone": {
+    "Silicon Motion Net Income (thousand USD)": 172510,
+    "Silicon Motion employee count": 1643,
+    "Silicon Motion per capita net profit (USD)": 104996.96,
+    "2022 average exchange rate (USD/RMB)": 6.73,
+    "Silicon Motion per capita net profit (ten thousand RMB)": 70.66,
+    "Semiconductor industry private enterprise sample count": 111,
+    "Semiconductor industry state-owned enterprise sample count": 32,
+    "Private enterprise median per capita net profit (ten thousand yuan)": 8.58,
+    "State-owned enterprise median per capita net profit (ten thousand yuan)": 14.75,
+    "Silicon Motion vs. private ratio (x)": 8.23,
+    "Silicon Motion vs. state-owned ratio (x)": 4.79
+  },
+  "answer": [
+    70.66,
+    8.23,
+    4.79
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/hard006_result.json b/assets/qa_raw/international_comparison/hard006_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..323e9a71c5fe1e4d338afeaa983badeb804de3ab
--- /dev/null
+++ b/assets/qa_raw/international_comparison/hard006_result.json
@@ -0,0 +1,39 @@
+{
+  "id": "hard006",
+  "question": "A global technology thematic portfolio uses a \"technological moat + policy diffusion\" framework for its semiconductor sub-portfolio. For candidate companies, first calculate: 1. Technological moat gap = company advanced process revenue ratio - median R&D investment ratio of A-share semiconductor industry top 10% by revenue in 2022, where advanced process revenue ratio = (5nm + 7nm revenue) ÷ wafer revenue; 2. Policy diffusion ratio = count of China semiconductor industry policies ÷ number of provincial-level administrative regions covered by non-national policies; 3. Theme conviction score = 0.6 × technological moat gap + 0.4 × policy diffusion ratio. If technological moat gap > 40 and policy diffusion ratio > 2.5, list as core overweight with active weight = min(3.00%, theme conviction score ÷ 10); otherwise do not include in core overweight list. Using TSMC's 2022 annual report and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: technological moat gap (percentage points), policy diffusion ratio, theme conviction score, active weight, most appropriate conclusion (conclusion must specify position action and active weight). Retain 2 decimal places and return as an array. If relevant data cannot be found, please answer \"No relevant data found\".",
+  "steps": [
+    "Web search TSMC 2022 annual report; from the process revenue disclosure section extract 2022 5nm revenue = 508,689.9, 7nm revenue = 535,153.8, wafer revenue = 1,991,855.9, all in NT$ million.",
+    "Calculate advanced process revenue ratio = (508,689.9 + 535,153.8) ÷ 1,991,855.9 × 100% = 52.4056%, rounded to 2 decimals = 52.41%.",
+    "Filter from company_profile.csv A-share companies with industry=\"semiconductor industry\", 172 total; then from company_operation_status.csv extract 2022 operating revenue amount and R&D investment ratio for these companies, retaining 169 companies with both fields valid.",
+    "Sort the 169 companies by operating revenue descending, take top 10% = 17 companies, calculate median R&D investment ratio = 5.81%.",
+    "Calculate technological moat gap per question definition = 52.41% - 5.81% = 46.60 percentage points.",
+    "Filter semiconductor industry policies from policy_release_status.csv, 44 records; after excluding province=\"national\", 15 provincial-level administrative regions covered.",
+    "Calculate policy diffusion ratio = 44 ÷ 15 = 2.9333, rounded to 2.93; then theme conviction score = 0.6 × 46.60 + 0.4 × 2.93 = 29.1320, rounded to 29.13. Since technological moat gap > 40 and policy diffusion ratio > 2.5, core overweight condition is met; active weight = min(3.00%, 29.13 ÷ 10) = 2.91%."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Web search obtained raw data: 5nm, 7nm, wafer revenue = 508,689.9, 535,153.8, 1,991,855.9 respectively.",
+    "company_profile.csv and company_operation_status.csv provided 172 A-share semiconductor industry companies and 169 with valid 2022 R&D investment ratio data.",
+    "policy_release_status.csv provided 44 semiconductor-related policies and 15 non-national provincial-level administrative region coverage information."
+  ],
+  "milestone": {
+    "TSMC advanced process revenue ratio (%)": 52.41,
+    "A-share semiconductor industry top 10% by revenue median R&D investment ratio (%)": 5.81,
+    "Technological moat gap (percentage points)": 46.6,
+    "Policy diffusion ratio": 2.93,
+    "Theme conviction score": 29.13,
+    "Active weight (%)": 2.91
+  },
+  "answer": [
+    46.6,
+    2.93,
+    29.13,
+    "Core overweight, active weight 2.91%"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/hard007_result.json b/assets/qa_raw/international_comparison/hard007_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..43a5626d997778ae89ecb4972da71c849f122dde
--- /dev/null
+++ b/assets/qa_raw/international_comparison/hard007_result.json
@@ -0,0 +1,44 @@
+{
+  "id": "hard007",
+  "question": "A global electric vehicle growth portfolio uses an \"innovation offsets earnings deficit\" framework for loss-making but high-R&D complete vehicle companies. For each candidate, first calculate: 1. Innovation excess = company R&D-to-revenue ratio − median R&D-to-revenue ratio among A-share automotive manufacturing firms in the top 10% by operating revenue in 2022; 2. Profit gap = median net profit margin among those top-10%-by-revenue A-share automotive manufacturing firms − company net profit margin; 3. Policy leverage = number of China automotive manufacturing industry policies ÷ number of provincial-level administrative regions covered by non-national policies. If innovation excess > 8, profit gap < 10, and policy leverage > 1.5, tactical overweight is allowed with active weight = min(2.00%, innovation excess ÷ 5 − profit gap ÷ 10 + policy leverage ÷ 10); otherwise only the watch list applies. Using Li Auto (LI) 2022 annual report and the local database, compute and state the most appropriate conclusion.",
+  "guidelines": "Answer in order: innovation excess (percentage points), profit gap (percentage points), policy leverage, most appropriate conclusion. Retain 2 decimal places and return as an array; the conclusion must specify position action and active weight. If relevant data cannot be found, please answer \"No relevant data found\".",
+  "steps": [
+    "Web search Li Auto 2022 annual report; take Total revenues 45,286,816, Research and development expenses 6,780,032, Net loss 2,032,348 (RMB thousand); R&D-to-revenue ratio = 6,780,032 ÷ 45,286,816 × 100% = 14.9713%, rounded to 14.97%; net profit margin = −2,032,348 ÷ 45,286,816 × 100% = −4.4877%, rounded to −4.49%.",
+    "In company_profile.csv, filter industry=\"automotive manufacturing industry\" and A-share exchanges (SZSE, SSE, BSE), excluding Hong Kong listings; join company_operation_status.csv on bmCode with year=2022. The profile lists 230 automotive manufacturing firms (including 7 Hong Kong–listed); the valid A-share sample is 187 firms. Among rows with non-empty R&D-to-revenue ratio, operating revenue, and net profit, sort by operating revenue descending.",
+    "Top-10% count k = max(1, ⌈10% × N⌉); for N = 187, k = 19. For these 19 firms: median R&D-to-revenue ratio is 4.78%; for each firm compute net profit margin = net profit ÷ operating revenue × 100%, median net profit margin is 3.24%.",
+    "Innovation excess = 14.97% − 4.78% = 10.19 percentage points; profit gap = 3.24% − (−4.49%) = 7.73 percentage points.",
+    "In policy_release_status.csv, count as automotive manufacturing policies those whose industry field contains \"automotive manufacturing industry\", 69 records; after excluding province=\"national\", deduplicate province on the remainder to obtain 20 provincial-level regions covered by non-national policies.",
+    "Policy leverage = 69 ÷ 20 = 3.45.",
+    "Because innovation excess 10.19 > 8, profit gap 7.73 < 10, and policy leverage 3.45 > 1.5, tactical overweight conditions are met; active weight = min(2.00%, 10.19 ÷ 5 − 7.73 ÷ 10 + 3.45 ÷ 10) = min(2.00%, 1.61%) = 1.61%."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Web search yields Li Auto 2022 revenue, R&D expense, and net loss (thousand RMB) for company-level R&D ratio and net margin.",
+    "company_profile.csv and company_operation_status.csv: merged A-share automotive manufacturing 2022 sample 187 firms; for the top 10% by revenue (19 firms), median R&D-to-revenue ratio 4.78% and median net profit margin 3.24%. In national_industry_status.csv, the industry \"automotive manufacturing industry\" with district \"national\" has median R&D ratio 4.615% under a full-industry definition and does not substitute the revenue top-10% stratification required by the problem.",
+    "policy_release_status.csv: 69 policies whose industry field contains automotive manufacturing; 20 deduplicated non-national provinces; policy leverage 3.45."
+  ],
+  "milestone": {
+    "Li Auto R&D-to-revenue ratio (%)": 14.97,
+    "A-share automotive manufacturing valid sample count (2022)": 187,
+    "Top-10%-by-revenue firm count": 19,
+    "Median R&D-to-revenue ratio of A-share automotive top 10% by revenue (%)": 4.78,
+    "Innovation excess (percentage points)": 10.19,
+    "Median net profit margin of A-share automotive top 10% by revenue (%)": 3.24,
+    "Profit gap (percentage points)": 7.73,
+    "China automotive manufacturing related policy count": 69,
+    "Non-national policy covered provincial regions (deduped)": 20,
+    "Policy leverage": 3.45,
+    "Active weight (%)": 1.61
+  },
+  "answer": [
+    10.19,
+    7.73,
+    3.45,
+    "Tactical overweight, active weight 1.61%"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/hard008_result.json b/assets/qa_raw/international_comparison/hard008_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..625b84817f7d92ed225b9d8a623c865ac47dc5b8
--- /dev/null
+++ b/assets/qa_raw/international_comparison/hard008_result.json
@@ -0,0 +1,38 @@
+{
+  "id": "hard008",
+  "question": "A global consumer defensive portfolio, when screening platform retailers, views fulfillment expenses as having quasi-fixed cost characteristics. For candidate companies, apply the following stress test: ① Fulfillment expense ratio = Fulfillment ÷ Net revenues; ② Stressed net profit margin = company net profit margin - 0.2 × fulfillment expense ratio; ③ Defense gap = stressed net profit margin - median net profit margin of A-share wholesale and retail industry top 10% by revenue in 2022. If stressed net profit margin < 0, do not include in defensive core position; if stressed net profit margin is between 0 and industry median, benchmark hold only; if above industry median, overweight is allowed. Using JD.com (JD) 2022 annual report and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: fulfillment expense ratio (%), stressed net profit margin (%), defense gap (percentage points), most appropriate conclusion. Retain 2 decimal places and return as an array; conclusion must specify position action. If relevant data cannot be found, please answer \"No relevant data found\".",
+  "steps": [
+    "Web search JD.com 2022 annual report; extract Net revenues = 1,046,236, Fulfillment = 63,011, Net income = 9,691, all in RMB million.",
+    "Fulfillment expense ratio = 63,011 ÷ 1,046,236 × 100%; net profit margin = 9,691 ÷ 1,046,236 × 100%. Carry full precision through intermediate steps—do not round these ratios to two decimals before the next calculation.",
+    "Stressed net profit margin = net profit margin − 0.2 × fulfillment expense ratio, computed directly from the unrounded ratio values above; only then round the final stressed margin to two decimal places (rounding pre-rounded 0.93% and 6.02% would incorrectly yield −0.27%).",
+    "From company_profile.csv, filter industry consistent with A-share wholesale and retail (批发和零售业) where exchange is Shanghai, Shenzhen, or Beijing (exclude Hong Kong Stock Exchange listings). Merge with company_operation_status.csv for year = 2022; keep records with valid operating revenue and net profit and operating revenue > 0 (n = 185).",
+    "For each company, net profit margin = net profit amount ÷ operating revenue amount × 100%. Sort by operating revenue descending; take the top ⌈n × 10%⌉ = 19 companies; median net profit margin ≈ 0.880508%, rounded to two decimals = 0.88%.",
+    "Defense gap = stressed net profit margin − industry median (in percentage points), using full-precision values before rounding the gap to two decimals. Because stressed net profit margin < 0, JD.com must not be included in the defensive core position."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "JD.com 2022 annual report figures for revenue, fulfillment, and net income (1,046,236; 63,011; 9,691 RMB million) as given or verified by web search.",
+    "company_profile.csv supplies industry and exchange fields to define the A-share universe excluding Hong Kong–listed firms.",
+    "company_operation_status.csv supplies year-2022 operating revenue and net profit for percentile sorting and median calculation."
+  ],
+  "milestone": {
+    "Fulfillment expense ratio (%)": 6.02,
+    "Net profit margin (%)": 0.93,
+    "Stressed net profit margin (%)": -0.28,
+    "A-share wholesale and retail top 10% by revenue median net profit margin (%)": 0.88,
+    "Defense gap (percentage points)": -1.16,
+    "A-share sample note": "exchange ∈ {Shanghai, Shenzhen, Beijing}; Hong Kong excluded; n = 185; top decile count = 19"
+  },
+  "answer": [
+    6.02,
+    -0.28,
+    -1.16,
+    "Do not include in defensive core position"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/hard009_result.json b/assets/qa_raw/international_comparison/hard009_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3be774ef6e9e2b5bb71904a432ee985e032f82f8
--- /dev/null
+++ b/assets/qa_raw/international_comparison/hard009_result.json
@@ -0,0 +1,39 @@
+{
+  "id": "hard009",
+  "question": "An active equity manager uses a \"high-profit reinvestment\" framework for the internet retail growth sub-portfolio. For candidate companies, first calculate: ① High-quality growth score = difference between company net profit margin and median net profit margin of A-share wholesale and retail industry top 10% by revenue in 2022 + 0.5 × transaction services revenue ratio + 0.5 × R&D investment ratio; ② Policy diffusion ratio = count of wholesale and retail industry policies ÷ number of provincial-level administrative regions covered by non-national policies. If high-quality growth score > 35 and policy diffusion ratio > 2.0, list as strategic overweight with active weight = min(4.00%, high-quality growth score ÷ 10); otherwise ordinary position only. Using Pinduoduo (PDD) 2022 annual report and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: high-quality growth score, policy diffusion ratio, active weight (%), most appropriate conclusion. Retain 2 decimal places and return as an array; conclusion must specify position action. If relevant data cannot be found, please answer \"No relevant data found\".",
+  "steps": [
+    "Web search Pinduoduo 2022 annual report; extract Transaction services = 27,626,494, Total revenues = 130,557,589, Research and development expenses = 10,384,716, Net income = 31,538,062, all in RMB thousand.",
+    "Calculate transaction services revenue ratio = 27,626,494 ÷ 130,557,589 × 100% = 21.1604%, rounded to 21.16%; calculate R&D investment ratio = 10,384,716 ÷ 130,557,589 × 100% = 7.9541%, rounded to 7.95%; calculate net profit margin = 31,538,062 ÷ 130,557,589 × 100% = 24.1564%, rounded to 24.16%.",
+    "Filter from company_profile.csv A-share companies with industry=\"wholesale and retail industry\", 273 total; then from company_operation_status.csv extract 2022 operating revenue amount and net profit amount for these companies.",
+    "Among 273 companies with valid operating revenue and net profit, calculate each company's net profit margin = net profit amount ÷ operating revenue amount × 100%, then sort by operating revenue descending and take top 10% = 28 companies, median net profit margin = 0.97%; thus net profit margin difference = 24.16% - 0.97% = 23.18 percentage points.",
+    "Calculate high-quality growth score per question definition = 23.18 + 0.5 × 21.16 + 0.5 × 7.95 = 23.18 + 10.58 + 3.975 = 37.7350, rounded to 37.73.",
+    "Filter from policy_release_status.csv policies whose industry field contains wholesale and retail industry, 28 records; after excluding province=\"national\", 14 provincial-level administrative regions covered. Calculate policy diffusion ratio = 28 ÷ 14 = 2.0000, rounded to 2.00.",
+    "Since high-quality growth score > 35 but policy diffusion ratio 2.00 does not meet the > 2.0 threshold, strategic overweight condition is not met; ordinary position only, strategic overweight active weight formula not applicable (recorded as 0.00%)."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Web search obtained Pinduoduo 2022 annual report transaction services revenue, total revenues, R&D expenses, and net profit: 27,626,494, 130,557,589, 10,384,716, 31,538,062 respectively.",
+    "company_profile.csv and company_operation_status.csv provided 273 A-share wholesale and retail industry companies and their 2022 operating revenue and net profit samples.",
+    "policy_release_status.csv provided 28 wholesale and retail industry policies and 14 non-national provincial-level administrative region coverage information."
+  ],
+  "milestone": {
+    "Transaction services revenue ratio (%)": 21.16,
+    "R&D investment ratio (%)": 7.95,
+    "Net profit margin difference (percentage points)": 23.18,
+    "High-quality growth score": 37.73,
+    "Policy diffusion ratio": 2.0,
+    "Active weight (%)": 0.0
+  },
+  "answer": [
+    37.73,
+    2.0,
+    0.0,
+    "Ordinary position"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/hard010_result.json b/assets/qa_raw/international_comparison/hard010_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..9ea6ba5c7f0bb976e43e0cd16f17950610ddc87d
--- /dev/null
+++ b/assets/qa_raw/international_comparison/hard010_result.json
@@ -0,0 +1,37 @@
+{
+  "id": "hard010",
+  "question": "A private wealth global consumption themed account wishes to incorporate the feature of 'overseas cash flow hedging domestic cycle' for China optional consumption. For candidate companies, apply the following rules: ① Overseas Hedging Quality Score = the difference between the company's net profit margin and the median net profit margin of A-share wholesale and retail industry in 2022 + 0.5 × international market revenue share; ② If the score ≥ 15, list as core holding, with active weight = min(3.00%, Overseas Hedging Quality Score ÷ 8); otherwise, only satellite holding is permitted. Based on MINISO (MNSO) FY2022 annual report and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: International market revenue share (%), the difference between net profit margin and the median net profit margin of A-share wholesale and retail industry (percentage points), Overseas Hedging Quality Score, and the most appropriate conclusion. Retain 2 decimal places for numerical values and return as an array. The conclusion must include position action and active weight. Example: [\"26.20\", \"4.79\", \"17.89\", \"Core holding, active weight 2.24%\"]. If relevant data cannot be found, respond with \"Relevant data not found\".",
+  "steps": [
+    "Search the web for MINISO's FY2022 annual report; extract Revenue of 10,085,649 and Profit for the year of 639,743, both in RMB thousand; and extract international market revenue contribution of 26.2% from segment disclosure.",
+    "International market revenue share is directly disclosed in the annual report as 26.20%. Calculate net profit margin = 639,743 ÷ 10,085,649 × 100% ≈ 6.34% (two decimal places).",
+    "From national_industry_status.csv, select the row where the industry is \"批发和零售业\" (wholesale and retail) and district is \"全国\" (national). Compute the industry-level median net profit margin as median net profit amount ÷ median revenue amount × 100% = 78,000,847 ÷ 3,970,535,358 × 100% ≈ 1.96%.",
+    "Net profit margin difference vs. the industry benchmark = 6.34% − 1.96% = 4.38 percentage points (two decimal places).",
+    "Per the problem definition, Overseas Hedging Quality Score = 4.38 + 0.5 × 26.20 = 17.48.",
+    "Because the score is ≥ 15, the position qualifies as a core holding; active weight = min(3.00%, 17.48 ÷ 8) = 2.18% (two decimal places)."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "Web search yielded MINISO FY2022 revenue, profit for the year, and international market revenue share of 10,085,649 (RMB thousand), 639,743 (RMB thousand), and 26.2% respectively.",
+    "national_industry_status.csv provides the national wholesale-and-retail sample's median net profit amount and median revenue amount used to define the industry median net profit margin."
+  ],
+  "milestone": {
+    "International market revenue share (%)": 26.2,
+    "Net profit margin (%)": 6.34,
+    "Median net profit margin of A-share wholesale and retail industry (%)": 1.96,
+    "Net profit margin difference (percentage points)": 4.38,
+    "Overseas Hedging Quality Score": 17.48,
+    "Active weight (%)": 2.18
+  },
+  "answer": [
+    26.2,
+    4.38,
+    17.48,
+    "Core holding, active weight 2.18%"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/hard011_result.json b/assets/qa_raw/international_comparison/hard011_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f4ac2b0fc6248d06c2bd0e392cffad825af35be
--- /dev/null
+++ b/assets/qa_raw/international_comparison/hard011_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "hard011",
+  "question": "When an industrial internet fund evaluation distribution platform transforms toward high value-added services, it adopts a two-step method: ? Intangible input excess = company's R&D investment ratio - median R&D investment ratio of the A-share wholesale and retail industry in 2022; ? Transformation score = net service revenue ratio + intangible input excess. Only when net service revenue ratio >= 5 and transformation score >= 5 can a company enter the platform-based overweight list; otherwise, it enters the watch list. Based on ZKH's 2022 annual report and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: Net service revenue ratio (%), intangible input excess (percentage points), transformation score, and the most appropriate conclusion. Retain 2 decimal places for numerical values and return as an array. The conclusion must clearly indicate whether to enter the platform-based overweight list or the watch list. If relevant data cannot be found, respond with \"Relevant data not found\".",
+  "steps": [
+    "Search the web for ZKH's 2022 annual report, extract Net service revenues of 179,508, Total revenues of 8,315,236, and Research and development expenses of 240,534, all in RMB thousand.",
+    "Calculate net service revenue ratio = 179,508 ÷ 8,315,236 × 100% = 2.1588%, rounded to 2.16% with two decimal places; calculate R&D investment ratio = 240,534 ÷ 8,315,236 × 100% = 2.8927%, rounded to 2.89% with two decimal places.",
+    "Filter A-share companies with industry=\"wholesale and retail\" from company_profile.csv, totaling 273 companies; then extract the 2022 R&D investment ratio field for these companies from company_operation_status.csv. Among the 141 companies with valid R&D investment ratio, the median is 0.40%.",
+    "According to the problem definition, calculate intangible input excess = 2.89% - 0.40% = 2.49 percentage points.",
+    "Calculate transformation score = 2.16 + 2.49 = 4.65.",
+    "Since net service revenue ratio of 2.16% is below 5%, and transformation score of 4.65 is also below 5, the platform-based overweight list conditions are not met; therefore, the conclusion is watch list."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "Web search obtained ZKH's 2022 annual report data for net service revenue, total revenue, and R&D expenses of 179,508, 8,315,236, and 240,534 respectively.",
+    "company_profile.csv and company_operation_status.csv provide data for 273 A-share wholesale and retail companies and 141 valid 2022 R&D investment ratio samples."
+  ],
+  "milestone": {
+    "Net service revenue ratio (%)": 2.16,
+    "R&D investment ratio (%)": 2.89,
+    "Median R&D investment ratio of A-share wholesale and retail industry (%)": 0.4,
+    "Intangible input excess (percentage points)": 2.49,
+    "Transformation score": 4.65
+  },
+  "answer": [
+    "2.16",
+    "2.49",
+    "4.65",
+    "Watch list"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/hard012_result.json b/assets/qa_raw/international_comparison/hard012_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..809c6439c4b4da1d27421380b3ea012f139ee02c
--- /dev/null
+++ b/assets/qa_raw/international_comparison/hard012_result.json
@@ -0,0 +1,38 @@
+{
+  "id": "hard012",
+  "question": "A real estate transformation special account evaluates companies that replace development cycles with existing property services. For candidate companies, apply the following rules: ① Transformation buffer = home renovation and furnishing revenue ratio + R&D investment ratio; ② Profit gap = median net profit margin of top 10% A-share real estate companies by revenue in 2022 - company net profit margin; ③ Net transformation score = transformation buffer - profit gap. If net transformation score ≤ 0, exclude; if 0 < net transformation score < 5, only tactical small overweight is permitted, with active weight = min(1.50%, net transformation score ÷ 10); if net transformation score ≥ 5, standard overweight is permitted. Based on KE Holdings (BEKE) 2022 annual report and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: Transformation buffer, profit gap (percentage points), net transformation score, and the most appropriate conclusion. Retain 2 decimal places for numerical values and return as an array. The conclusion must include position action and active weight. If relevant data cannot be found, respond with \"Relevant data not found\".",
+  "steps": [
+    "Search the web for KE Holdings' 2022 annual report, extract Home renovation and furnishing revenue of 5,046,627, Total net revenues of 60,668,779, Research and development expenses of 2,545,549, and Net loss of 1,397,284, all in RMB thousand.",
+    "Calculate home renovation and furnishing revenue ratio = 5,046,627 ÷ 60,668,779 × 100% = 8.3148%, rounded to 8.31% with two decimal places; calculate R&D investment ratio = 2,545,549 ÷ 60,668,779 × 100% = 4.1958%, rounded to 4.20% with two decimal places; calculate net profit margin = -1,397,284 ÷ 60,668,779 × 100% = -2.3032%, rounded to -2.30% with two decimal places.",
+    "According to the problem definition, calculate transformation buffer = 8.31 + 4.20 = 12.51.",
+    "From company_profile.csv, filter A-share companies with companyType=\"沪深\" and exchange in SZSE/SSE/BSE and industry=\"房地产业\"; merge with company_operation_status.csv records where year=2022 on bmCode.",
+    "Among samples with valid operating revenue > 0 (110 companies in total), sort by operating revenue descending, take top 10% = ceil(110×10%) = 11 companies, compute net profit margin = net profit amount ÷ operating revenue amount × 100% for each; median net profit margin is 1.92%; therefore profit gap = 1.92% - (-2.30%) = 4.22 percentage points.",
+    "Calculate net transformation score = 12.51 - 4.22 = 8.29. Since net transformation score ≥ 5, KE Holdings is eligible for standard overweight; the problem only specifies active weight formula min(1.50%, net transformation score ÷ 10) for tactical small overweight, and does not specify a concrete active weight percentage for standard overweight."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "Web search obtained KE Holdings' 2022 annual report figures for home renovation and furnishing revenue, total revenue, R&D expenses, and net loss: 5,046,627, 60,668,779, 2,545,549, and -1,397,284 respectively.",
+    "After merging and filtering company_profile.csv and company_operation_status.csv, they provide 2022 operating revenue and net profit samples for A-share (CSI) real estate companies; national_industry_status.csv and regional_industry_status.csv contain industry/regional aggregates and do not include the median net profit margin for the top 10% by revenue, so the benchmark net profit margin is derived from the firm-level tables above."
+  ],
+  "milestone": {
+    "Home renovation and furnishing revenue ratio (%)": 8.31,
+    "R&D investment ratio (%)": 4.2,
+    "Transformation buffer": 12.51,
+    "Median net profit margin of top 10% A-share real estate companies by revenue (%)": 1.92,
+    "Profit gap (percentage points)": 4.22,
+    "Net transformation score": 8.29,
+    "Active weight (%)": null
+  },
+  "answer": [
+    12.51,
+    4.22,
+    8.29,
+    "Standard overweight; the rules do not specify a concrete active weight percentage"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/hard013_result.json b/assets/qa_raw/international_comparison/hard013_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..8ad912db58a7aceb5a319b92f84e1d700fb044be
--- /dev/null
+++ b/assets/qa_raw/international_comparison/hard013_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "hard013",
+  "question": "A quasi-infrastructure growth portfolio allows allocation to data center operators during accounting loss periods, but requires a significant operating profit buffer. For candidate companies, apply the following rules: ① Profit conversion penalty = |net profit margin| ÷ adjusted EBITDA margin × 100; ② Policy diffusion ratio = number of data center, Eastern Data Western Computing, or computing power related policies ÷ number of provincial-level administrative regions covered by non-national policies; ③ Infrastructure capacity score = adjusted EBITDA margin - profit conversion penalty + 5 × policy diffusion ratio. If profit conversion penalty < 35 and infrastructure capacity score > 25, the company may be listed for satellite overweight 1.00%; otherwise, benchmark allocation only. Based on GDS Holdings (GDS) 2022 annual report and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: Profit conversion penalty, policy diffusion ratio, infrastructure capacity score, and the most appropriate conclusion. Retain 2 decimal places for numerical values and return as an array. The conclusion must clearly indicate position action. If relevant data cannot be found, respond with \"Relevant data not found\".",
+  "steps": [
+    "Search the web for GDS Holdings' 2022 annual report, extract Net revenue of 9,325,631 and Net loss of 1,266,118, both in RMB thousand; and read Adjusted EBITDA margin of 45.6% from the annual report.",
+    "Calculate net profit margin = −1,266,118 ÷ 9,325,631 ≈ −0.135763, i.e. about −13.58% (two decimal places for the percentage).",
+    "Per the problem definition, adjusted EBITDA margin 45.6% corresponds to decimal 0.456; profit conversion penalty = (1,266,118 ÷ 9,325,631) ÷ 0.456 × 100 ≈ 29.77 (retain two decimal places; avoid bias from rounding net profit margin before dividing).",
+    "From policy_release_status.csv, filter policies whose title or industry field mentions data centers, Eastern Data Western Computing, or computing power, totaling 5; treat non-national policies as \"local policies\". For rows with province column \"national\" but where the title or issuing authority indicates a provincial scope (Ningxia Hui Autonomous Region, Guizhou Province related plans), count the actual covered provinces; together with Shanghai Municipality and Yunnan Province this yields 4 provincial-level administrative regions; policy diffusion ratio = 5 ÷ 4 = 1.25.",
+    "Calculate infrastructure capacity score = 45.60 − 29.77 + 5 × 1.25 = 45.60 − 29.77 + 6.25 = 22.08.",
+    "Profit conversion penalty 29.77 < 35, but infrastructure capacity score 22.08 is not greater than 25, so the dual conditions for satellite overweight are not met; the most appropriate conclusion is benchmark allocation only."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "Web search obtained GDS Holdings' 2022 annual report figures for revenue, net loss, and adjusted EBITDA margin: 9,325,631, −1,266,118, and 45.6% respectively.",
+    "policy_release_status.csv provides 5 data center/Eastern Data Western Computing/computing power related policies; non-national local policies are counted as covering 4 provincial-level administrative regions (Shanghai Municipality, Yunnan Province, Ningxia Hui Autonomous Region, Guizhou Province) for the policy diffusion ratio denominator."
+  ],
+  "milestone": {
+    "Adjusted EBITDA margin (%)": 45.6,
+    "Net profit margin (%)": -13.58,
+    "Profit conversion penalty": 29.77,
+    "Policy diffusion ratio": 1.25,
+    "Infrastructure capacity score": 22.08
+  },
+  "answer": [
+    29.77,
+    1.25,
+    22.08,
+    "Benchmark allocation only"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/hard014_result.json b/assets/qa_raw/international_comparison/hard014_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..829677ce18fcd7dbe10210dd421c962cdc8ae401
--- /dev/null
+++ b/assets/qa_raw/international_comparison/hard014_result.json
@@ -0,0 +1,37 @@
+{
+  "id": "hard014",
+  "question": "A distressed-reversal fund applies a \"screen first, value second\" rule to platform retail stocks. For each candidate company, first compute: ① Survival score = technology service revenue ratio + R&D intensity - 0.1×|company net margin - median net margin of A-share wholesale and retail industry in 2022|; ② If company net margin is below -100%, trigger one-vote veto and remove directly; otherwise, only enter the watch pool when survival score ≥ 20. Based on Mogujie (MOGU) annual report for the fiscal year ended March 31, 2022 and local database, calculate and determine the most appropriate conclusion.",
+  "guidelines": "Answer in order: technology service revenue ratio (%), R&D intensity (%), survival score, and most appropriate conclusion. Values rounded to 2 decimal places; conclusion must clearly state whether the stock is removed; return as an array. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Web search Mogujie annual report for fiscal year ended March 31, 2022: Technology service revenues = 46,077, Total revenues = 337,469, Research and development expenses = 82,641, Net loss = 642,374, all in thousands of CNY.",
+    "Calculate technology service revenue ratio = 46,077 ÷ 337,469 × 100% = 13.6537%, rounded to 13.65%; R&D intensity = 82,641 ÷ 337,469 × 100% = 24.4885%, rounded to 24.49%; net margin = -642,374 ÷ 337,469 × 100% = -190.3505%, rounded to -190.35%.",
+    "Filter A-share companies with industry=\"wholesale and retail\" from company_profile.csv, 273 companies; extract 2022 operating revenue and net profit from company_operation_status.csv, retain all 273 companies with valid data.",
+    "Compute net margin for each company: net profit ÷ operating revenue × 100%; median net margin of A-share wholesale and retail industry in 2022 = 1.55%. Absolute deviation from industry median = | -190.35% - 1.55% | = 191.90 percentage points.",
+    "Calculate survival score per definition: 13.65 + 24.49 - 0.1×191.90 = 38.14 - 19.19 = 18.95.",
+    "Since company net margin -190.35% is below -100%, one-vote veto is triggered; regardless of survival score, the conclusion is to remove."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "Web search yields Mogujie fiscal year ended March 31, 2022: technology service revenues = 46,077, total revenues = 337,469, R&D expenses = 82,641, net loss = -642,374.",
+    "company_profile.csv and company_operation_status.csv provide 273 A-share wholesale and retail companies and their 2022 operating revenue and net profit samples."
+  ],
+  "milestone": {
+    "Technology service revenue ratio(%)": 13.65,
+    "R&D intensity(%)": 24.49,
+    "Net margin(%)": -190.35,
+    "Median net margin of A-share wholesale and retail industry(%)": 1.55,
+    "Absolute deviation of net margin from median(percentage points)": 191.9,
+    "Survival score": 18.95
+  },
+  "answer": [
+    13.65,
+    24.49,
+    18.95,
+    "Remove"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium001_result.json b/assets/qa_raw/international_comparison/medium001_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..51053efd9aa8775745f9bcfb10ed022b6bc9d54e
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium001_result.json
@@ -0,0 +1,33 @@
+{
+  "id": "medium001",
+  "question": "What was NetEase's R&D intensity (R&D expenses as a percentage of revenue) in 2022? Compared with the median R&D intensity of listed companies in China's information transmission, software and IT services industry, how many percentage points higher or lower is it?",
+  "guidelines": "Answer both sub-questions: 1) NetEase 2022 R&D intensity (2 decimal places, %); 2) Difference from industry median (2 decimal places, percentage points; positive value means above industry median; return as an array). If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from NetEase 2022 annual report (20-F) consolidated income statement: total revenue = 96,495,809 thousand CNY, R&D expenses = 15,039,014 thousand CNY; R&D intensity = 15,039,014 / 96,495,809 × 100 = 15.59%.",
+    "Filter company_profile.csv for companies with industry=\"information transmission, software and information technology services\" (information transmission, software and IT services), 644 companies.",
+    "Obtain 2022 \"R&D intensity\" for these companies from company_operation_status.csv; after removing missing values, 606 valid records; median = 11.82%.",
+    "Calculate difference: 15.59% - 11.82% = 3.77 percentage points; NetEase R&D intensity is 3.77 percentage points above industry median."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "NetEase 2022 annual report (SEC 20-F filing): total revenue 96,495,809 thousand CNY, R&D expenses 15,039,014 thousand CNY",
+    "company_profile.csv: 644 companies in information transmission, software and IT services industry",
+    "company_operation_status.csv: 606 records of 2022 R&D intensity for above companies, median 11.82%"
+  ],
+  "milestone": {
+    "NetEase 2022 revenue(thousand CNY)": 96495809,
+    "NetEase 2022 R&D expenses(thousand CNY)": 15039014,
+    "NetEase R&D intensity(%)": 15.59,
+    "Industry median R&D intensity(%)": 11.82,
+    "Difference(percentage points)": 3.77
+  },
+  "answer": [
+    15.59,
+    3.77
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium002_result.json b/assets/qa_raw/international_comparison/medium002_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ece4437034bee4f971ed285d8c74b5aa7f3370a
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium002_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium002",
+  "question": "What is the net profit margin (net profit / operating revenue × 100%) of Trip.com Group according to its 2022 annual report? What is the difference in percentage points compared with the median net profit margin of listed companies in China's transport, storage and postal services industry?",
+  "guidelines": "Answer in order: Trip.com net profit margin (%), and the difference from domestic industry median (percentage points; negative value means Trip.com is below industry median). Values rounded to 2 decimal places, returned as an array, e.g. [-3.50, -1.20]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from Trip.com Group (TCOM) 2022 20-F annual report, Selected Consolidated Statements of Income/(Loss) Data section: Net revenues = RMB 20,039 million, Net income = RMB 1,367 million. Net profit margin = 1,367 / 20,039 × 100% = 6.82%.",
+    "Filter company_profile.csv for companies with industry=\"transport, storage and postal services\" (transport, storage and postal services), 176 companies.",
+    "Join the 176 companies with company_operation_status.csv by bmCode, obtain net profit and operating revenue for each, compute net profit margin (net profit / operating revenue × 100%); all 176 have valid data. Median net profit margin = 8.28%.",
+    "Calculate difference between Trip.com and domestic transport industry median: 6.82% - 8.28% = -1.46 percentage points; Trip.com net profit margin is below domestic industry median."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "Trip.com Group (TCOM) 2022 20-F annual report (SEC Filing), including 2022 consolidated net revenues RMB 20,039 million, consolidated net profit RMB 1,367 million.",
+    "company_profile.csv: 176 companies in transport, storage and postal services industry.",
+    "company_operation_status.csv: net profit and operating revenue for the 176 transport industry companies."
+  ],
+  "milestone": {
+    "Trip.com 2022 net revenue(RMB million)": 20039,
+    "Trip.com 2022 net profit(RMB million)": 1367,
+    "Trip.com net profit margin(%)": 6.82,
+    "Domestic transport industry company count": 176,
+    "Domestic transport industry median net profit margin(%)": 8.28,
+    "Net profit margin difference(percentage points)": -1.46
+  },
+  "answer": [
+    6.82,
+    -1.46
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium003_result.json b/assets/qa_raw/international_comparison/medium003_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2a3827c68e38ae0d796268d16018e88976bd497c
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium003_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium003",
+  "question": "What is the revenue per employee (operating revenue / total employees, unit: ten thousand CNY) of Bilibili according to its 2022 annual report? Compared with the median revenue per employee of listed companies in China's \"information transmission, software and IT services\" industry, how many times the industry median is Bilibili's revenue per employee?",
+  "guidelines": "Answer in order: Bilibili revenue per employee (ten thousand CNY), and the multiple of industry median. Values rounded to 2 decimal places, e.g. [197.43, 2.16]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from Bilibili (BILI) 2022 20-F annual report, Consolidated Results of Operations: 2022 Net revenues = RMB 21,899,167 thousand (i.e. 21.899 billion); from Item 6D. Employees, total employees as of Dec 31, 2022 = 11,092. Revenue per employee = 21,899,167,000 / 11,092 / 10,000 = 197.43 ten thousand CNY.",
+    "Filter company_profile.csv for companies with industry==\"information transmission, software and information technology services\" (information transmission, software and IT services), 644 companies.",
+    "Join the 644 companies with company_operation_status.csv by company name; obtain operating revenue and total employees for each. After excluding missing values and records with zero employees, 642 valid companies remain. Revenue per employee = operating revenue / total employees / 10000 (ten thousand CNY); median = 91.24 ten thousand CNY.",
+    "Calculate multiple of Bilibili revenue per employee vs industry median: 197.43 / 91.24 = 2.16x."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "Bilibili (BILI) 2022 20-F annual report (SEC Filing), including 2022 consolidated net revenues RMB 21,899,167 thousand, total employees 11,092.",
+    "company_profile.csv: 644 companies in information transmission, software and IT services industry.",
+    "company_operation_status.csv: operating revenue and total employees for 642 valid companies; revenue per employee computed."
+  ],
+  "milestone": {
+    "Bilibili 2022 net revenue(RMB thousand)": 21899167,
+    "Bilibili 2022 total employees": 11092,
+    "Bilibili revenue per employee(ten thousand CNY)": 197.43,
+    "Domestic IT industry valid company count": 642,
+    "Domestic IT industry median revenue per employee(ten thousand CNY)": 91.24,
+    "Multiple": 2.16
+  },
+  "answer": [
+    197.43,
+    2.16
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium004_result.json b/assets/qa_raw/international_comparison/medium004_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dc1685a1a0bdb50cb7921af281e40a6fe763803
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium004_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium004",
+  "question": "What is the net profit margin (net profit ÷ operating revenue × 100%) of Vipshop according to its 2022 annual report? How many percentage points higher is it compared with the median net profit margin of listed companies in China's wholesale and retail industry?",
+  "guidelines": "Answer in order: Vipshop net profit margin (%, 2 decimal places) and the difference from domestic industry median (percentage points, 2 decimal places; positive value means Vipshop is higher), e.g. [6.12, 4.57]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from Vipshop 2022 20-F annual report PDF, Consolidated Statements of Income: 2022 Total net revenues = RMB 103,152,489 thousand, Net income = RMB 6,311,835 thousand. Net profit margin = 6,311,835 / 103,152,489 × 100% = 6.12%.",
+    "Filter company_profile.csv for companies with industry=='wholesale and retail', 273 companies.",
+    "Obtain operating revenue and net profit for the 273 companies from company_operation_status.csv; compute net profit margin (net profit ÷ revenue × 100%) for each; all 273 have valid data. Median net profit margin = 1.55%.",
+    "Calculate difference between Vipshop and domestic wholesale and retail industry median: 6.12% - 1.55% = 4.57 percentage points; Vipshop is significantly above domestic industry median."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "Vipshop 2022 20-F annual report (SEC Filing), including 2022 Total net revenues RMB 103,152,489 thousand, Net income RMB 6,311,835 thousand.",
+    "company_profile.csv: 273 companies in wholesale and retail industry.",
+    "company_operation_status.csv: operating revenue and net profit for 273 wholesale and retail companies; net profit margin computed."
+  ],
+  "milestone": {
+    "Vipshop 2022 total net revenue(RMB thousand)": 103152489,
+    "Vipshop 2022 net profit(RMB thousand)": 6311835,
+    "Vipshop net profit margin(%)": 6.12,
+    "Domestic wholesale and retail industry company count": 273,
+    "Domestic wholesale and retail industry median net profit margin(%)": 1.55,
+    "Net profit margin difference(percentage points)": 4.57
+  },
+  "answer": [
+    6.12,
+    4.57
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium005_result.json b/assets/qa_raw/international_comparison/medium005_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0caa7bb95ed465964326216241c4ce902acaca6b
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium005_result.json
@@ -0,0 +1,35 @@
+{
+  "id": "medium005",
+  "question": "ZTO Express is a profit leader in China's express delivery industry. Based on ZTO Express 2022 annual report, answer in order: (1) ZTO Express 2022 net profit per employee (net profit ÷ total employees, unit: ten thousand CNY); (2) Median net profit per employee of listed companies in China's transport, storage and postal services industry (ten thousand CNY); (3) How many times the industry median is ZTO Express net profit per employee?",
+  "guidelines": "Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [26.76, 13.26, 2.02]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from ZTO Express 2022 20-F annual report PDF: 2022 consolidated net profit (Net income) = RMB 6,658,966 thousand (approx. 6.659 billion), total employees (Employees) = 24,888.",
+    "Calculate ZTO net profit per employee: 6,658,966,000 CNY / 24,888 persons / 10,000 = 26.76 ten thousand CNY per person.",
+    "Filter company_profile.csv for industry=='transport, storage and postal services' (transport, storage and postal services), 176 companies. Obtain net profit and total employees for these 176 from company_operation_status.csv; after excluding missing values, 174 valid. Net profit per employee = net profit / total employees / 10000; median = 13.26 ten thousand CNY.",
+    "Calculate multiple: 26.76 / 13.26 = 2.02x. ZTO Express net profit per employee is about 2 times the domestic transport, storage and postal services industry median."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "ZTO Express 2022 20-F annual report, including 2022 consolidated net profit RMB 6,658,966 thousand, total employees 24,888.",
+    "company_profile.csv: 176 companies in transport, storage and postal services industry.",
+    "company_operation_status.csv: net profit and employee data for 174 companies; median net profit per employee computed."
+  ],
+  "milestone": {
+    "ZTO Express 2022 net profit(RMB thousand)": 6658966,
+    "ZTO Express 2022 total employees": 24888,
+    "ZTO Express net profit per employee(ten thousand CNY)": 26.76,
+    "Transport storage postal industry valid company count": 174,
+    "Industry median net profit per employee(ten thousand CNY)": 13.26,
+    "ZTO vs industry median multiple": 2.02
+  },
+  "answer": [
+    26.76,
+    13.26,
+    2.02
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium006_result.json b/assets/qa_raw/international_comparison/medium006_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..48836066be7c43e0f38edc248c3a2885fa2cac37
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium006_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium006",
+  "question": "For MINISO (MNSO) FY2022 (ended June 30, 2022), answer in order: (1) What is the total asset turnover ratio? (2) What is the median total asset turnover ratio of listed companies in China's wholesale and retail industry in 2022? (3) What is the difference between the two (positive value means industry median is higher)?",
+  "guidelines": "Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [0.89, 1.02, 0.13]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from MINISO (MNSO) FY2022 annual report (20-F): Revenue = RMB 10,085,649 thousand, Total assets = RMB 11,281,788 thousand. Total asset turnover = 10,085,649 / 11,281,788 = 0.89.",
+    "Filter company_profile.csv for industry=\"wholesale and retail\", 273 companies.",
+    "Obtain 2022 operating revenue and total assets for the 273 companies from company_operation_status.csv; after excluding invalid data, compute total asset turnover = operating revenue / total assets for each company.",
+    "Median total asset turnover of domestic wholesale and retail industry = 1.02.",
+    "Calculate difference: 1.02 - 0.89 = 0.13; domestic wholesale and retail industry median is 0.13 higher than MINISO."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "MINISO (MNSO) FY2022 20-F annual report (SEC Filing), including revenue RMB 10,085,649 thousand, total assets RMB 11,281,788 thousand.",
+    "company_profile.csv: 273 companies in wholesale and retail industry.",
+    "company_operation_status.csv: operating revenue and total assets for 273 wholesale and retail companies; median total asset turnover computed."
+  ],
+  "milestone": {
+    "MINISO operating revenue(RMB thousand)": 10085649,
+    "MINISO total assets(RMB thousand)": 11281788,
+    "MINISO total asset turnover": 0.89,
+    "Wholesale and retail industry company count": 273,
+    "Wholesale and retail industry median total asset turnover": 1.02,
+    "Difference": 0.13
+  },
+  "answer": [
+    0.89,
+    1.02,
+    0.13
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium007_result.json b/assets/qa_raw/international_comparison/medium007_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..55827b62f48ff85532b632ecaee8e4ba2d69c82a
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium007_result.json
@@ -0,0 +1,35 @@
+{
+  "id": "medium007",
+  "question": "H World Group (HTHT) is one of China's leading hotel groups. Based on H World Group 2022 annual report, answer in order: (1) Revenue per employee (total operating revenue ÷ total employees, unit: ten thousand CNY); (2) Median revenue per employee of listed companies in China's accommodation and catering industry (ten thousand CNY); (3) How many times the industry median is H World Group's revenue per employee?",
+  "guidelines": "Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [56.96, 41.29, 1.38]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from H World Group 2022 20-F annual report: 2022 Total revenues = RMB 13,862 million (approx. 13.862 billion), total employees (Employees) = 24,335.",
+    "Calculate H World Group revenue per employee: 13,862,000,000 CNY / 24,335 persons / 10,000 = 56.96 ten thousand CNY per person.",
+    "Filter company_profile.csv for industry=\"accommodation and catering\", 35 companies. Obtain operating revenue and total employees from company_operation_status.csv; after excluding invalid data, 32 valid. Revenue per employee = operating revenue / total employees / 10000; median = 41.29 ten thousand CNY.",
+    "Calculate multiple: 56.96 / 41.29 = 1.38x. H World Group revenue per employee is about 1.38 times the domestic accommodation and catering industry median."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "H World Group 2022 20-F annual report, including 2022 total operating revenue RMB 13,862 million, total employees 24,335.",
+    "company_profile.csv: 35 companies in accommodation and catering industry.",
+    "company_operation_status.csv: operating revenue and employee data for 32 companies; median revenue per employee computed."
+  ],
+  "milestone": {
+    "H World Group 2022 total operating revenue(RMB million)": 13862,
+    "H World Group 2022 total employees": 24335,
+    "H World Group revenue per employee(ten thousand CNY)": 56.96,
+    "Accommodation and catering industry valid company count": 32,
+    "Industry median revenue per employee(ten thousand CNY)": 41.29,
+    "H World Group vs industry median multiple": 1.38
+  },
+  "answer": [
+    56.96,
+    41.29,
+    1.38
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium008_result.json b/assets/qa_raw/international_comparison/medium008_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc808241c28bb7fe8183f485e8b1dd0e3e590926
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium008_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium008",
+  "question": "Yum China (YUMC) is one of China's largest restaurant chains, with brands including KFC and Pizza Hut. Based on Yum China 2022 annual report, answer in order: (1) Net profit margin (net profit ÷ operating revenue, %); (2) Median net profit margin of listed companies in China's accommodation and catering industry (%); (3) How many percentage points higher is Yum China's net profit margin than that median?",
+  "guidelines": "Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [4.62, -16.17, 20.79]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from Yum China 2022 annual report (10-K) consolidated income statement: 2022 Total revenues = 9,569 million USD, Net Income – Yum China Holdings, Inc. = 442 million USD.",
+    "Calculate Yum China net profit margin: 442 / 9,569 × 100% = 4.62%.",
+    "Filter company_profile.csv for accommodation and catering industry, 35 companies. Join with company_operation_status.csv for net profit and operating revenue; after excluding records with zero or missing revenue, compute net profit margin (net profit / operating revenue × 100%) for each; median = -16.17%.",
+    "Calculate difference: 4.62% - (-16.17%) = 20.79 percentage points. Yum China net profit margin is significantly above the domestic accommodation and catering industry median."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "Yum China (YUMC) 2022 10-K annual report, including consolidated income statement: Total revenues 9,569 million USD, Net Income 442 million USD.",
+    "company_profile.csv: 35 accommodation and catering companies; company_operation_status.csv provides net profit and operating revenue; median net profit margin computed."
+  ],
+  "milestone": {
+    "Yum China 2022 operating revenue(million USD)": 9569,
+    "Yum China 2022 net profit(million USD)": 442,
+    "Yum China net profit margin(%)": 4.62,
+    "Accommodation and catering industry valid company count": 35,
+    "Industry median net profit margin(%)": -16.17,
+    "Yum China above industry median(percentage points)": 20.79
+  },
+  "answer": [
+    4.62,
+    -16.17,
+    20.79
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium009_result.json b/assets/qa_raw/international_comparison/medium009_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a668df4a384188251b85a9e2d4dbbd11d51ac25a
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium009_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium009",
+  "question": "What is ASE Technology Holding's net profit margin in 2022? Among listed companies in China's semiconductor industry, the top 10% by operating revenue (count rounded up) are defined as industry leaders. What is ASE Technology Holding's net profit margin? How many percentage points does it differ from the median net profit margin of industry leaders?",
+  "guidelines": "Answer in order: 1) ASE Technology Holding's 2022 net profit margin (2 decimal places, %); 2) Difference between ASE's net profit margin and the median net profit margin of domestic semiconductor industry leaders (top 10% by revenue) in percentage points (2 decimal places). Return as a list. If relevant data cannot be found, answer \"No relevant data found\".",
+  "steps": [
+    "From ASE Technology Holding's 2022 annual report (Form 20-F), extract consolidated income statement data: 2022 Net revenues = 670,872,643 thousand TWD, Net profit attributable to owners of the parent = 61,501,545 thousand TWD. Net profit margin = 61,501,545 / 670,872,643 × 100 = 9.17%.",
+    "From company_profile.csv, filter industry = \"semiconductor industry\", 172 companies.",
+    "From company_operation_status.csv, obtain 2022 net profit and operating revenue for these companies; 172 valid records after excluding invalid data. Sort by operating revenue descending; take the top 10% (18 companies, rounded up) as industry leaders, remaining 154 as non-leaders.",
+    "Compute median net profit margins for industry leaders and non-leaders separately; industry leaders median = 9.60%.",
+    "Compute the difference between ASE and the industry leaders' median: 9.17% − 9.60% = −0.43 percentage points; ASE's net profit margin is 0.43 percentage points below the median of domestic semiconductor industry leaders."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "ASE Technology Holding 2022 annual report (SEC Form 20-F filing): Net revenues 670,872,643 thousand TWD, net profit attributable to owners of the parent 61,501,545 thousand TWD",
+    "company_profile.csv: 172 semiconductor industry companies",
+    "company_operation_status.csv: 172 net profit margin records for 2022; industry leaders (top 10%, 18 companies) median 9.60%"
+  ],
+  "milestone": {
+    "ASE Technology Holding 2022 operating revenue (thousand TWD)": 670872643,
+    "ASE Technology Holding 2022 net profit attributable to owners of the parent (thousand TWD)": 61501545,
+    "ASE Technology Holding net profit margin (%)": 9.17,
+    "Semiconductor industry total enterprise count": 172,
+    "Industry leader enterprise count (top 10%, rounded up)": 18,
+    "Industry leaders median net profit margin (%)": 9.6,
+    "Gap between ASE Technology Holding and industry leaders median (percentage points)": -0.43
+  },
+  "answer": [
+    9.17,
+    -0.43
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium010_result.json b/assets/qa_raw/international_comparison/medium010_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d60eba1776d4b314a5f2ee3d170d23c3b1fb534c
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium010_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium010",
+  "question": "What was Futu Holdings' (FUTU) net profit per employee in 2022, in ten thousand CNY? What is the multiple of Futu's net profit per employee compared with the median net profit per employee of listed companies in China's capital market services industry (securities and diversified financials)?",
+  "guidelines": "Answer both sub-questions: 1) Futu Holdings 2022 net profit per employee (2 decimal places, ten thousand CNY; convert HKD to CNY using 2022 average rate 1 HKD ≈ 0.86 RMB); 2) Multiple of Futu net profit per employee vs domestic capital market services industry median (2 decimal places). If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from Futu Holdings 2022 annual report (20-F) consolidated financial data: 2022 net profit attributable to shareholders = 2,926,944 thousand HKD, total employees = 2,784.",
+    "Convert net profit to CNY: 2,926,944 × 1000 × 0.86 = 2,517,171,840 CNY; net profit per employee = 2,517,171,840 / 2,784 / 10,000 = 90.42 ten thousand CNY.",
+    "Filter company_profile.csv for capital market services with secondary industry \"securities\" and \"diversified financials\", 202 companies.",
+    "Obtain 2022 net profit and total employees for these companies from company_operation_status.csv; after excluding invalid data, 201 valid records. Compute net profit per employee for each; median = 16.09 ten thousand CNY.",
+    "Calculate multiple: 90.42 / 16.09 = 5.62x; Futu net profit per employee is about 5.62 times the domestic capital market services industry median."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Futu Holdings 2022 annual report (SEC 20-F filing): net profit attributable to shareholders 2,926,944 thousand HKD, total employees 2,784",
+    "company_profile.csv: 202 companies in securities and diversified financials industry",
+    "company_operation_status.csv: 201 valid 2022 net profit per employee records for above companies, median 16.09 ten thousand CNY"
+  ],
+  "milestone": {
+    "Futu 2022 net profit(thousand HKD)": 2926944,
+    "Futu 2022 employee count": 2784,
+    "Futu net profit per employee(ten thousand CNY)": 90.42,
+    "Domestic capital market services median net profit per employee(ten thousand CNY)": 16.09,
+    "Multiple": 5.62
+  },
+  "answer": [
+    90.42,
+    5.62
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium011_result.json b/assets/qa_raw/international_comparison/medium011_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c730644107d5e384c0a7724d90446d2a41e5cd8
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium011_result.json
@@ -0,0 +1,35 @@
+{
+  "id": "medium011",
+  "question": "Atour Lifestyle Group (ATAT) is a representative mid-to-high-end chain hotel brand in China. Based on Atour 2022 annual report, answer in order: (1) Asset-liability ratio (total liabilities ÷ total assets, %); (2) Median asset-liability ratio of listed companies in China's accommodation and catering industry (%); (3) How many percentage points higher is Atour's asset-liability ratio than that median?",
+  "guidelines": "Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [75.07, 63.83, 11.24]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from Atour Lifestyle Group (ATAT) 2022 annual report PDF consolidated balance sheet: 2022 year-end Total liabilities = RMB 3,574,620 thousand, Total assets = RMB 4,762,026 thousand.",
+    "Calculate Atour asset-liability ratio: 3,574,620 / 4,762,026 × 100% = 75.07%.",
+    "Filter company_profile.csv for industry=\"accommodation and catering\", 35 companies. Obtain 2022 asset-liability ratio for these 35 from company_operation_status.csv; after excluding missing values, 35 valid; median = 63.83%.",
+    "Calculate difference: 75.07% - 63.83% = 11.24 percentage points. Atour's asset-liability ratio is 11.24 percentage points above the domestic accommodation and catering industry median, mainly affected by operating lease liabilities (IFRS 16/ASC 842) recognition."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "Atour Lifestyle Group (ATAT) 2022 20-F annual report, including consolidated balance sheet: total liabilities RMB 3,574,620 thousand, total assets RMB 4,762,026 thousand.",
+    "company_profile.csv: 35 companies in accommodation and catering industry.",
+    "company_operation_status.csv: 2022 asset-liability ratio for 35 companies; industry median computed."
+  ],
+  "milestone": {
+    "Atour 2022 total liabilities(RMB thousand)": 3574620,
+    "Atour 2022 total assets(RMB thousand)": 4762026,
+    "Atour asset-liability ratio(%)": 75.07,
+    "Accommodation and catering industry valid company count": 35,
+    "Industry median asset-liability ratio(%)": 63.83,
+    "Atour above industry median(percentage points)": 11.24
+  },
+  "answer": [
+    75.07,
+    63.83,
+    11.24
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium012_result.json b/assets/qa_raw/international_comparison/medium012_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..4d0b884f7901ed24892e65b7dc5fefc480eecbbe
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium012_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium012",
+  "question": "Full Truck Alliance (NYSE: YMM) is a leading digital freight platform in China. Based on Full Truck Alliance 2022 annual report, answer in order: (1) R&D intensity (R&D expenses ÷ operating revenue, %); (2) Median R&D intensity of listed companies in China's transport, storage and postal services industry (%); (3) How many times the industry median is Full Truck Alliance's R&D intensity?",
+  "guidelines": "Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [13.58, 0.56, 24.25]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from Full Truck Alliance 2022 20-F annual report: 2022 Net revenues = RMB 6,733,644 thousand (approx. 6.734 billion), Research and development expenses = RMB 914,151 thousand (approx. 914 million).",
+    "Calculate Full Truck Alliance R&D intensity: 914,151 / 6,733,644 × 100% = 13.58%.",
+    "Filter company_profile.csv for industry=='transport, storage and postal services' (transport, storage and postal services), 176 companies. Obtain R&D intensity for these 176 from company_operation_status.csv; after excluding missing and invalid data, 107 companies have valid R&D intensity.",
+    "Industry median R&D intensity: median for the 107 companies = 0.56%.",
+    "Calculate multiple: 13.58 / 0.56 = 24.25x. As a digital platform company, Full Truck Alliance's R&D intensity is far above traditional transport, storage and postal services listed companies."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Full Truck Alliance 2022 20-F annual report (Form 20-F), including 2022 net operating revenue RMB 6,733,644 thousand, R&D expenses RMB 914,151 thousand.",
+    "company_profile.csv: 176 companies in transport, storage and postal services industry.",
+    "company_operation_status.csv: valid R&D intensity for 107 companies; median computed."
+  ],
+  "milestone": {
+    "Full Truck Alliance 2022 operating revenue(RMB thousand)": 6733644,
+    "Full Truck Alliance 2022 R&D expenses(RMB thousand)": 914151,
+    "Full Truck Alliance R&D intensity(%)": 13.58,
+    "Transport storage postal industry valid company count": 107,
+    "Industry median R&D intensity(%)": 0.56,
+    "Full Truck Alliance vs industry median multiple": 24.25
+  },
+  "answer": [
+    13.58,
+    0.56,
+    24.25
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium013_result.json b/assets/qa_raw/international_comparison/medium013_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..45d4de41c53f8da48418bee50d8abe7d7719e699
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium013_result.json
@@ -0,0 +1,35 @@
+{
+  "id": "medium013",
+  "question": "Himax Technologies (HIMX) is a leading display driver IC design company globally. Based on Himax 2022 annual report, answer in order: (1) R&D intensity (R&D expenses as percentage of operating revenue, %); (2) Median R&D intensity of listed semiconductor companies in China (%); (3) How many times the domestic semiconductor industry median is Himax's R&D intensity?",
+  "guidelines": "Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [14.61, 7.17, 2.04]. Reasoning may be shown in the process, but final answer should include only these three items. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from Himax Technologies (HIMX) 2022 20-F annual report consolidated income statement: 2022 Total revenues = $1,201,339 thousand USD, Research and development = $175,557 thousand USD.",
+    "Calculate Himax R&D intensity: $175,557 thousand / $1,201,339 thousand × 100% = 14.61%.",
+    "Filter company_profile.csv for industry=\"semiconductor industry\" (semiconductors), 172 companies. Join with company_operation_status.csv for R&D intensity; after excluding invalid data, 169 valid. Industry median R&D intensity = 7.17%.",
+    "Calculate multiple: 14.61% / 7.17% = 2.04x. Himax R&D intensity is about 2 times the domestic semiconductor industry median."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "Himax Technologies (HIMX) 2022 20-F annual report, including 2022 operating revenue $1,201,339 thousand USD, R&D expenses $175,557 thousand USD.",
+    "company_profile.csv: 172 semiconductor industry companies.",
+    "company_operation_status.csv: R&D intensity for 169 valid companies; industry median computed."
+  ],
+  "milestone": {
+    "Himax 2022 R&D expenses(thousand USD)": 175557,
+    "Himax 2022 operating revenue(thousand USD)": 1201339,
+    "Himax R&D intensity(%)": 14.61,
+    "Semiconductor industry valid company count": 169,
+    "Industry median R&D intensity(%)": 7.17,
+    "Himax vs industry median multiple": 2.04
+  },
+  "answer": [
+    14.61,
+    7.17,
+    2.04
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium014_result.json b/assets/qa_raw/international_comparison/medium014_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..2fe63e3dc299b0e7f3ea8ee1b921051830c0ee07
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium014_result.json
@@ -0,0 +1,37 @@
+{
+  "id": "medium014",
+  "question": "Weibo (WB / Weibo Corporation) is a leading social media platform in China. Based on Weibo 2022 annual report, compute its net profit margin (net profit attributable to Weibo shareholders ÷ operating revenue, in percentage), and conduct peer analysis against listed companies in China's information transmission, software and IT services industry. Rank by operating revenue; top 10% are defined as industry leaders. Answer in order: Weibo's net profit margin, and the difference (in percentage points) between Weibo's net profit margin and the median net profit margin of industry leaders.",
+  "guidelines": "Answer in order: Weibo 2022 net profit margin (%), and the difference between Weibo and industry leaders (top 10% by revenue) median net profit margin (percentage points). Values rounded to 2 decimal places, e.g. [4.66, 1.21]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from Weibo (WB) 2022 annual report (20-F) consolidated income statement: 2022 Net income attributable to Weibo's shareholders = 85,555 thousand USD, Total net revenues = 1,836,332 thousand USD.",
+    "Calculate Weibo net profit margin: 85,555 / 1,836,332 × 100% = 4.66%.",
+    "Filter company_profile.csv for industry=\"information transmission, software and information technology services\" (information transmission, software and IT services), 644 companies. Join with company_operation_status.csv for 2022 net profit and operating revenue; after excluding invalid data (zero or missing revenue), 644 valid. Compute net profit margin for each.",
+    "Rank by operating revenue descending; top 10% (65 companies) as industry leaders, remaining 579 as non-leaders. Leaders median net profit margin = 3.45%, non-leaders median = 2.25%.",
+    "Calculate difference: 4.66% - 3.45% = 1.21 percentage points. Weibo net profit margin is 1.21 percentage points above domestic IT industry leaders median."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Weibo (WB) 2022 20-F annual report, including consolidated income statement: net profit attributable to Weibo shareholders 85,555 thousand USD, operating revenue 1,836,332 thousand USD.",
+    "company_profile.csv: 644 companies in information transmission, software and IT services industry.",
+    "company_operation_status.csv: 2022 net profit and operating revenue for 644 companies; net profit margin computed; leaders and non-leaders median by revenue ranking."
+  ],
+  "milestone": {
+    "Weibo 2022 net profit attributable to shareholders(thousand USD)": 85555,
+    "Weibo 2022 operating revenue(thousand USD)": 1836332,
+    "Weibo net profit margin(%)": 4.66,
+    "IT industry valid company count": 644,
+    "Industry leaders count": 65,
+    "Industry leaders median net profit margin(%)": 3.45,
+    "Non-leaders median net profit margin(%)": 2.25,
+    "Weibo vs leaders median difference(percentage points)": 1.21
+  },
+  "answer": [
+    4.66,
+    1.21
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium015_result.json b/assets/qa_raw/international_comparison/medium015_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..610078bad95e823e209832e4e52ea868a19eaabc
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium015_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium015",
+  "question": "Delta Electronics (2308.TW) is a leading global provider of power and thermal management solutions. Based on Delta Electronics' 2022 annual report, compute its consolidated net profit margin (net profit / operating revenue × 100%) and benchmark against listed companies in China's Electrical Machinery and Equipment Manufacturing industry: the top 10% by operating revenue are defined as industry leaders. What is Delta Electronics' net profit margin? What is the gap (in percentage points) between Delta's net profit margin and the median net profit margin of industry leaders?",
+  "guidelines": "Answer in order: Delta Electronics 2022 consolidated net profit margin (%), and the gap (percentage points) between Delta Electronics and the median net profit margin of domestic Electrical Machinery and Equipment Manufacturing industry leaders (top 10% by revenue). Round all values to 2 decimal places, e.g. [9.62, 3.30]. If relevant data cannot be found, answer \"No relevant data found\".",
+  "steps": [
+    "From Delta Electronics' 2022 annual report (ROC calendar year 111), consolidated condensed comprehensive income statement: consolidated operating revenue = 384,443,308 thousand TWD, consolidated net profit for the period = 36,990,738 thousand TWD.",
+    "Compute Delta consolidated net profit margin: 36,990,738 / 384,443,308 × 100% = 9.62%.",
+    "From company_profile.csv, filter industry = \"Electrical Machinery and Equipment Manufacturing\", 320 companies. Join company_operation_status.csv for 2022 net profit and operating revenue; compute each company's net profit margin (net profit / operating revenue × 100%); all 320 records valid.",
+    "Sort by operating revenue descending; top 10% (32 companies) as industry leaders, remaining 288 as non-leaders. Median net profit margin of industry leaders = 6.32%.",
+    "Compute the gap between Delta's net profit margin and the industry leaders' median: 9.62% − 6.32% = 3.30 percentage points. Delta's net profit margin is above the domestic industry leaders' median."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Delta Electronics 2022 annual report (ROC year 111), consolidated condensed comprehensive income statement: consolidated operating revenue 384,443,308 thousand TWD, consolidated net profit for the period 36,990,738 thousand TWD.",
+    "company_profile.csv: 320 companies in Electrical Machinery and Equipment Manufacturing.",
+    "company_operation_status.csv: 2022 net profit and operating revenue for 320 companies; net profit margin computed; industry leaders vs non-leaders distinguished."
+  ],
+  "milestone": {
+    "Delta Electronics 2022 consolidated operating revenue (thousand TWD)": 384443308,
+    "Delta Electronics 2022 consolidated net profit (thousand TWD)": 36990738,
+    "Delta Electronics net profit margin (%)": 9.62,
+    "Electrical Machinery and Equipment Manufacturing valid enterprise count": 320,
+    "Industry leader enterprise count": 32,
+    "Industry leaders median net profit margin (%)": 6.32,
+    "Gap between Delta Electronics and industry leaders median (percentage points)": 3.3
+  },
+  "answer": [
+    9.62,
+    3.3
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium016_result.json b/assets/qa_raw/international_comparison/medium016_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d8f6f0fcae4a388aa7f0aae6f9e236c392e6f1e5
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium016_result.json
@@ -0,0 +1,35 @@
+{
+  "id": "medium016",
+  "question": "In 2022, amid strong policy support for domestic semiconductor self-reliance in China, what was TSMC's R&D intensity (R&D expenses as percentage of revenue) in its annual report? What is the difference in percentage points compared with the median R&D intensity of listed semiconductor companies in China?",
+  "guidelines": "Answer in order: TSMC R&D intensity (%), and the difference from domestic median (percentage points; positive value means TSMC is higher). Values rounded to 2 decimal places. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from TSMC 2022 20-F annual report PDF: 2022 Net Revenue = NT$2,263,891 million, Research and Development Expenses = NT$163,262 million. R&D intensity = 163,262 / 2,263,891 × 100% = 7.21%.",
+    "Filter company_profile.csv for industry=='semiconductor industry' (semiconductors), 172 companies.",
+    "Obtain 'R&D intensity' for the 172 semiconductor companies from company_operation_status.csv; 169 have valid data. Median = 7.17%.",
+    "Calculate difference between TSMC and domestic semiconductor industry median: 7.21% - 7.17% = 0.04 percentage points; TSMC is slightly above domestic median.",
+    "TSMC R&D intensity (7.21%) differs very little from domestic semiconductor industry median (7.17%)—only 0.04 percentage points."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "TSMC 2022 20-F annual report (SEC Filing), including 2022 net revenue NT$2,263,891 million, R&D expenses NT$163,262 million.",
+    "company_profile.csv: 172 semiconductor industry companies.",
+    "company_operation_status.csv: R&D intensity for 169 semiconductor companies."
+  ],
+  "milestone": {
+    "TSMC 2022 net revenue(NT$ million)": 2263891,
+    "TSMC 2022 R&D expenses(NT$ million)": 163262,
+    "TSMC R&D intensity(%)": 7.21,
+    "Domestic semiconductor industry company count": 172,
+    "Domestic semiconductor industry median R&D intensity(%)": 7.17,
+    "R&D intensity difference(percentage points)": 0.04
+  },
+  "answer": [
+    7.21,
+    0.04
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium017_result.json b/assets/qa_raw/international_comparison/medium017_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..959e4e216b22ac12ce7155b8f8bd770dc63d0140
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium017_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "medium017",
+  "question": "In 2022, under national policies promoting innovative drug R&D, what was BeiGene's R&D expense as a percentage of revenue in its annual report? How many percentage points does that ratio differ from the median R&D-to-revenue ratio among domestic pharmaceutical manufacturing firms that are (1) private enterprises and (2) state-owned enterprises (including centrally administered SOEs, locally administered SOEs, and institute-type SOEs), after excluding zeros and invalid data? What is the gap between the median R&D ratios of private vs. state-owned domestic pharmaceutical firms?",
+  "guidelines": "Answer in order: BeiGene's R&D-to-revenue ratio (%); gap vs. private enterprises' median (percentage points); gap vs. state-owned enterprises' median (percentage points); gap between private and state-owned medians (percentage points). Use 2 decimal places; return as a list. If relevant data cannot be found, answer \"No relevant data found\".",
+  "steps": [
+    "From BeiGene's 2022 annual report (PDF), Consolidated Statements of Operations: Total revenues = $1,415,921 thousand USD, Research and development expense = $1,640,508 thousand USD. R&D-to-revenue ratio = 1,640,508 / 1,415,921 × 100 = 115.86%.",
+    "From company_profile.csv, filter industry = \"Pharmaceutical Manufacturing\", 449 companies. By ownership: private enterprises 346; state-owned enterprises (centrally administered SOEs 16 + locally administered SOEs 49 + institute-type SOEs 2) = 67 in total.",
+    "Join company_operation_status.csv; filter valid R&D-to-revenue ratios > 0. Private enterprises: 326 valid samples, median R&D ratio = 7.46%. State-owned enterprises: 66 valid samples, median R&D ratio = 4.63%.",
+    "Compute gaps vs. benchmarks: BeiGene vs. private median = 115.86% − 7.46% = 108.40 percentage points; BeiGene vs. state-owned median = 115.86% − 4.63% = 111.23 percentage points.",
+    "Gap between private and state-owned medians = 7.46% − 4.63% = 2.83 percentage points; private enterprises show higher R&D intensity than state-owned enterprises."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "BeiGene 2022 annual report PDF (Consolidated Statements of Operations): Total revenues = $1,415,921 thousand USD, R&D expense = $1,640,508 thousand USD.",
+    "company_profile.csv: 449 Pharmaceutical Manufacturing companies; 346 private enterprises, 67 state-owned enterprises.",
+    "company_operation_status.csv: R&D-to-revenue ratios — 326 valid private samples, 66 valid state-owned samples."
+  ],
+  "milestone": {
+    "BeiGene Total revenues (thousand USD)": 1415921,
+    "BeiGene R&D expense (thousand USD)": 1640508,
+    "BeiGene R&D-to-revenue ratio (%)": 115.86,
+    "Private enterprises valid sample count": 326,
+    "Private enterprises median R&D-to-revenue ratio (%)": 7.46,
+    "State-owned enterprises valid sample count": 66,
+    "State-owned enterprises median R&D-to-revenue ratio (%)": 4.63,
+    "BeiGene vs. private median gap (percentage points)": 108.4,
+    "BeiGene vs. state-owned median gap (percentage points)": 111.23,
+    "Private vs. state-owned median gap (percentage points)": 2.83
+  },
+  "answer": [
+    115.86,
+    108.4,
+    111.23,
+    2.83
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium018_result.json b/assets/qa_raw/international_comparison/medium018_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..d4cf313b03cc90b529b58e0836b3bd4b7181a8d9
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium018_result.json
@@ -0,0 +1,35 @@
+{
+  "id": "medium018",
+  "question": "In 2022, amid China's new energy vehicle industry policy support, what was XPeng's net profit margin (net profit ÷ operating revenue × 100%) in its annual report? What is the difference in percentage points compared with the median net profit margin of domestic automotive manufacturing industry leaders (top 10% by revenue)?",
+  "guidelines": "Answer in order: XPeng net profit margin (%) and the difference from industry leaders median (percentage points; negative value means XPeng is lower). Values rounded to 2 decimal places, e.g. [-34.03, -37.27]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from XPeng 2022 annual report (20-F) PDF: Total revenues = RMB 26,855,119 thousand (approx. 26.855 billion), Net loss = RMB 9,138,972 thousand (approx. 9.139 billion). Net profit margin = -9,138,972 / 26,855,119 × 100% = -34.03%.",
+    "Filter company_profile.csv for industry=\"automotive manufacturing\", 230 companies; obtain operating revenue and net profit from company_operation_status.csv; after excluding invalid data, 230 companies have complete financial data.",
+    "Rank by operating revenue descending; take top 10% (23 companies) as industry leaders. Compute net profit margin (net profit / operating revenue × 100%) for each leader; median = 3.24%.",
+    "Calculate difference: XPeng net profit margin (-34.03%) - leaders median (3.24%) = -37.27 percentage points."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "XPeng 2022 annual report (20-F): total revenue RMB 26,855,119 thousand, net loss RMB 9,138,972 thousand.",
+    "company_profile.csv: 230 automotive manufacturing companies.",
+    "company_operation_status.csv: operating revenue and net profit for 230 companies; industry leaders (top 10%) 23 companies."
+  ],
+  "milestone": {
+    "XPeng 2022 operating revenue(hundred million CNY)": 268.55,
+    "XPeng 2022 net loss(hundred million CNY)": 91.39,
+    "XPeng net profit margin(%)": -34.03,
+    "Automotive manufacturing total company count": 230,
+    "Industry leaders count": 23,
+    "Industry leaders median net profit margin(%)": 3.24,
+    "Difference(percentage points)": -37.27
+  },
+  "answer": [
+    -34.03,
+    -37.27
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium019_result.json b/assets/qa_raw/international_comparison/medium019_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e82ce9806cad3be943be6fee021fb013c06f80fb
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium019_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium019",
+  "question": "In 2022, amid policy support for digital economy and e-commerce development in China, what was PDD Holdings' net profit margin (net profit ÷ operating revenue × 100%) in its annual report? How many percentage points higher is it compared with the median net profit margin of listed companies in China's wholesale and retail industry?",
+  "guidelines": "Answer in order: PDD Holdings net profit margin (%), and the number of percentage points above domestic median. Values rounded to 2 decimal places, e.g. [24.16, 22.60]. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from PDD Holdings 2022 20-F annual report PDF: 2022 Total Revenues = RMB 130,557,589 thousand (approx. 130.558 billion), Net Income = RMB 31,538,100 thousand (approx. 31.538 billion). Net profit margin = 31,538,100 / 130,557,589 × 100% = 24.16%.",
+    "Filter company_profile.csv for industry=='wholesale and retail', 273 companies.",
+    "Obtain net profit and operating revenue for the 273 wholesale and retail companies from company_operation_status.csv; compute net profit margin = net profit / operating revenue × 100% for each; all 273 have valid data. Industry median net profit margin = 1.55%.",
+    "Calculate difference between PDD and domestic wholesale and retail industry median: 24.16% - 1.55% = 22.60 percentage points; PDD is significantly above industry median."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "PDD Holdings 2022 20-F annual report: total revenue RMB 130,557,589 thousand, net profit RMB 31,538,100 thousand.",
+    "company_profile.csv: 273 wholesale and retail industry companies.",
+    "company_operation_status.csv: net profit and operating revenue for 273 companies."
+  ],
+  "milestone": {
+    "PDD 2022 total revenue(RMB thousand)": 130557589,
+    "PDD 2022 net profit(RMB thousand)": 31538100,
+    "PDD net profit margin(%)": 24.16,
+    "Wholesale and retail industry company count": 273,
+    "Wholesale and retail industry median net profit margin(%)": 1.55,
+    "Net profit margin difference(percentage points)": 22.6
+  },
+  "answer": [
+    24.16,
+    22.6
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium020_result.json b/assets/qa_raw/international_comparison/medium020_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3dddfac08ff9596b625b00ed4298a25d4780735b
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium020_result.json
@@ -0,0 +1,39 @@
+{
+  "id": "medium020",
+  "question": "In 2022, amid national carbon peaking and clean energy development policies, what was JinkoSolar's asset-liability ratio (total liabilities ÷ total assets × 100%) in its annual report? Compared with the median asset-liability ratios among listed companies in China's Electricity, Heat, Gas and Water Supply industry—separately for state-owned enterprises (including centrally administered SOEs, locally administered SOEs, and other SOEs) and for private enterprises—how many percentage points higher is JinkoSolar's ratio in each case?",
+  "guidelines": "Answer in order: JinkoSolar's asset-liability ratio (%); percentage points above the state-owned median; percentage points above the private enterprise median. Use 2 decimal places; return as an array. If relevant data cannot be found, answer \"No relevant data found\".",
+  "steps": [
+    "From JinkoSolar (JKS) 2022 annual report PDF (English Form 20-F), Consolidated Balance Sheets as of December 31, 2022: Total assets = RMB 108,662,182 thousand, Total liabilities = RMB 81,658,289 thousand.",
+    "Compute JinkoSolar's 2022 asset-liability ratio = 81,658,289 / 108,662,182 × 100% = 75.15%.",
+    "From company_profile.csv, filter industry = \"Electricity, Heat, Gas and Water Supply\", 189 companies. By ownership: state-owned enterprises (centrally administered SOEs 44 + locally administered SOEs 81 + other SOEs 1) = 126; private enterprises 56.",
+    "From company_operation_status.csv, read each company's asset-liability ratio; compute group medians: state-owned median = 61.27%, private enterprise median = 52.58%.",
+    "Compute gaps vs. JinkoSolar: above state-owned median 75.15% − 61.27% = 13.88 percentage points; above private enterprise median 75.15% − 52.58% = 22.57 percentage points."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "JinkoSolar (JKS) 2022 annual report PDF (Form 20-F): total assets RMB 108,662,182 thousand, total liabilities RMB 81,658,289 thousand.",
+    "company_profile.csv: 189 companies in Electricity, Heat, Gas and Water Supply — 126 state-owned, 56 private.",
+    "company_operation_status.csv: asset-liability ratio data for 189 companies."
+  ],
+  "milestone": {
+    "JinkoSolar total assets (thousand RMB)": 108662182,
+    "JinkoSolar total liabilities (thousand RMB)": 81658289,
+    "JinkoSolar asset-liability ratio (%)": 75.15,
+    "State-owned enterprise count (Electricity, Heat, Gas and Water Supply)": 126,
+    "Private enterprise count (Electricity, Heat, Gas and Water Supply)": 56,
+    "State-owned enterprises median asset-liability ratio (%)": 61.27,
+    "Private enterprises median asset-liability ratio (%)": 52.58,
+    "Above state-owned median (percentage points)": 13.88,
+    "Above private enterprise median (percentage points)": 22.57
+  },
+  "answer": [
+    75.15,
+    13.88,
+    22.57
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium021_result.json b/assets/qa_raw/international_comparison/medium021_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c6c8ef9c08f9ca458447cb19d222aaba6d0030a0
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium021_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium021",
+  "question": "In 2022, amid policy support for digital economy development and healthy platform economy regulation in China, what was Alibaba's revenue per employee (operating revenue ÷ total employees) in ten thousand CNY? How many times the median revenue per employee of domestic IT industry leaders (top 10% (round down) by operating revenue in information transmission, software and IT services) is it?",
+  "guidelines": "Answer in order: Alibaba revenue per employee (ten thousand CNY), and the multiple of industry leaders median revenue per employee. Values rounded to 2 decimal places, returned as an array. If relevant data cannot be found, reply \"No relevant data found\".",
+  "steps": [
+    "Extract from Alibaba (BABA) 2022 annual report PDF (20-F, fiscal year ended March 31, 2023, corresponding to calendar year 2022): Consolidated Total Revenue = RMB 868,687 million (i.e. 8,686.87 hundred million), full-time employees as of March 31, 2023 = 235,216.",
+    "Calculate Alibaba revenue per employee = 868,687,000,000 / 235,216 = 3,693,064.34 CNY = 369.31 ten thousand CNY.",
+    "Filter company_profile.csv for industry=\"information transmission, software and information technology services\" (information transmission, software and IT services), 644 companies. Join with company_operation_status.csv; exclude records with empty/zero revenue or employees; 642 valid companies.",
+    "Compute revenue per employee = operating revenue / total employees for each. Rank by operating revenue descending; top 10% (64 companies) as industry leaders. Leaders median revenue per employee = 2,456,389.49 CNY = 245.64 ten thousand CNY.",
+    "Calculate multiple: Alibaba revenue per employee / leaders median = 369.31 / 245.64 = 1.50x."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Alibaba (BABA) 2022 annual report PDF (20-F, fiscal year ended March 31, 2023): operating revenue RMB 868,687 million, total employees 235,216.",
+    "company_profile.csv: 644 IT industry companies, 642 valid.",
+    "company_operation_status.csv: operating revenue and total employees for 642 companies."
+  ],
+  "milestone": {
+    "Alibaba operating revenue(million RMB)": 868687,
+    "Alibaba total employees": 235216,
+    "Alibaba revenue per employee(ten thousand CNY)": 369.31,
+    "IT industry valid company count": 642,
+    "Industry leaders count(top 10%)": 64,
+    "Industry leaders median revenue per employee(ten thousand CNY)": 245.64,
+    "Alibaba vs leaders median multiple": 1.5
+  },
+  "answer": [
+    369.31,
+    1.5
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium022_result.json b/assets/qa_raw/international_comparison/medium022_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c9645e7a1484d1c695dcec4870a9b2b5951e4f75
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium022_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "medium022",
+  "question": "In 2022, against the backdrop of intensive policy rollouts for the new energy vehicle industry, what was the proportion of R&D personnel to total employees in NIO's annual report? By how many percentage points did it differ from the median R&D personnel ratio among private enterprises and state-owned enterprises (including central and local state-owned enterprises) in China's automobile manufacturing industry?",
+  "guidelines": "Answer in order: NIO's R&D personnel ratio (%), the number of percentage points above the private enterprise median, and the number of percentage points above the state-owned enterprise median. Retain 2 decimal places for values. E.g. [37.46, 24.19, 22.02]. If the relevant data cannot be found, answer \"Relevant data not found\"",
+  "steps": [
+    "From NIO's 2022 Annual Report PDF (20-F) employee section (Item 6.D Employees, As of December 31, 2022), extract: Total employees = 26,763; Product and software development = 10,025 (i.e., R&D personnel).",
+    "Calculate NIO's 2022 R&D personnel ratio = 10,025 / 26,763 × 100% = 37.46%.",
+    "From company_profile.csv, filter enterprises with industry \"Automobile Manufacturing\", totaling 230. Group by ownership: 161 private enterprises (149 with R&D personnel ratio data); 48 state-owned enterprises (17 central + 31 local, 43 with data).",
+    "From company_operation_status.csv, read the R&D personnel ratio field for the above enterprises and calculate the median for each group: Private enterprise median = 13.27%; State-owned enterprise median = 15.44%.",
+    "Calculate NIO's gap from each group: 37.46% - 13.27% = 24.19 percentage points above private enterprise median; 37.46% - 15.44% = 22.02 percentage points above state-owned enterprise median."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "NIO 2022 Annual Report PDF (20-F): Total employees 26,763, R&D personnel (Product and software development) 10,025.",
+    "From company_profile.csv: 230 automobile manufacturing enterprises, including 161 private enterprises and 48 state-owned enterprises.",
+    "From company_operation_status.csv: R&D personnel ratio data for 230 enterprises."
+  ],
+  "milestone": {
+    "NIO total employees (persons)": 26763,
+    "NIO R&D personnel (persons)": 10025,
+    "NIO R&D personnel ratio (%)": 37.46,
+    "Automobile manufacturing private enterprises (count)": 161,
+    "Automobile manufacturing private enterprises with data (count)": 149,
+    "Automobile manufacturing state-owned enterprises (count)": 48,
+    "Automobile manufacturing state-owned enterprises with data (count)": 43,
+    "Private enterprise R&D personnel ratio median (%)": 13.27,
+    "State-owned enterprise R&D personnel ratio median (%)": 15.44,
+    "Above private enterprise median (percentage points)": 24.19,
+    "Above state-owned enterprise median (percentage points)": 22.02
+  },
+  "answer": [
+    37.46,
+    24.19,
+    22.02
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium023_result.json b/assets/qa_raw/international_comparison/medium023_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..fef2dcf533f8a3e63a00646715891d10b867aeee
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium023_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium023",
+  "question": "Against the backdrop of the state promoting high-quality development of the digital economy and platform economy, what were JD.com's total revenue and total number of employees disclosed in its 2022 annual report? What is the per capita revenue in 10,000 yuan calculated therefrom? Among industries with no fewer than 30 domestic enterprises (excluding financial, real estate and diversified industries), how many times is JD.com's per capita revenue compared to the industry with the highest median per capita revenue?",
+  "guidelines": "Please answer in order: (1) JD.com's 2022 total revenue (100 million yuan, 2 decimal places); (2) JD.com's 2022 total employees (persons); (3) JD.com's per capita revenue (10,000 yuan, 2 decimal places); (4) The ratio of JD.com's per capita revenue to the industry with highest median per capita revenue (2 decimal places). Return as an array. If the relevant data cannot be found, answer \"Relevant data not found\".",
+  "steps": [
+    "From JD.com's 2022 Annual Report (20-F) PDF, extract financial data: Total net revenues RMB 1,046,236 million (i.e., 1046.236 billion yuan), total employees 450,679 (as of December 31, 2022).",
+    "Calculate JD.com's per capita revenue: 1046.236 billion yuan / 450,679 employees = 232.15 ten thousand yuan per person.",
+    "From company_profile.csv obtain industry classification for each enterprise, merge with company_operation_status.csv, calculate per capita revenue for each enterprise (operating revenue / total employees / 10000, unit: 10,000 yuan). Exclude financial, real estate, and diversified industries; filter industries with enterprise count >= 30.",
+    "For the 36 qualifying industries, rank by median per capita revenue in descending order; confirm the highest industry is metal smelting and rolling, median 335.86 ten thousand yuan (145 enterprises).",
+    "Calculate the ratio of JD.com's per capita revenue to this industry's median: 232.15 / 335.86 = 0.69."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "JD.com 2022 Annual Report (20-F): Total net revenues RMB 1,046,236 million and total employees 450,679.",
+    "company_profile.csv merged with company_operation_status.csv to calculate per capita revenue by industry; after filtering, 36 industries; metal smelting and rolling has the highest median per capita revenue."
+  ],
+  "milestone": {
+    "JD.com 2022 total revenue (100 million yuan)": 10462.36,
+    "JD.com 2022 total employees (persons)": 450679,
+    "JD.com per capita revenue (10,000 yuan)": 232.15,
+    "Qualifying industries count": 36,
+    "Metal smelting and rolling median per capita revenue (10,000 yuan)": 335.86,
+    "Ratio of JD.com per capita revenue to industry median": 0.69
+  },
+  "answer": [
+    10462.36,
+    450679,
+    232.15,
+    0.69
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium024_result.json b/assets/qa_raw/international_comparison/medium024_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e8d9d2cb03bb55b2643dd1f436b58f005b343b77
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium024_result.json
@@ -0,0 +1,42 @@
+{
+  "id": "medium024",
+  "question": "Against the backdrop of national policies promoting high-level AI application and supporting digital economy enterprises in strengthening hard-tech innovation, what was Baidu's (BIDU, listed on NASDAQ) R&D intensity (R&D expenses as a percentage of revenue) in 2022? Compared with A-share listed companies in China's information transmission, software, and information technology services industry, when divided into leading enterprises (revenue ≥ 90th percentile) and non-leading enterprises, by how many percentage points did Baidu's R&D intensity exceed the median R&D ratio of each group?",
+  "guidelines": "Answer in order: Baidu's R&D intensity (%), the number of percentage points above the A-share leading group (revenue ≥ sample 90th percentile) median R&D ratio, and the number of percentage points above the A-share non-leading group. Retain 2 decimal places, return as an array. If the relevant data cannot be found, answer \"Relevant data not found\".",
+  "steps": [
+    "From Baidu (BIDU) 2022 Annual Report PDF (20-F, filed with SEC) Consolidated Statements of Comprehensive Income, extract: Total revenues = RMB 123,675 million (123.675 billion yuan); Research and development = RMB 23,315 million (23.315 billion yuan).",
+    "Calculate Baidu's 2022 R&D intensity = 23,315 / 123,675 × 100% = 18.85%.",
+    "From company_profile.csv, filter A-share listed companies with industry \"Information transmission, software, and information technology services\" and companyType \"Shanghai-Shenzhen\", 433 enterprises; merge with company_operation_status.csv, filter valid samples with operating revenue > 0 and non-null R&D ratio, 430 enterprises.",
+    "Calculate sample 90th percentile threshold for operating revenue ≈ 5.653 billion yuan; leading group (43 enterprises) with revenue ≥ threshold, non-leading group (387 enterprises) for the rest. Calculate median R&D ratio by group: leading group median = 6.26%, non-leading group median = 13.06%.",
+    "Calculate Baidu's difference from each group median: 18.85% - 6.26% = 12.59 percentage points; 18.85% - 13.06% = 5.79 percentage points."
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "Baidu (BIDU) 2022 Annual Report PDF (20-F): Total revenues RMB 123,675 million, R&D expenses RMB 23,315 million.",
+    "From company_profile.csv: 433 A-share (Shanghai-Shenzhen) information transmission, software, and IT services listed companies.",
+    "From company_operation_status.csv: R&D ratio and operating revenue for 430 valid enterprises."
+  ],
+  "milestone": {
+    "Baidu total revenue (million RMB)": 123675,
+    "Baidu R&D expenses (million RMB)": 23315,
+    "Baidu R&D intensity (%)": 18.85,
+    "A-share IT industry profile count": 433,
+    "A-share IT industry valid samples": 430,
+    "Revenue 90th percentile threshold (billion yuan)": 56.53,
+    "Leading group enterprises count": 43,
+    "Non-leading group enterprises count": 387,
+    "Leading group R&D intensity median (%)": 6.26,
+    "Non-leading group R&D intensity median (%)": 13.06,
+    "Baidu above leading group median (percentage points)": 12.59,
+    "Baidu above non-leading group median (percentage points)": 5.79
+  },
+  "answer": [
+    18.85,
+    12.59,
+    5.79
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/international_comparison/medium025_result.json b/assets/qa_raw/international_comparison/medium025_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5c7224799e2ce5b0d1ed31d9285b247b063a5d7
--- /dev/null
+++ b/assets/qa_raw/international_comparison/medium025_result.json
@@ -0,0 +1,34 @@
+{
+  "id": "medium025",
+  "question": "Against the backdrop of intensive policy rollouts for the new energy vehicle industry, what was ZEEKR's asset turnover ratio (revenue divided by total assets) in 2022? How did it compare to the median asset turnover ratio of China's automobile manufacturing industry as a whole?",
+  "guidelines": "Answer in order: ZEEKR's asset turnover ratio (2 decimal places), and the difference from the industry median (2 decimal places; positive means ZEEKR is higher). Return as an array. If the relevant data cannot be found, answer \"Relevant data not found\".",
+  "steps": [
+    "From ZEEKR (ZK) F-1 prospectus PDF, extract 2022 financial data: Net revenues RMB 31,899,448 thousand, Total assets RMB 19,477,316 thousand. Calculate asset turnover ratio = 31,899,448 / 19,477,316 = 1.64.",
+    "From company_profile.csv, filter enterprises with industry == \"Automobile manufacturing\", 230 enterprises.",
+    "From company_operation_status.csv, obtain 'operating revenue' and 'total assets' for the above 230 automobile manufacturing enterprises; all 230 have valid data. Calculate asset turnover ratio for each = operating revenue / total assets; industry median = 0.59.",
+    "Calculate the difference between ZEEKR and industry median: 1.64 - 0.59 = 1.05; ZEEKR is significantly higher than the industry median, reflecting its asset-light operation and rapid scaling as a new EV brand."
+  ],
+  "steps_num": 4,
+  "evidence": [
+    "ZEEKR F-1 prospectus (SEC Filing) containing 2022 net revenues RMB 31,899,448 thousand, total assets RMB 19,477,316 thousand, etc.",
+    "From company_profile.csv: 230 automobile manufacturing enterprises.",
+    "From company_operation_status.csv: operating revenue and total assets for 230 automobile manufacturing enterprises."
+  ],
+  "milestone": {
+    "ZEEKR 2022 net revenue (RMB thousand)": 31899448,
+    "ZEEKR 2022 total assets (RMB thousand)": 19477316,
+    "ZEEKR asset turnover ratio": 1.64,
+    "Automobile manufacturing enterprises count": 230,
+    "Industry asset turnover ratio median": 0.59,
+    "ZEEKR difference from industry median": 1.05
+  },
+  "answer": [
+    1.64,
+    1.05
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "international_comparison"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/hard001_result.json b/assets/qa_raw/risk_assessment/hard001_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..24e9043b149545f46e2e40edb0a1be9090738c08
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/hard001_result.json
@@ -0,0 +1,39 @@
+{
+  "id": "hard001",
+  "question": "In 2022, in the automobile manufacturing industry, focusing on provinces with a total of 8 or more relevant enterprises, what proportion of the qualifying provinces simultaneously meet both the 'high policy dependence' risk (government subsidies as a proportion of operating profit >30%) and the 'high market concentration' risk (operating revenue CR4>60%)?",
+  "guidelines": "Answer format: percentage value (2 decimal places). E.g. 25.67. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "From regional_industry_status.csv, filter records with industry=\"Automobile Manufacturing\" and total enterprises>=8, extract province, total operating profit amount, total government reward and subsidy funds, and total operating revenue amount fields, finding 5 provinces.",
+    "Filter provinces with total operating profit amount>0 and all required fields non-null as provinces meeting the statistical criteria, totaling 5 provinces.",
+    "For each province meeting the statistical criteria, calculate government subsidies as a proportion of operating profit = (total government reward and subsidy funds / total operating profit amount) × 100%, and filter provinces with high policy dependence (>30%).",
+    "From company_profile.csv, filter enterprises with industry=\"Automobile Manufacturing\"; from company_operation_status.csv, obtain 2022 operating revenue amount, and group by province.",
+    "For each province meeting the statistical criteria, filter enterprises with non-null operating revenue amount, sort by operating revenue amount in descending order, calculate CR4 = (sum of top 4 enterprises' operating revenue / total operating revenue of all enterprises in that province) × 100%, and filter provinces with high market concentration (>60%).",
+    "Take the intersection of high policy dependence provinces and high market concentration provinces, and count the number of dual-risk provinces as 1.",
+    "Calculate the proportion = (number of dual-risk provinces / total number of provinces meeting statistical criteria) × 100% = (1 / 5) × 100% = 20.00%."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Found 5 automobile manufacturing provinces with total enterprises>=8 from regional_industry_status.csv.",
+    "After filtering, obtained 5 provinces meeting the statistical criteria (operating profit>0 and all required fields non-null)."
+  ],
+  "milestone": {
+    "Total number of provinces meeting statistical criteria": 5,
+    "High policy dependence provinces": [
+      "Shanghai"
+    ],
+    "High market concentration provinces": [
+      "Guangdong",
+      "Shanghai",
+      "Zhejiang",
+      "Shandong"
+    ],
+    "Number of dual-risk provinces": 1,
+    "Proportion (%)": 20.0
+  },
+  "answer": 20.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/hard002_result.json b/assets/qa_raw/risk_assessment/hard002_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..597cafd6cf6235397d67d4c59586cbd19313c3fc
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/hard002_result.json
@@ -0,0 +1,33 @@
+{
+  "id": "hard002",
+  "question": "In 2022, in the communication transmission equipment industry, assume that enterprises with R&D investment intensity below the national median for that industry are eliminated from the market within the next three years, and only enterprises with R&D investment intensity not below the national median remain. What is the ratio of remaining enterprises' operating revenue after elimination to operating revenue before elimination for the province with the highest such ratio?",
+  "guidelines": "Answer format: numeric value (2 decimal places, unit %). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "From national_industry_status.csv, filter industry=\"Communication Transmission Equipment\", extract the median R&D investment intensity as 9.92, used as the national benchmark.",
+    "From company_profile.csv, filter industry=\"Communication Transmission Equipment\" for all enterprise records, extract enterprise name, bmCode, and province fields; 120 enterprises found.",
+    "From company_operation_status.csv, join by bmCode to the enterprises above, extract operating revenue amount and R&D investment intensity; 120 enterprises matched.",
+    "Group by province and compute each province's total operating revenue before elimination (all enterprises, including those with missing R&D intensity); 19 provinces in total.",
+    "Filter enterprises with non-null R&D intensity not below the national median 9.92 (enterprises with null R&D intensity are treated as below the median and eliminated); group by province and compute total operating revenue of surviving enterprises after elimination. 13 provinces have at least one surviving enterprise.",
+    "For each valid province, compute revenue retention ratio = (total operating revenue after elimination / total operating revenue before elimination) × 100. For Anhui Province, revenue retention ratio = 2727186878.01 / 2727186878.01 × 100 = 100.00%.",
+    "Sort by revenue retention ratio in descending order; the province with the highest ratio is Anhui Province, at 100.00%."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "From national_industry_status.csv, the national median R&D investment intensity for communication transmission equipment is 9.92.",
+    "From company_profile.csv, 120 communication transmission equipment enterprises were found.",
+    "From company_operation_status.csv, operating revenue and R&D investment intensity were matched for 120 enterprises."
+  ],
+  "milestone": {
+    "National median R&D investment intensity": 9.92,
+    "Total enterprises in communication transmission equipment industry": 120,
+    "Anhui Province total operating revenue before elimination (CNY)": 2727186878.01,
+    "Anhui Province total operating revenue after elimination (CNY)": 2727186878.01,
+    "Anhui Province revenue retention ratio (%)": 100.0
+  },
+  "answer": 100.0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/hard003_result.json b/assets/qa_raw/risk_assessment/hard003_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..88e0ef53a4b69f2faa4f8ccb77f9cfa2c89a2ff3
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/hard003_result.json
@@ -0,0 +1,44 @@
+{
+  "id": "hard003",
+  "question": "2022年，国家层面发布了数字经济发展相关规划，同时部分省份也配套出台了支持通信传输设备业的政策，形成中央-地方协同支持格局。在同时受益于上述两级政策支持、且通信传输设备业上市企业数量不低于8家的省份中（计算政府补贴依赖度时，仅纳入政府补贴金额、营业利润金额、营业收入金额三项同时有完整记录的企业，依赖度=省内企业政府补贴总额÷省内企业营业利润总额），哪个省份对政府补贴的财务依赖最为突出？进一步模拟：一旦该省所有通信传输设备企业同时遭遇50%补贴缩减，且利润等额受损，全省通信传输设备业的营业利润将萎缩多大比例？",
+  "guidelines": "依次回答省份名称、政府补贴依赖度和营业利润下降比例。依赖度和下降比例均以百分数表示，保留2位小数。如[\"广东省\", 25.33, 12.67]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"通信传输设备业\"的政策记录，找到70条相关政策，其中国务院政策1条、部委政策23条、地方政策46条。",
+    "从policy_resource.csv中读取46条地方政策全文，筛选出44条明确涉及特定省份的地方政策（排除2条标注为\"全国\"的地方政策），覆盖17个省份：广东省11条（含广东省促进工业经济平稳增长措施、深圳市战略性新兴产业发展意见、广东省数字经济工作要点等）、安徽省4条、山东省4条、上海市3条、重庆市3条、四川省3条、云南省2条、贵州省2条、湖南省2条、陕西省2条、北京市2条（含北京市数字经济全产业链开放发展行动方案等）、河南省1条、海南省1条、湖北省1条（湖北省推进北斗产业高质量发展若干措施）、新疆维吾尔自治区1条、福建省1条、江西省1条。",
+    "从company_profile.csv筛选行业=\"通信传输设备业\"的企业共120家，按省份统计企业数。拥有不少于8家企业的省份有6个：广东省38家、江苏省18家、湖北省11家、北京市10家、浙江省9家、四川省8家。",
+    "取同时满足\"有地方通信产业政策支持\"和\"企业总数>=8\"的省份交集，得到4个符合条件的省份：广东省、北京市、湖北省、四川省。江苏省和浙江省虽有足够企业但无地方通信传输设备业相关政策。",
+    "从company_operation_status.csv获取这4个省份通信传输设备业企业的政府奖励资金补贴、营业利润金额和营业收入金额数据，筛选三项数据均完整的企业。广东省36家、北京市9家、湖北省11家、四川省8家数据完整。",
+    "计算各省份的政府补贴依赖度=省内企业政府补贴总额/省内企业营业利润总额×100%。北京市：补贴总额8.63亿元/营业利润总额44.22亿元=19.52%；广东省：76.22亿元/434.77亿元=17.53%；湖北省：7.84亿元/46.27亿元=16.94%；四川省：0.61亿元/15.24亿元=4.02%。北京市补贴依赖度最高，达19.52%。",
+    "模拟北京市政府补贴削减50%的冲击：补贴削减额=8.63亿元×50%=4.32亿元，营业利润等额减少4.32亿元。营业利润下降比例=4.32/44.22×100%=9.76%。"
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "从policy_release_status.csv中找到70条涉及通信传输设备业的政策记录，其中国务院政策1条、部委政策23条、地方政策46条。",
+    "从policy_resource.csv中分析国务院数字经济发展规划和23条部委政策全文，确认国家层面政策支持框架；分析44条有明确省份归属的地方政策全文，覆盖17个省份。",
+    "从company_profile.csv中筛选出120家通信传输设备业企业，6个省份拥有不少于8家企业。",
+    "从company_operation_status.csv中获取4个符合条件省份共64家数据完整企业的政府补贴和营业利润数据。"
+  ],
+  "milestone": {
+    "通信传输设备业相关政策总数（条）": 70,
+    "地方政策数（条）": 46,
+    "有地方政策的省份数（个）": 17,
+    "企业总数>=8的省份数（个）": 6,
+    "同时满足两项条件的省份数（个）": 4,
+    "北京市数据完整企业数（家）": 9,
+    "北京市政府补贴总额（亿元）": 8.63,
+    "北京市营业利润总额（亿元）": 44.22,
+    "北京市补贴依赖度（%）": 19.52,
+    "补贴削减额（亿元）": 4.32,
+    "营业利润下降比例（%）": 9.76
+  },
+  "answer": [
+    "北京市",
+    19.52,
+    9.76
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/hard004_result.json b/assets/qa_raw/risk_assessment/hard004_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..17308a46544b03d0abf5b7a090a8f9d3fc3e16e3
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/hard004_result.json
@@ -0,0 +1,33 @@
+{
+  "id": "hard004",
+  "question": "In 2022, suppose an industrial policy researcher is screening vulnerable provinces for the chemical raw materials and chemical products manufacturing industry: she defines \"government subsidy amount exceeding 5% of that province's industry total operating revenue\" as excessive subsidy dependence, and \"the single largest enterprise's operating revenue as a share of that province's industry total operating revenue above 40%\" as market structure imbalance. Only among provinces with total enterprises not less than 12 (operating profit margin = total operating profit / total operating revenue × 100%), among provinces that simultaneously trigger both alerts, what is the operating profit margin of the province with the weakest profitability?",
+  "guidelines": "Answer format: percentage value (2 decimal places). E.g. \"8.56\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "From regional_industry_status.csv, filter records with industry=\"Chemical Raw Materials and Chemical Products Manufacturing\" and total enterprises>=12, extract province, total government reward and subsidy funds, total operating revenue amount, maximum operating revenue amount, and total operating profit amount fields; 7 provinces found.",
+    "Filter provinces where all required fields are non-null and total operating revenue amount>0; 7 provinces in total.",
+    "For each province, compute government subsidy to operating revenue ratio = (total government reward and subsidy funds / total operating revenue amount) × 100%, and determine high government subsidy dependence (>5%); 0 provinces.",
+    "For each province, compute largest enterprise revenue share = (maximum operating revenue amount / total operating revenue amount) × 100%, and determine high revenue concentration (>40%); 1 province.",
+    "Filter provinces that simultaneously satisfy both high government subsidy dependence and high revenue concentration (dual-risk provinces); 0 provinces.",
+    "For each dual-risk province, compute operating profit margin = (total operating profit amount / total operating revenue amount) × 100%.",
+    "Sort by operating profit margin in ascending order; the province with the lowest operating profit margin was not found; operating profit margin is 0.00%."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Found 7 provinces with total enterprises>=12 for chemical raw materials and chemical products manufacturing from regional_industry_status.csv.",
+    "After filtering, obtained 7 provinces with complete required fields.",
+    "Found 0 dual-risk provinces."
+  ],
+  "milestone": {
+    "Number of provinces with complete required fields": 7,
+    "Number of high government subsidy dependence provinces": 0,
+    "Number of high revenue concentration provinces": 1,
+    "Number of dual-risk provinces": 0,
+    "Lowest operating profit margin (%)": 0
+  },
+  "answer": "No relevant data found",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/hard005_result.json b/assets/qa_raw/risk_assessment/hard005_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5c1ca920f734953eb5d52bd6e9cc7d28251e589f
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/hard005_result.json
@@ -0,0 +1,30 @@
+{
+  "id": "hard005",
+  "question": "At the end of 2022, semiconductor industry analysts modeled 2023 scenarios: revenue across enterprises is projected to decline by 20%, while fixed operating costs (defined as operating revenue minus operating profit) rise by 15% from their baseline. Under these assumptions, compute the change in operating profit margin for each enterprise in Jiangsu Province, Guangdong Province, and Shanghai (original operating profit margin = operating profit / operating revenue × 100%; new operating profit margin = (new operating revenue − new operating cost) / new operating revenue × 100%; decline magnitude = original operating profit margin − new operating profit margin), then take the average decline per region—limited to enterprises with complete operating revenue and operating profit data and positive operating revenue. Among the three regions, which province/municipality has the largest average decline in enterprise operating profit margin?",
+  "guidelines": "Answer format: province or city name. E.g. \"Zhejiang Province\" or \"Beijing\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "From company_profile.csv, filter enterprise records with industry=\"Semiconductor industry\" and province in (\"Jiangsu\", \"Guangdong\", \"Shanghai\"), extract enterprise name, bmCode, and province fields; 104 enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by enterprise name and bmCode, extract operating revenue amount and operating profit amount; 104 enterprises successfully matched.",
+    "Filter enterprise records with non-null operating revenue and operating profit and operating revenue amount>0; 104 enterprises in total.",
+    "For each enterprise, compute original operating profit margin = (operating profit amount / operating revenue amount) × 100%, original operating cost = operating revenue amount − operating profit amount.",
+    "For each enterprise, compute new operating revenue = operating revenue amount × 0.8, new operating cost = original operating cost × 1.15.",
+    "For each enterprise, compute new operating profit = new operating revenue − new operating cost, new operating profit margin = (new operating profit / new operating revenue) × 100%, decline magnitude = original operating profit margin − new operating profit margin.",
+    "Group enterprises by province; for each province compute average decline magnitude, sort by average decline magnitude in descending order; Guangdong Province has the largest average decline."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Found 104 semiconductor industry enterprises from company_profile.csv.",
+    "Found 104 enterprises with complete data from company_operation_status.csv."
+  ],
+  "milestone": {
+    "Number of enterprises with complete data": 104,
+    "Province/municipality with largest average decline": "Guangdong Province",
+    "Largest average decline (percentage points)": 42.12
+  },
+  "answer": "Guangdong",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/hard006_result.json b/assets/qa_raw/risk_assessment/hard006_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..af56c7a65f6e286ff87a65ae893420823b4e669b
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/hard006_result.json
@@ -0,0 +1,33 @@
+{
+  "id": "hard006",
+  "question": "In 2022, in the comprehensive assessment of the semiconductor industry ecosystem, the scope is limited to provinces with total enterprises not less than 6. The assessment framework has three dimensions: ? \"Industry Scale Foundation\" (weight 30%): half the sum of each province's share of national semiconductor enterprise count and each province's share of national semiconductor operating revenue as the raw score for this dimension; ? \"Policy Ecosystem Density\" (weight 30%): each province's count of policies involving the Chinese terms for semiconductors, chips, or integrated circuits divided by the national total of such policies; ? \"Technological Accumulation Depth\" (weight 40%): each province's share of cumulative Chinese invention patent grants in the national total for the industry. Each dimension is min-max normalized (mapped to 0-100), then weighted to yield the comprehensive development potential index. Which province has the highest comprehensive score?",
+  "guidelines": "Answer format: province name. E.g. \"Guangdong Province\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "From national_industry_status.csv, filter records with industry=\"semiconductor industry\", and extract national benchmarks: total enterprises 172, total operating revenue amount 1494377201673.42, and total cumulative Chinese invention patent grants 54069.",
+    "From policy_release_status.csv, filter policy records whose policy name or full text involves the Chinese terms for semiconductors, chips, or integrated circuits, extract the province field, and count the number of semiconductor-related policies per province as well as the national total of such policies.",
+    "From regional_industry_status.csv, filter records with industry=\"semiconductor industry\" and total enterprises >= 6, extract the province, total enterprises, total operating revenue amount, and total cumulative Chinese invention patent grants fields; 6 provinces found.",
+    "Filter provinces where all required fields are non-null; 5 provinces in total.",
+    "For each province, compute raw industry scale foundation = (total enterprises / national total enterprises + total operating revenue amount / national total operating revenue amount) / 2.",
+    "For each province, compute raw policy ecosystem = that province's semiconductor-related policy count / national total semiconductor-related policies, and raw technological accumulation = total cumulative Chinese invention patent grants / national total cumulative Chinese invention patent grants.",
+    "Perform min-max normalization to 0–100 for industry scale foundation, policy ecosystem, and technological accumulation; for each province compute industry ecosystem resilience index = industry scale foundation score × 0.3 + policy ecosystem score × 0.3 + technological accumulation score × 0.4; sort by index in descending order; the highest-scoring province is Shanghai."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "From national_industry_status.csv, obtained national benchmark data for the semiconductor industry, including total enterprises 172, national total operating revenue, and total patent grants.",
+    "From policy_release_status.csv, counted 4 semiconductor-related policies nationally.",
+    "From regional_industry_status.csv, found 5 qualifying provinces."
+  ],
+  "milestone": {
+    "National total enterprises": 172,
+    "National total semiconductor-related policies": 4,
+    "Number of qualifying provinces": 5,
+    "Highest-scoring province": "Shanghai",
+    "Highest score": 87.38
+  },
+  "answer": "Shanghai",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/hard007_result.json b/assets/qa_raw/risk_assessment/hard007_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..5d4443115ebec552ebda2edfdc6c93d06e30b34b
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/hard007_result.json
@@ -0,0 +1,30 @@
+{
+  "id": "hard007",
+  "question": "In 2022, Guangdong Province's automobile manufacturing industry faces pressure from subsidy phase-out. A three-year phase-out path is set: compared with the actual subsidy amount in 2022, cut 20% in 2023, 40% in 2024, and 60% in 2025. For each Guangdong automobile manufacturing enterprise that has a recorded government reward/subsidy amount, its net profit loss in each year is exactly equal to that year's subsidy reduction; summing the losses for 2023–2025 gives the enterprise's cumulative net profit loss. Under this definition, how many hundred million yuan is the three-year total loss for the enterprise with the heaviest cumulative damage?",
+  "guidelines": "Answer format: a numeric value (2 decimal places). For example, \"5.67\" means a cumulative loss of 5.67 hundred million yuan. If relevant data cannot be found, answer \"No relevant data found\".",
+  "steps": [
+    "From company_profile.csv, filter records with province = \"Guangdong Province\" and industry = \"Automobile Manufacturing\", extract enterprise name and bmCode fields, 27 enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by enterprise name and bmCode, extract the government reward/subsidy field, 27 enterprises matched.",
+    "Filter to records where government reward/subsidy is not null, 26 enterprises.",
+    "For each enterprise, compute 2023 subsidy reduction = government reward/subsidy × 0.2.",
+    "For each enterprise, compute 2024 subsidy reduction = government reward/subsidy × 0.4.",
+    "For each enterprise, compute 2025 subsidy reduction = government reward/subsidy × 0.6.",
+    "For each enterprise, compute cumulative net profit loss = 2023 subsidy reduction + 2024 subsidy reduction + 2025 subsidy reduction; sort by cumulative loss descending. The enterprise with the largest cumulative loss is Bei Qi Lu Yuan Xin Neng Yuan Qi Che Co., Ltd.; convert the amount from yuan to hundred million yuan, cumulative loss 20.53 hundred million yuan."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Found 27 Guangdong Province automobile manufacturing enterprises in company_profile.csv.",
+    "Found 26 enterprises with complete government subsidy data in company_operation_status.csv."
+  ],
+  "milestone": {
+    "Enterprises with complete government subsidy data": 26,
+    "Maximum cumulative loss (yuan)": 2052595485.6,
+    "Maximum cumulative loss (hundred million yuan)": 20.53
+  },
+  "answer": "20.53",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/hard008_result.json b/assets/qa_raw/risk_assessment/hard008_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..6efcc8a15d8d7270f55af27d52e57bd2e308adba
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/hard008_result.json
@@ -0,0 +1,29 @@
+{
+  "id": "hard008",
+  "question": "In 2022, suppose the chemical raw materials and chemical products manufacturing industry faces dual pressure in 2023: environmental protection investment must increase by an amount equivalent to 8% of operating revenue, and raw material price increases lead to a 10% increase in existing costs. If existing costs account for 75% of operating revenue, among enterprises in this industry in Jiangsu Province, how many enterprises will have negative new operating profit?",
+  "guidelines": "Answer format: integer. E.g. 15. If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "From company_profile.csv, filter enterprise records with province=\"Jiangsu Province\" and industry=\"Chemical Raw Materials and Chemical Products Manufacturing\", extract enterprise name and bmCode fields; 55 enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by enterprise name and bmCode, extract operating revenue amount and operating profit amount; 55 enterprises successfully matched.",
+    "Filter enterprise records with non-null operating revenue and operating profit; 55 enterprises in total.",
+    "For each enterprise, compute original cost = operating revenue amount × 75%.",
+    "For each enterprise, compute new cost = original cost × 1.1.",
+    "For each enterprise, compute environmental protection investment increase = operating revenue amount × 8%.",
+    "For each enterprise, compute new operating profit = operating revenue amount − new cost − environmental protection investment increase; count enterprises with new operating profit < 0: 0 enterprises."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Found 55 chemical raw materials and chemical products manufacturing enterprises in Jiangsu Province from company_profile.csv.",
+    "Found 55 enterprises with complete data from company_operation_status.csv."
+  ],
+  "milestone": {
+    "Number of enterprises with complete data": 55,
+    "Number of enterprises with negative new operating profit": 0
+  },
+  "answer": 0,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/hard009_result.json b/assets/qa_raw/risk_assessment/hard009_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3b8cae638bf6e4112289c96b0737b79038f432e
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/hard009_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "hard009",
+  "question": "In 2022, build a comprehensive industry-level risk resilience scoring system covering all manufacturing industries with total enterprises not less than 25 (excluding Finance, Real Estate, and Conglomerate industries). The scoring framework has three dimensions with different weights: Asset Return Efficiency (weight 40%), measured as the industry's average operating profit per enterprise divided by average total assets per enterprise; R&D Investment Intensity (weight 30%), measured as the industry's average R&D investment per enterprise divided by average operating revenue per enterprise; and Financial Desensitization to Policy Subsidies (weight 30%), measured as the industry's total operating profit divided by (total government subsidies + 1)—adding 1 to the denominator avoids division by zero when subsidies are zero. Each of the three raw indicators is min-max normalized across industries, scaled to a 0–100 score, then weighted by the above weights to yield the final score. Under this system, what is the comprehensive score of the top-ranked industry?",
+  "guidelines": "Answer format: numeric value (2 decimal places). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "From national_industry_status.csv, extract total enterprises for all industries; filter manufacturing industries with total enterprises>=25, excluding non-manufacturing industries such as \"Finance\", \"Real Estate\", and \"Conglomerate\"; 38 qualifying industries found.",
+    "For each qualifying industry, from national_industry_status.csv extract mean operating profit amount, mean total assets, mean R&D investment amount, mean operating revenue amount, total operating profit amount, and total government reward and subsidy funds fields.",
+    "Filter industry records where all required fields are non-null; 38 industries in total.",
+    "For each industry, compute raw financial robustness = mean operating profit amount / mean total assets, raw innovation capacity = mean R&D investment amount / mean operating revenue amount, raw policy independence = total operating profit amount / (total government reward and subsidy funds + 1).",
+    "Perform min-max normalization to 0–100 for the three dimensions (financial robustness, innovation capacity, policy independence); normalization formula = (value − min) / (max − min) × 100.",
+    "For each industry, compute comprehensive risk resilience index = financial robustness score × 0.4 + innovation capacity score × 0.3 + policy independence score × 0.3.",
+    "Sort by comprehensive risk resilience index in descending order; the top-ranked industry is Mining, with a score of 65.11."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Found 38 manufacturing industries with total enterprises>=25 from national_industry_status.csv.",
+    "After filtering, obtained 38 industries with complete required fields."
+  ],
+  "milestone": {
+    "Number of qualifying industries": 38,
+    "Number of industries with complete required fields": 38,
+    "Top-ranked industry": "Mining",
+    "Highest score": 65.11
+  },
+  "answer": 65.11,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/hard010_result.json b/assets/qa_raw/risk_assessment/hard010_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..b9955eee123751ec89f7903dca0f4a18b23fecbe
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/hard010_result.json
@@ -0,0 +1,45 @@
+{
+  "id": "hard010",
+  "question": "In 2022, to measure the resilience reserve of each province's pharmaceutical manufacturing industry under extreme external shocks, a research team established the following three-dimensional evaluation framework (including only provinces with total enterprises not less than 10): Dimension 1 \"Policy Shield Thickness\" (weight 35%): for each province, the number of local policies involving the Chinese terms for pharmaceuticals or bio-industry divided by the national total of such policies yields the raw relative policy density; Dimension 2 \"Independent Profitability Buffer\" (weight 35%): each province's total pharmaceutical manufacturing operating profit minus total government subsidies, divided by total assets, reflecting actual asset profitability excluding subsidies; Dimension 3 \"Technological Innovation Reserve Thickness\" (weight 30%): patent grants per unit of R&D investment by province, measuring R&D output efficiency. Each dimension is min-max normalized across provinces (mapped to 0-100), then weighted by the above weights to produce the comprehensive score. Among the top three ranked provinces, what is the average comprehensive score?",
+  "guidelines": "Answer format: numeric value (2 decimal places). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "From policy_release_status.csv, filter policy records whose policy name or full text involves the Chinese terms for pharmaceuticals or bio-industry, extract the province field, and count the number of pharmaceutical-related policies per province as well as the national total of such policies.",
+    "From regional_industry_status.csv, filter records with industry=\"Pharmaceutical Manufacturing\" and total enterprises>=10, extract province, total operating profit amount, total government reward and subsidy funds, total assets, total cumulative Chinese invention patent grants, and total R&D investment amount fields; 11 provinces found.",
+    "Filter provinces where all required fields are non-null, total assets>0, and total R&D investment amount>0; 10 provinces in total.",
+    "For each province, compute raw policy support intensity = that province's pharmaceutical-related policy count / national total of pharmaceutical-related policies.",
+    "For each province, compute raw financial buffer capacity = (total operating profit amount − total government reward and subsidy funds) / total assets, and raw innovation reserve = total cumulative Chinese invention patent grants / total R&D investment amount.",
+    "Perform min-max normalization to 0–100 for each of the three dimensions: policy support intensity, financial buffer capacity, and innovation reserve.",
+    "Sort by crisis response capability index in descending order; the top 3 provinces and their scores are: Guangdong Province (70.12), Shanghai (68.56), Beijing (69.23).",
+    "Compute average score = (70.12 + 68.56 + 69.23) / 3 = 69.40."
+  ],
+  "steps_num": 8,
+  "evidence": [
+    "From policy_release_status.csv, national total of pharmaceutical-related policies is 21.",
+    "From regional_industry_status.csv, found 10 qualifying provinces."
+  ],
+  "milestone": {
+    "National total of pharmaceutical-related policies": 21,
+    "Number of qualifying provinces": 10,
+    "Top 3 provinces and scores": [
+      {
+        "Province": "Guangdong Province",
+        "Score": 70.12
+      },
+      {
+        "Province": "Shanghai",
+        "Score": 68.56
+      },
+      {
+        "Province": "Beijing",
+        "Score": 69.23
+      }
+    ],
+    "Average score": 69.4
+  },
+  "answer": 69.4,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/hard011_result.json b/assets/qa_raw/risk_assessment/hard011_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3fd4eee94bad7999e51432532352396af89962e0
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/hard011_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "hard011",
+  "question": "2022年，国家从消费品工业数字化转型、质量标准提升等多个维度出台了政策支持，部分省份也因势利导推出了地方消费电子产业相关政策方案。现聚焦受上述双重政策覆盖的省份（同时须满足：省内消费电子及电气业上市企业不少于5家，且全省总营业利润大于零）。在这批省份中，哪个省份的企业负债结构对利率最为敏感——即以全省总负债相对于总营业利润的倍率（总负债/总营业利润）来衡量，哪个省份的这一比值最高？在此基础上，若利率水平抬升2个百分点，以各企业总负债×2%估算其新增利息负担，全省新增利息成本总额将相当于该省消费电子及电气业总营业利润的百分之多少？",
+  "guidelines": "依次回答省份名称和额外利息成本占总营业利润的比例。比例以百分数表示，保留2位小数。如[\"广东省\", 23.36]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"消费电子\"的政策记录，找到10条消费电子及电气业相关政策，其中4条为部委政策（国家层面），6条为地方政策。",
+    "经过对地方政策内容的分析，识别出台了地方消费电子产业发展政策的省份共5个：广东省（促进工业经济平稳增长措施，提出促进消费电子产品促销、推动8500家企业技术改造等）、江西省（打造全国新兴产业培育发展高地方案，提出电子信息产业规模突破12000亿元目标、新兴产业研发投入强度达2%）、四川省（承接制造业有序转移实施意见和成都制造业规划，提出承接电子信息配套产业、打造世界级电子信息产业集群）、重庆市（促进大中小企业融通发展方案）、海南省（激励企业上规模奖励资金细则）。",
+    "从company_profile.csv筛选行业=\"消费电子及电气业\"的企业，按省份统计，在上述省份中筛选企业总数>=5且总营业利润>0的省份：广东省150家、江西省5家、四川省8家符合条件。重庆市仅1家、海南省0家，不满足条件。",
+    "从company_operation_status.csv提取这3个省份消费电子及电气业企业的总负债和营业利润数据，计算各省份负债/营业利润倍数：江西省总负债166.30亿元/总营业利润5.51亿元=30.18倍、四川省总负债740.05亿元/总营业利润36.27亿元=20.40倍、广东省总负债16211.57亿元/总营业利润1387.98亿元=11.68倍。",
+    "江西省负债/营业利润倍数最高（30.18倍），利率敏感性风险最大。",
+    "计算江西省利率上升2个百分点的影响：额外利息成本=总负债166.30亿元×2%=3.33亿元，占总营业利润5.51亿元的比例=3.33/5.51×100%=60.36%。"
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "从policy_release_status.csv中找到10条消费电子及电气业相关政策，其中4条国家层面政策、6条地方政策。",
+    "从policy_resource.csv中分析10条政策全文，识别出4条国家层面消费品工业发展政策和5个有地方消费电子政策的省份。",
+    "从company_profile.csv中找到3个符合条件省份（企业>=5、利润>0）共163家消费电子及电气业企业。",
+    "从company_operation_status.csv中获取这163家企业的总负债和营业利润数据。"
+  ],
+  "milestone": {
+    "消费电子相关政策总数（条）": 10,
+    "国家层面政策数（条）": 4,
+    "地方政策涉及省份数（个）": 5,
+    "符合条件省份数（企业>=5且利润>0）": 3,
+    "江西省消费电子企业数（家）": 5,
+    "江西省总负债（亿元）": 166.3,
+    "江西省总营业利润（亿元）": 5.51,
+    "江西省负债/营业利润倍数": 30.18,
+    "江西省额外利息成本（亿元）": 3.33,
+    "江西省额外利息成本占总营业利润比例（%）": 60.36
+  },
+  "answer": [
+    "江西省",
+    60.36
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/hard012_result.json b/assets/qa_raw/risk_assessment/hard012_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..83c530d1195f224a1655feda78b01537307c7aaf
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/hard012_result.json
@@ -0,0 +1,31 @@
+{
+  "id": "hard012",
+  "question": "In 2022, in the chemical raw materials and chemical products manufacturing industry, comprehensively assess each province's ability to withstand external shocks. What is the comprehensive score of the lowest-scoring province? (Assessment indicators: Profit Diversification 30%, Financial Safety Cushion 35%, Policy Buffer 35%. Profit Diversification = 1 − (largest enterprise operating profit / total operating profit) normalized score; Financial Safety Cushion = (total operating profit / total liabilities) normalized score; Policy Buffer = (total operating profit / (total government subsidies + 1)) normalized score; each dimension min-max normalized to 0–100; only provinces with total enterprises>=15 are included)",
+  "guidelines": "Answer format: numeric value (2 decimal places). If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "From regional_industry_status.csv, filter records with industry=\"Chemical Raw Materials and Chemical Products Manufacturing\" and total enterprises>=15, extract province, maximum operating profit amount, total operating profit amount, total liabilities, and total government reward and subsidy funds fields; 6 provinces found.",
+    "Filter provinces where all required fields are non-null, total operating profit amount>0, and total liabilities>0; 6 provinces in total.",
+    "For each province, compute raw profit diversification = 1 − (maximum operating profit amount / total operating profit amount).",
+    "For each province, compute raw financial safety cushion = total operating profit amount / total liabilities.",
+    "For each province, compute raw policy buffer = total operating profit amount / (total government reward and subsidy funds + 1).",
+    "Perform min-max normalization to 0–100 for profit diversification, financial safety cushion, and policy buffer.",
+    "For each province, compute regional shock resilience index = profit diversification score × 0.3 + financial safety cushion score × 0.35 + policy buffer score × 0.35; sort by index in ascending order; the lowest-scoring province is Shanghai, with a score of 15.78."
+  ],
+  "steps_num": 7,
+  "evidence": [
+    "Found 6 chemical raw materials and chemical products manufacturing provinces with total enterprises>=15 from regional_industry_status.csv.",
+    "After filtering, obtained 6 provinces with complete required fields."
+  ],
+  "milestone": {
+    "Number of provinces with total enterprises>=15": 6,
+    "Number of provinces with complete required fields": 6,
+    "Lowest-scoring province": "Shanghai",
+    "Lowest score": 15.78
+  },
+  "answer": 15.78,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/hard013_result.json b/assets/qa_raw/risk_assessment/hard013_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..f88f48670774a4af64a97740a3b84552200fb18a
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/hard013_result.json
@@ -0,0 +1,35 @@
+{
+  "id": "hard013",
+  "question": "In 2022 regional data for the general purpose equipment manufacturing industry, for the province with the highest total government reward and subsidy funds: comprehensively rate its industrial competitiveness along two strategic dimensions. \"Industrial Upgrading Capacity\" combines the province's average rank across three sub-indicators (R&D investment growth rate, year-on-year change in Chinese patent applications, R&D personnel share); \"Industrial Base\" combines the province's average rank across three sub-indicators (total enterprises, total operating revenue, subsidy efficiency = total operating revenue / total government subsidies). The overall comprehensive performance rank is the average of the two dimension ranks. Answer only: Can this province's comprehensive industrial performance rank among the top 5 nationally?",
+  "guidelines": "Answer format: \"Yes\" or \"No\". If relevant data cannot be found, please answer \"No relevant data found\"",
+  "steps": [
+    "From regional_industry_status.csv, filter industry=\"General Purpose Equipment Manufacturing\", extract province and total government reward and subsidy funds fields; after filtering nulls, 16 provinces.",
+    "Sort by total subsidies in descending order; the province with the highest total government subsidies is Shanghai (2466523519.29 yuan).",
+    "From the same table, extract per-province mean year-on-year R&D investment change, mean R&D personnel share, total enterprises, total operating revenue amount, and total annual Chinese patent applications fields.",
+    "Compute per-province subsidy efficiency = total operating revenue amount / total government reward and subsidy funds (filter out zero denominator); 14 valid provinces.",
+    "Rank all provinces in descending order for each of the six indicators (R&D growth rate, annual patent applications, R&D personnel share, enterprise scale, revenue scale, subsidy efficiency); compute each province's rank on the six indicators.",
+    "Compute industrial upgrading capacity rank = (R&D growth rank + patent application rank + talent share rank) / 3; Shanghai's industrial upgrading capacity rank = 8.67.",
+    "Compute industrial base rank = (enterprise scale rank + revenue scale rank + subsidy efficiency rank) / 3; Shanghai's industrial base rank = 4.33.",
+    "Comprehensive performance rank = (industrial upgrading rank + industrial base rank) / 2; Shanghai's comprehensive performance rank value = 6.50.",
+    "Sort all provinces by comprehensive performance rank in ascending order; Shanghai's comprehensive rank is 6th, > 5; output \"No\"."
+  ],
+  "steps_num": 9,
+  "evidence": [
+    "Found government subsidy data for 16 provinces in general purpose equipment manufacturing from regional_industry_status.csv.",
+    "Found complete statistics for 14 provinces in general purpose equipment manufacturing from regional_industry_status.csv (subsidy efficiency computable)."
+  ],
+  "milestone": {
+    "Province with highest government subsidies": "Shanghai",
+    "Shanghai total government subsidies (yuan)": 2466523519.29,
+    "Shanghai industrial upgrading capacity rank": 8.67,
+    "Shanghai industrial base rank": 4.33,
+    "Shanghai comprehensive performance rank value": 6.5,
+    "Shanghai comprehensive rank": 6
+  },
+  "answer": "No",
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "hard",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/medium001_result.json b/assets/qa_raw/risk_assessment/medium001_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f374b73f3b34ae6d12ac0eed77e5c7a016ec4e0
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/medium001_result.json
@@ -0,0 +1,36 @@
+{
+  "id": "medium001",
+  "question": "Based on 2022 data, consider the following policy effect scenario: For provinces that have promulgated local industrial policies whose titles contain the keywords \"semiconductor\" or \"integrated circuit\", policy support will accelerate their semiconductor industry enterprises' R&D expansion pace to 2× the current growth rate over the next 3 years; for provinces that have not yet issued such policies, R&D growth remains unchanged. Using the median year-on-year change in enterprise R&D investment by province as the baseline growth rate, projected with 3-year compound growth, which province will rank first nationwide in total semiconductor industry R&D investment by 2025? What is the projected amount?",
+  "guidelines": "Answer format: [Province name, value (2 decimal places, unit: yuan)]. If relevant data cannot be found, answer \"No relevant data found\".",
+  "steps": [
+    "From policy_resource.csv, filter policies whose titles contain \"semiconductor\" or \"integrated circuit\", 4 records. Extract the involved province list (deduplicated, excluding national policies), yielding 4 provinces: Guangdong, Zhejiang, Shanghai, Anhui.",
+    "From regional_industry_status.csv, filter records with industry = \"semiconductor industry\" and exclude provinces where either R&D investment amount or year-on-year R&D change is missing, 16 provinces: Guangdong Province, Beijing Municipality, Jiangsu Province, Shanghai Municipality, Zhejiang Province, Shandong Province, Sichuan Province, Anhui Province, Hong Kong SAR, Hunan Province, Hebei Province, Liaoning Province, Jilin Province, Xinjiang Uygur Autonomous Region, Shanxi Province.",
+    "Group by province, calculate total R&D investment and median year-on-year R&D change for each province, 16 provinces.",
+    "For each province, determine if it is a policy province: policy provinces use adjusted growth rate = median growth rate × 2; non-policy provinces use adjusted growth rate = median growth rate. The top-ranked province, Shanghai, is a policy province with adjusted growth rate 64.38%.",
+    "Calculate each province's projected 2025 total R&D investment = total R&D investment × (1 + adjusted growth rate/100)^3. Shanghai's projected 2025 R&D investment = 22003461800.09 × (1+64.38/100)^3 = 97732260069.03.",
+    "The top-ranked province is Shanghai, with projected 2025 total R&D investment of 97732260069.03 yuan."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "From policy_release_status.csv: 4 policy records containing \"semiconductor\" or \"integrated circuit\" keywords, involving 4 provinces.",
+    "From regional_industry_status.csv: 16 provinces with valid data.",
+    "From regional_industry_status.csv: R&D investment amounts and year-on-year R&D change data for 168 semiconductor industry enterprises."
+  ],
+  "milestone": {
+    "Provinces with policy count": 4,
+    "Provinces with valid data count": 16,
+    "Shanghai 2022 total R&D investment (yuan)": 22003461800.09,
+    "Shanghai 2022 median YoY R&D change (%)": 32.19,
+    "Shanghai adjusted growth rate (%)": 64.38,
+    "Shanghai projected 2025 R&D investment (yuan)": 97732260069.03
+  },
+  "answer": [
+    "Shanghai",
+    97732260069.03
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/medium002_result.json b/assets/qa_raw/risk_assessment/medium002_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..237af39be634560a3ac9e6cebf269bb68f151056
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/medium002_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "medium002",
+  "question": "In 2022, focusing on pharmaceutical manufacturing: among provinces included in the statistics (requiring total related enterprises in the province ≥ 10), if a province has R&D investment concentration CR3 greater than 60% and cumulative granted Chinese invention patent concentration CR3 also greater than 60%, classify that province as a high-risk \"R&D–patent dual head concentration\" province. How many provinces satisfy both dual-concentration conditions?",
+  "guidelines": "Answer format: an integer, e.g. \"3\". If relevant data cannot be found, answer \"No relevant data found\".",
+  "steps": [
+    "From company_profile.csv, group enterprise records by province and count enterprises per province; filter to provinces with total enterprises ≥ 10, obtaining 11 provinces: Beijing Municipality, Shanghai Municipality, Jiangsu Province, Guangdong Province, Zhejiang Province, Shandong Province, Sichuan Province, Hubei Province, Hunan Province, Jilin Province, Hong Kong SAR.",
+    "From company_profile.csv, filter records in the above provinces with industry = \"Pharmaceutical Manufacturing\", extract enterprise name, bmCode, and province fields, 340 enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by enterprise name and bmCode, extract R&D investment amount and cumulative granted Chinese invention patents.",
+    "For each qualifying province, filter enterprises with non-null R&D investment amount, sort by R&D investment amount descending, take the top 3 enterprises, compute R&D concentration CR3 = (sum of top 3 enterprises' R&D / sum of all enterprises' R&D in the province) × 100%, and select provinces with R&D CR3 > 60%.",
+    "For each qualifying province, filter enterprises with non-null cumulative granted Chinese invention patents, sort by cumulative patents descending, take the top 3 enterprises, compute patent output concentration CR3 = (sum of top 3 enterprises' patents / sum of all enterprises' patents in the province) × 100%, and select provinces with patent CR3 > 60%.",
+    "Take the intersection of provinces with R&D CR3 > 60% and provinces with patent CR3 > 60%; count provinces satisfying both dual-concentration conditions: 3."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "From company_profile.csv, grouped by province and filtered to provinces with total enterprises ≥ 10, 11 provinces obtained.",
+    "From company_profile.csv, found 340 pharmaceutical manufacturing enterprises in the above 11 provinces.",
+    "From company_operation_status.csv, obtained 2022 R&D investment amounts and cumulative granted Chinese invention patent counts for 340 enterprises."
+  ],
+  "milestone": {
+    "Provinces with total enterprises ≥ 10": 11,
+    "Provinces with R&D CR3 > 60%": [
+      "Hong Kong SAR",
+      "Jilin Province",
+      "Sichuan Province",
+      "Hubei Province"
+    ],
+    "Provinces with patent output CR3 > 60%": [
+      "Hong Kong SAR",
+      "Shanghai Municipality",
+      "Hubei Province",
+      "Sichuan Province"
+    ],
+    "Count of dual high-concentration risk provinces": 3
+  },
+  "answer": 3,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/medium003_result.json b/assets/qa_raw/risk_assessment/medium003_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..e37258ff9166f5194bc65240e8d52058e8f20e93
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/medium003_result.json
@@ -0,0 +1,32 @@
+{
+  "id": "medium003",
+  "question": "In 2022, assume all national R&D tax incentives are cancelled. Among manufacturing industries (only industries with total enterprises ≥ 20 and complete R&D investment and operating revenue data, positive revenue, excluding \"Financial Services\", \"Real Estate\", and \"Conglomerates\"), assume a corporate income tax rate of 25%; R&D tax benefits include 100% additional deduction. What is the sum of the declines (in percentage points) for the three industries with the largest average net profit margin decline, where net profit margin impact = R&D × 100% × 25% / operating revenue × 100%?",
+  "guidelines": "Answer format: a numeric value (2 decimal places). For example, 15.67 means the sum of the three industries' declines is 15.67 percentage points. If relevant data cannot be found, answer \"No relevant data found\".",
+  "steps": [
+    "From national_industry_status.csv, extract total enterprises for all industries; filter to manufacturing industries with total enterprises ≥ 20; exclude non-manufacturing industries such as \"Financial Services\", \"Real Estate\", and \"Conglomerates\", obtaining 41 qualifying industries. Identify 30 valid provinces: Guangdong Province, Zhejiang Province, Jiangsu Province, Beijing Municipality, Shanghai Municipality, Shandong Province, Hong Kong SAR, Fujian Province, Sichuan Province, Anhui Province, Hubei Province, Hunan Province, Henan Province, Liaoning Province, Hebei Province, Shaanxi Province, Tianjin Municipality, Chongqing Municipality, Xinjiang Uygur Autonomous Region, Jilin Province, Shanxi Province, Guangxi Zhuang Autonomous Region, Yunnan Province, Heilongjiang Province, Gansu Province, Guizhou Province, Inner Mongolia Autonomous Region, Hainan Province, Tibet Autonomous Region.",
+    "From company_profile.csv, filter all enterprises in the qualifying industries, extract enterprise name, bmCode, and industry fields; related enterprises found.",
+    "From company_operation_status.csv, filter 2022 data for these enterprises by enterprise name and bmCode; keep records where R&D investment amount and operating revenue amount are both non-null and operating revenue amount > 0, 5416 enterprises.",
+    "For each enterprise, compute net profit margin impact = (R&D investment amount × 100% × 25%) / operating revenue amount × 100%.",
+    "For each industry, compute average enterprise net profit margin decline = Σ(enterprise net profit margin impact) / number of enterprises in that industry; average decline computed for 41 industries.",
+    "Sort by average net profit margin decline descending; take the top 3 industries (Pharmaceutical Manufacturing, Specialized Equipment Manufacturing, Information Transmission, Software and IT Services); sum their declines = 601.08 percentage points."
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "From national_industry_status.csv: 41 manufacturing industries with total enterprises ≥ 20, 30 valid provinces.",
+    "From company_profile.csv and company_operation_status.csv: 5416 enterprises with complete data.",
+    "Average decline computed for 41 industries."
+  ],
+  "milestone": {
+    "Qualifying industries count": 41,
+    "Qualifying provinces count": 30,
+    "Enterprises with complete data": 5416,
+    "Industries with computed average decline": 41,
+    "Sum of top-3 industry declines (percentage points)": 601.08
+  },
+  "answer": 601.08,
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/medium004_result.json b/assets/qa_raw/risk_assessment/medium004_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..c59a0da66d5a06acfd22059e71dd52331158ea59
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/medium004_result.json
@@ -0,0 +1,40 @@
+{
+  "id": "medium004",
+  "question": "2022年，《建材行业碳达峰实施方案》由多部委联合印发，各省也相继出台了碳达峰或节能减排地方行动方案，建材类企业由此面临双层合规压力。在同时处于上述两级政策约束之下的省份中，进一步筛选非金属矿物制品业样本：省内上市企业总数至少5家，且整体营业利润为正。对于符合条件的省份，若将碳排放合规成本设定为各企业运营成本（运营成本=营业收入-营业利润）的5%，并全部计入利润扣减项，那么，哪个省份的非金属矿物制品业总营业利润下降幅度最大？该省合规成本冲击后，总营业利润究竟会下降多少个百分点？",
+  "guidelines": "依次回答省份名称和营业利润下降百分比。百分比保留2位小数。如[\"湖南省\", 28.63]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"非金属矿物\"的政策记录，找到39条非金属矿物制品业相关政策。",
+    "在39条政策中，筛选policyClassification为\"地方政策\"且政策名称包含\"碳达峰\"或\"节能减排\"的地方政策，找到8条，涉及7个省份：四川省、宁夏回族自治区、安徽省、江西省、湖南省、贵州省、辽宁省。",
+    "从company_profile.csv筛选行业=\"非金属矿物制品业\"的企业，按省份统计，在上述7个省份中筛选有营业收入的企业总数>=5且总营业利润>0的省份，得到3个符合条件的省份：四川省（5家）、安徽省（5家）、湖南省（5家）。",
+    "从company_operation_status.csv提取这3个省份非金属矿物制品业企业的营业收入和营业利润数据，计算各省的运营成本（=营业收入-营业利润）和碳排放合规成本（=运营成本×5%），再计算营业利润下降百分比（=碳排放合规成本/总营业利润×100%）：四川省总营业利润11.78亿元，总运营成本19.30亿元，碳排放合规成本0.96亿元，下降8.19%；安徽省总营业利润263.50亿元，总运营成本2410.28亿元，碳排放合规成本120.51亿元，下降45.74%；湖南省总营业利润23.53亿元，总运营成本149.90亿元，碳排放合规成本7.49亿元，下降31.85%。",
+    "安徽省营业利润下降幅度45.74%最大，原因是安徽省非金属矿物制品业企业整体利润率较低（约9.86%），运营成本规模远大于利润，因此碳成本冲击的相对影响最严重。"
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "从policy_release_status.csv中找到39条非金属矿物制品业相关政策。",
+    "从policy_resource.csv中分析9条政策全文（1条国家碳达峰政策+8条地方碳达峰/节能减排政策），提取碳达峰要求和建材行业具体措施，涉及7个省份。",
+    "从company_profile.csv中筛选出3个符合条件省份的15家非金属矿物制品业企业。",
+    "从company_operation_status.csv中获取这15家企业的营业收入和营业利润数据，用于计算运营成本和碳排放合规成本。"
+  ],
+  "milestone": {
+    "非金属矿物制品业相关政策总数（条）": 39,
+    "国家碳达峰政策数（条）": 1,
+    "地方碳达峰/节能减排政策数（条）": 8,
+    "涉及省份数（个）": 7,
+    "符合条件省份数（企业>=5且总利润>0）": 3,
+    "安徽省企业数（家）": 5,
+    "安徽省总营业利润（亿元）": 263.5,
+    "安徽省总运营成本（亿元）": 2410.28,
+    "安徽省碳排放合规成本（亿元）": 120.51,
+    "安徽省营业利润下降百分比（%）": 45.74
+  },
+  "answer": [
+    "安徽省",
+    45.74
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/medium005_result.json b/assets/qa_raw/risk_assessment/medium005_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..ccb5d33aef8c230d507df24fc325362e9beba114
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/medium005_result.json
@@ -0,0 +1,38 @@
+{
+  "id": "medium005",
+  "question": "2022年，相关部委出台了促进钢铁工业高质量发展的指导意见，明确提出金属冶炼行业研发投入强度力争达到1.5%的政策目标。现聚焦以下范围：同时受到国家层面金属冶炼产业发展政策和地方金属冶炼和压延加工业相关政策覆盖的省份，且该省上市企业总数（以营业收入、营业利润和研发投入数据均不为空、营业收入大于零为准）不低于6家，同时全省总营业利润为正值。在此范围内，模拟地方政府出台强制合规要求——凡研发投入强度（研发投入÷营业收入）未达1.5%门槛的企业，须将研发投入强制补足至1.5%，补足部分直接计入成本并从营业利润中扣除。请问：在符合条件的省份中，哪个省份的企业需要补足的研发投入缺口总量最大？执行该合规要求后，该省金属冶炼和压延加工业的总营业利润预计下降多少个百分点？",
+  "guidelines": "依次回答省份名称和营业利润下降比例。下降比例以百分数表示，保留2位小数。如[\"山东省\", 3.75]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"金属冶炼\"的政策记录，找到41条金属冶炼和压延加工业相关政策，其中地方政策32条，涉及17个省份。",
+    "从company_profile.csv筛选行业=\"金属冶炼和压延加工业\"的企业，按省份统计，在上述17个省份中筛选营业收入金额、营业利润金额和研发投入金额均不为空且营业收入金额>0的企业，保留企业数>=6且总营业利润>0的省份，得到5个符合条件的省份：云南省（7家）、安徽省（7家）、山东省（9家）、江西省（8家）、河南省（6家）。",
+    "从company_operation_status.csv获取这5个省份金属冶炼企业的营业收入金额和研发投入金额，计算每家企业的研发投入占比（研发投入金额/营业收入金额×100%），筛选研发投入占比<1.5%的企业，计算每家不达标企业的额外研发投入=营业收入金额×1.5%-实际研发投入金额，汇总得到各省份额外研发投入总额：江西省37.69亿元、山东省9.93亿元、云南省8.94亿元、河南省6.31亿元、安徽省0.39亿元。",
+    "江西省需要额外增加的研发投入总额最多（37.69亿元），其中3家不达标企业分别需增加约17.61亿元、17.38亿元和2.70亿元。",
+    "计算江西省营业利润下降比例=额外研发投入总额/总营业利润×100%=3768902884.89元/63806804479.63元×100%=5.91%。"
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "从policy_release_status.csv中找到41条金属冶炼和压延加工业相关政策，其中32条地方政策。",
+    "从policy_resource.csv中分析32条地方政策全文，确认17个省份出台了涉及金属冶炼行业的地方产业发展政策。",
+    "从company_profile.csv中筛选出5个符合条件省份共37家金属冶炼和压延加工业企业。",
+    "从company_operation_status.csv中获取这37家企业的营业收入、营业利润和研发投入数据。"
+  ],
+  "milestone": {
+    "地方金属冶炼相关政策数（条）": 32,
+    "涉及省份数（个）": 17,
+    "符合条件省份数（个）": 5,
+    "江西省有效企业数（家）": 8,
+    "江西省研发不达标企业数（家）": 3,
+    "江西省额外研发投入总额（亿元）": 37.69,
+    "江西省总营业利润（亿元）": 638.07,
+    "江西省营业利润下降比例（%）": 5.91
+  },
+  "answer": [
+    "江西省",
+    5.91
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/medium006_result.json b/assets/qa_raw/risk_assessment/medium006_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..62720112dcc68d354dd9a44bcd9b4b8a09c2fe55
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/medium006_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "medium006",
+  "question": "2022年，国家将橡胶制品业列入新污染物治理试点，同时也要求塑料制品行业加快推进绿色低碳转型。在同时具备国家层面新污染物治理政策约束和地方橡胶和塑料制品政策支持、且省内上市企业总数不低于5家、总营业利润为正的省份中，平均营业利润率（以总营业利润除以总营业收入的比值衡量）最低的是哪个省份？鉴于该省企业本身盈利空间有限，一旦面临新污染物合规落地——以各企业运营成本（=营业收入-营业利润）的3%估算额外环保支出，且这部分费用全额从营业利润中扣减——请问该省橡胶和塑料制品业的总营业利润将因此下降多大比例？",
+  "guidelines": "依次回答省份名称和总营业利润下降的百分比。百分比保留2位小数。如[\"山东省\", 56.12]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段找到22条橡胶和塑料制品业相关政策，其中国家层面政策5条、地方政策17条。",
+    "分析17条地方政策覆盖的省份，共涉及12个省份。",
+    "从company_profile.csv筛选行业=\"橡胶和塑料制品业\"的企业共107家，按省份统计，在有地方政策覆盖的12个省份中筛选企业数>=5且总营业利润>0的省份，得到3个符合条件的省份：上海市（10家）、安徽省（8家）、山东省（5家）。",
+    "从company_operation_status.csv提取这3个省份橡胶和塑料制品业企业的营业收入金额和营业利润金额，计算各省平均营业利润率（总营业利润/总营业收入×100%）：上海市=7.49亿/215.69亿=3.47%、山东省=20.00亿/275.40亿=7.26%、安徽省=35.94亿/448.88亿=8.01%。上海市营业利润率最低。",
+    "计算上海市的新污染物合规成本冲击：各企业额外环保成本=各企业运营成本×3%，其中运营成本=营业收入-营业利润。上海市10家企业的额外环保成本总计=6.25亿元，总营业利润下降百分比=6.25/7.49×100%=83.39%。"
+  ],
+  "steps_num": 5,
+  "evidence": [
+    "从policy_release_status.csv中找到22条橡胶和塑料制品业相关政策，其中国家层面5条、地方17条覆盖12个省份。",
+    "从policy_resource.csv中分析22条政策全文，识别出新污染物治理和轻工业绿色转型两大国家级政策对橡胶和塑料制品业的具体要求。",
+    "从company_profile.csv中找到橡胶和塑料制品业企业107家，筛选出3个符合条件的省份共23家企业。",
+    "从company_operation_status.csv中获取这23家企业的营业收入和营业利润数据。"
+  ],
+  "milestone": {
+    "橡胶和塑料制品业相关政策总数（条）": 22,
+    "国家层面政策数（条）": 5,
+    "地方政策覆盖省份数（个）": 12,
+    "橡胶和塑料制品业企业总数（家）": 107,
+    "符合条件省份数（个）": 3,
+    "上海市企业数（家）": 10,
+    "上海市总营业收入（亿元）": 215.69,
+    "上海市总营业利润（亿元）": 7.49,
+    "上海市营业利润率（%）": 3.47,
+    "上海市额外环保成本总计（亿元）": 6.25,
+    "上海市总营业利润下降百分比（%）": 83.39
+  },
+  "answer": [
+    "上海市",
+    83.39
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/medium007_result.json b/assets/qa_raw/risk_assessment/medium007_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..88231a68ce68570e997f56d9150a55879beba554
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/medium007_result.json
@@ -0,0 +1,43 @@
+{
+  "id": "medium007",
+  "question": "2022年，若要评估各省半导体行业的市场结构风险，需重点关注市场集中程度——集中度越高，单一龙头企业的经营波动对全省产业的冲击越大。请在满足以下两个条件的省份范围内作答：一是该省已出台半导体行业政策；二是省内半导体业上市企业总数不少于10家（仅统计营业收入不为空且大于零的企业）。基于赫芬达尔-赫希曼指数（HHI=各企业营业收入占省内总营业收入百分比的平方和），哪个省份市场集中度最高？在此基础上，进一步模拟：若该省营业收入体量最大的企业遭遇外部供应链冲击，营业收入收缩30%而省内其余企业保持不变，新的HHI指数会变为多少？",
+  "guidelines": "依次回答省份名称、原HHI指数和冲击后HHI指数，HHI指数保留2位小数。如[\"上海市\", 1285.07, 1223.31]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"半导体\"的政策记录，找到44条半导体相关政策，其中全国性政策（province为\"全国\"）9条，地方政策35条，涉及15个省份。",
+    "从company_profile.csv筛选行业=\"半导体业\"的企业，按省份统计企业总数。在上述3个有集成电路专项政策的省份中，筛选企业总数不少于10家的省份：广东省54家、上海市27家、浙江省13家，均满足条件。",
+    "从company_operation_status.csv获取这3个省份半导体企业的营业收入金额，筛选营业收入金额不为空且大于0的企业。广东省54家、上海市27家、浙江省13家均全部满足条件。",
+    "计算各省份的HHI指数。HHI=Σ(企业营业收入/省内总营业收入×100)²。广东省：总营收2764.08亿元，HHI=680.27；上海市：总营收2474.39亿元，HHI=1285.07；浙江省：总营收255.75亿元，HHI=1596.76。浙江省HHI指数最高，表明市场集中度最高。",
+    "浙江省HHI最高，其营业收入排名第一的企业为华微创澜微电子公司，营收82.82亿元，市场份额占比32.38%。假设该企业因外部供应链冲击营收下降30%，新营收=82.82×0.70=57.98亿元。省内新总营收=255.75-82.82+57.98=230.90亿元。",
+    "重新计算冲击后浙江省各企业的市场份额并计算新HHI。龙头企业新份额=57.98/230.90×100=25.11%，其余12家企业份额相应调整。冲击后HHI=Σ(新份额)²=1302.75。HHI从1596.76下降至1302.75，下降293.01，虽然集中度有所降低，但仍处于中高集中度区间（HHI>1500为高集中度，1000-1500为中集中度），且省内总营收损失24.85亿元（占比9.72%），反映出浙江省半导体产业对龙头企业的高度依赖。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从policy_release_status.csv中找到44条半导体相关政策，其中地方政策35条涉及15个省份。",
+    "从policy_resource.csv中分析44条政策全文，筛选出3个省份（广东省、上海市、浙江省）出台了集成电路产业专项扶持政策，包含具体的产业补贴、流片补助、研发支持等措施。",
+    "从company_profile.csv中找到3个省份共94家半导体企业（广东省54家、上海市27家、浙江省13家）。"
+  ],
+  "milestone": {
+    "半导体相关政策总数（条）": 44,
+    "地方政策数（条）": 35,
+    "有集成电路专项政策的省份数（个）": 3,
+    "广东省HHI": 680.27,
+    "上海市HHI": 1285.07,
+    "浙江省HHI（原）": 1596.76,
+    "浙江省龙头企业营收（亿元）": 82.82,
+    "浙江省龙头企业市场份额（%）": 32.38,
+    "冲击后龙头企业营收（亿元）": 57.98,
+    "冲击后浙江省总营收（亿元）": 230.9,
+    "冲击后浙江省HHI": 1302.75,
+    "浙江省总营收损失占比（%）": 9.72
+  },
+  "answer": [
+    "浙江省",
+    1596.76,
+    1302.75
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/medium008_result.json b/assets/qa_raw/risk_assessment/medium008_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..60ed59de112500ba01fafdf1800777683bd26cfd
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/medium008_result.json
@@ -0,0 +1,49 @@
+{
+  "id": "medium008",
+  "question": "2022年，随着新能源汽车补贴政策进入退坡阶段，各省汽车制造业对政府补贴的依存程度成为衡量产业抗风险能力的关键变量。在已出台地方汽车产业专项扶持政策、且本省汽车制造业上市企业数量不低于10家的省份中（仅纳入政府奖励资金补贴、营业利润、营业收入三项数据均完整的企业，政府补贴依赖度定义为省内补贴总额与营业利润总额之比），哪个省份的补贴依赖度最高？如果对该省所有汽车制造企业同步实施50%补贴削减，且营业利润随之等额下降，则该省汽车制造业的整体营业利润将下降多少？",
+  "guidelines": "依次回答省份名称、补贴依赖度和营业利润下降比例。补贴依赖度和下降比例均以百分数表示，保留2位小数。如[\"广东省\", 45.20, 22.60]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"汽车\"的政策记录，找到69条汽车制造业相关政策，其中全国性政策16条，地方政策53条。",
+    "从policy_resource.csv中读取53条地方政策全文，分析哪些省份出台了汽车产业生产端专项扶持政策（含具体的产业补贴、补助、奖励或专项资金措施）。经分析，共有9个省份出台了此类政策：广东省5条（含广州市支持汽车及核心零部件产业稳链补链强链若干措施、广东省加快建设燃料电池汽车示范城市群行动计划等），上海市2条（含上海市制造业创新中心建设工程实施方案、充换电基础设施建设实施意见），山东省1条（青岛市加快新能源汽车产业高质量发展若干政策措施），江苏省1条（南京市新能源汽车换电模式应用试点实施方案），以及湖北省、湖南省、重庆市、辽宁省、吉林省各1条。",
+    "从company_profile.csv筛选行业=\"汽车制造业\"的企业共230家，按省份统计企业数。在上述9个有汽车产业扶持政策的省份中，筛选企业总数不少于10家的省份，得到4个符合条件的省份：上海市（22家）、山东省（18家）、广东省（27家）、江苏省（37家）。",
+    "从company_operation_status.csv获取这4个省份汽车制造业企业的政府奖励资金补贴、营业利润金额和营业收入金额数据，筛选三项数据均完整的企业。上海市20家、山东省18家、广东省26家、江苏省37家数据完整。",
+    "计算各省份的政府补贴依赖度=省内企业政府补贴总额/省内企业营业利润总额×100%。上海市：补贴总额50.91亿元/营业利润总额76.53亿元=66.53%；广东省：59.42亿元/453.02亿元=13.12%；江苏省：6.55亿元/60.89亿元=10.76%；山东省：15.85亿元/178.59亿元=8.87%。上海市补贴依赖度最高，达66.53%。",
+    "模拟上海市政府补贴削减50%的冲击：补贴削减额=50.91亿元×50%=25.46亿元，营业利润等额减少25.46亿元。原营业利润76.53亿元，新营业利润=76.53-25.46=51.07亿元，营业利润下降比例=25.46/76.53×100%=33.27%。上海市汽车制造业高补贴依赖度（66.53%）叠加低营业利润率（0.74%），使其在补贴退坡情景下面临的利润冲击最为显著。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从policy_release_status.csv中找到69条汽车制造业相关政策，其中全国性政策16条，地方政策53条。",
+    "从policy_resource.csv中分析3条全国性政策全文，确认新能源汽车购置税免征政策仅延续至2023年底，补贴退坡风险明确。",
+    "从policy_resource.csv中分析53条地方政策全文，筛选出9个省份出台了汽车产业生产端专项扶持政策。",
+    "从company_profile.csv中找到4个符合条件省份（企业>=10且有产业扶持政策）的104家汽车制造业企业。",
+    "从company_operation_status.csv中获取101家数据完整企业的政府补贴、营业利润和营业收入数据。"
+  ],
+  "milestone": {
+    "汽车相关政策总数（条）": 69,
+    "全国性政策数（条）": 16,
+    "地方政策数（条）": 53,
+    "有汽车产业扶持政策的省份数（个）": 9,
+    "符合条件省份数（企业>=10且有扶持政策）": 4,
+    "上海市数据完整企业数（家）": 20,
+    "上海市政府补贴总额（亿元）": 50.91,
+    "上海市营业利润总额（亿元）": 76.53,
+    "上海市营业收入总额（亿元）": 10279.21,
+    "上海市补贴依赖度（%）": 66.53,
+    "广东省补贴依赖度（%）": 13.12,
+    "江苏省补贴依赖度（%）": 10.76,
+    "山东省补贴依赖度（%）": 8.87,
+    "上海市补贴削减额（亿元）": 25.46,
+    "上海市新营业利润（亿元）": 51.07,
+    "上海市营业利润下降比例（%）": 33.27
+  },
+  "answer": [
+    "上海市",
+    66.53,
+    33.27
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/medium009_result.json b/assets/qa_raw/risk_assessment/medium009_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ab89bdccca73ab019115d97642de63658aadaa7
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/medium009_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "medium009",
+  "question": "2022年，纺织鞋服业受出口需求放缓与国内消费压力的双重影响，结构性风险进一步暴露。在同时受到国家层面纺织产业发展政策和地方纺织行业相关政策双重覆盖、且省内拥有正营业收入的上市企业总数不低于10家的省份范围内，哪个省份的纺织鞋服业市场结构最为集中——即按各企业营业收入占省内总营业收入的百分比计算市场份额后，所有市场份额百分比的平方加总（HHI）最大？进一步地，若该省第一大企业突遭外部需求骤降冲击，营业收入萎缩40%，而省内其余企业营收保持稳定，请重新计算此时的HHI指数。",
+  "guidelines": "依次回答省份名称和冲击后的HHI指数。HHI指数保留2位小数。如[\"广东省\", 1538.67]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"纺织\"的政策记录，找到21条纺织相关政策。地方层面共14条，涉及11个省份：上海市、广东省、福建省、山东省、四川省、辽宁省、广西壮族自治区、湖南省、陕西省、河北省、新疆维吾尔自治区。",
+    "从company_profile.csv筛选行业=\"纺织鞋服业\"的企业，关联company_operation_status.csv获取营业收入数据，仅保留营业收入大于0的企业，按省份统计企业数。在上述11个有地方政策覆盖的省份中，筛选企业数>=10的省份，得到4个符合条件的省份：上海市（13家）、山东省（11家）、广东省（33家）、福建省（17家）。",
+    "计算各省份的营业收入HHI指数（HHI=各企业营业收入占全省总营业收入百分比的平方和）：上海市HHI=2295.32、山东省HHI=2368.17、广东省HHI=1223.33、福建省HHI=2917.81。福建省HHI指数最高。",
+    "福建省纺织鞋服业总营业收入为1075.97亿元，营业收入排名第一的企业营业收入为536.51亿元，市场份额为49.86%。",
+    "模拟外部需求冲击：该企业营业收入下降40%后变为321.91亿元，福建省总营业收入变为861.36亿元。重新计算各企业市场份额并求HHI指数，冲击后HHI=2069.92。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从policy_release_status.csv中找到21条纺织相关政策。",
+    "从policy_resource.csv中分析21条政策全文，识别出7条国家层面纺织产业政策和14条地方纺织产业政策，覆盖11个省份。",
+    "从company_profile.csv中筛选出纺织鞋服业企业，关联company_operation_status.csv后得到4个符合条件省份共74家有正营业收入的企业。"
+  ],
+  "milestone": {
+    "纺织相关政策总数（条）": 21,
+    "国家层面纺织政策数（条）": 7,
+    "地方层面纺织政策数（条）": 14,
+    "有地方政策覆盖的省份数（个）": 11,
+    "符合条件省份数（企业>=10且有地方政策）": 4,
+    "福建省纺织企业数（家）": 17,
+    "福建省总营业收入（亿元）": 1075.97,
+    "福建省原始HHI": 2917.81,
+    "福建省排名第一企业营业收入（亿元）": 536.51,
+    "冲击后排名第一企业营业收入（亿元）": 321.91,
+    "冲击后福建省总营业收入（亿元）": 861.36,
+    "冲击后福建省HHI": 2069.92
+  },
+  "answer": [
+    "福建省",
+    2069.92
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/medium010_result.json b/assets/qa_raw/risk_assessment/medium010_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..079d81c6a3b876a5d4cdc71927f16d53e3f9f3f1
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/medium010_result.json
@@ -0,0 +1,41 @@
+{
+  "id": "medium010",
+  "question": "2022年，食品饮料业上市企业在受到国家轻工业高质量发展政策指引的同时，部分省份也出台了针对性的地方食品产业政策。在同时满足两项条件的省份中（条件一：省内有国家轻工业政策与地方食品政策的双重覆盖；条件二：省内有营业收入记录的食品饮料业上市企业不少于6家），哪个省份企业的平均资产负债率水平最高，意味着其对外部融资依赖最深？在此前提下，若货币政策收紧、基准利率上升2个百分点，以各企业总负债额乘以2%估算新增利息负担，所有企业的额外利息成本加总后，这一总额相当于该省食品饮料业总营业利润的多少？",
+  "guidelines": "依次回答省份名称和额外利息成本占总营业利润的比例。比例以百分数表示，保留2位小数。如[\"湖北省\", 35.20]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"食品\"的政策记录，找到16条食品饮料业相关政策，其中国家层面（部委政策）3条，地方政策12条（不含标注为全国的1条）。",
+    "经过对地方政策内容的分析，提取出台了涉及食品饮料业的地方产业发展政策的省份共9个：甘肃省、河南省、云南省、四川省、湖南省、海南省、宁夏回族自治区、河北省、贵州省。",
+    "从company_profile.csv筛选行业为食品饮料业的企业，按省份统计，在上述9个有地方食品政策的省份中筛选有营业收入的企业数>=6的省份，得到4个符合条件的省份：河南省（8家）、四川省（9家）、湖南省（14家）、河北省（6家）。",
+    "从company_operation_status.csv提取这4个省份食品饮料业企业的资产负债率，计算各省份企业平均资产负债率：湖南省45.72%、四川省42.37%、河南省32.82%、河北省30.40%。湖南省平均资产负债率最高。",
+    "从company_operation_status.csv提取湖南省14家食品饮料业企业的总负债数据，计算总负债合计为415.40亿元，总营业利润合计为16.46亿元。",
+    "计算利率上升2个百分点带来的额外利息成本：额外利息成本=总负债×2%=415.40×2%=8.31亿元。额外利息成本占总营业利润的比例=8.31/16.46×100%=50.47%。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从policy_release_status.csv中找到16条食品饮料业相关政策，其中国家层面3条、地方政策12条。",
+    "从policy_resource.csv中分析16条政策全文，确认3条国家层面轻工业政策涉及食品行业发展要求，9个省份出台了涉及食品饮料业的地方产业政策。",
+    "从company_profile.csv中筛选出4个符合条件省份（有地方食品政策且企业数>=6）的食品饮料业企业共37家。",
+    "从company_operation_status.csv中获取这37家企业的资产负债率、总负债和营业利润数据。"
+  ],
+  "milestone": {
+    "食品饮料业相关政策总数（条）": 16,
+    "国家层面政策数（条）": 3,
+    "地方食品政策涉及省份数（个）": 9,
+    "符合条件省份数（企业>=6）": 4,
+    "湖南省食品饮料业企业数（家）": 14,
+    "湖南省平均资产负债率（%）": 45.72,
+    "湖南省食品饮料业总负债（亿元）": 415.4,
+    "湖南省食品饮料业总营业利润（亿元）": 16.46,
+    "额外利息成本（亿元）": 8.31,
+    "额外利息占总营业利润比例（%）": 50.47
+  },
+  "answer": [
+    "湖南省",
+    50.47
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  }
+}
diff --git a/assets/qa_raw/risk_assessment/medium011_result.json b/assets/qa_raw/risk_assessment/medium011_result.json
new file mode 100644
index 0000000000000000000000000000000000000000..266bd5dd550750f2accb58f16fe6f3bc03fca060
--- /dev/null
+++ b/assets/qa_raw/risk_assessment/medium011_result.json
@@ -0,0 +1,42 @@
+{
+  "id": "medium011",
+  "question": "2022年，国家发布一系列采矿业相关产业政策，同时各省也出台了采矿业相关的产业发展政策。在双重政策覆盖之下，且有正营业收入的上市企业总数不低于6家、全省采矿业总营业利润大于零的省份中，哪个省份的采矿业市场份额分布最为集中（市场集中度以营业收入HHI衡量）？此外，鉴于采矿业对国际大宗商品价格高度敏感，若该省营业收入体量最大的采矿企业因国际市场剧烈波动遭遇营收下滑40%（其余企业营收保持不变），重新计算各企业市场份额后，全省采矿业的HHI指数将变动至多少？",
+  "guidelines": "依次回答省份名称和冲击后的HHI指数。HHI指数保留2位小数。如[\"内蒙古自治区\", 3200.50]。如果无法找到相关数据，请回答\"未查询到相关数据\"",
+  "steps": [
+    "从policy_release_status.csv筛选industry字段包含\"采矿\"的政策记录，找到27条采矿业相关政策。其中部委政策6条、地方政策21条。",
+    "分析21条地方采矿业相关政策，涉及17个省份。",
+    "从company_profile.csv筛选行业=\"采矿业\"的企业共143家，关联company_operation_status.csv获取运营数据。筛选营业收入>0且营业利润数据完整的企业，按省份统计。在上述17个有地方采矿业政策的省份中，筛选企业总数>=6家且总营业利润>0的省份，得到4个符合条件的省份：内蒙古自治区（8家）、山西省（8家）、新疆维吾尔自治区（6家）、河南省（8家）。",
+    "计算各省份采矿业营业收入HHI指数（HHI=各企业营业收入占省份总营业收入比例的百分比平方和）：内蒙古自治区HHI=3451.82，山西省HHI=2308.12，新疆维吾尔自治区HHI=4666.82，河南省HHI=3517.82。新疆维吾尔自治区HHI最高，为4666.82。",
+    "新疆维吾尔自治区6家采矿业企业营业收入分别为835.90亿元、594.09亿元、44.08亿元、20.15亿元、6.69亿元、1.97亿元，总营业收入1502.87亿元。排名第一企业占比55.62%。",
+    "模拟冲击：排名第一企业营业收入下降40%，从835.90亿元降至501.54亿元，其他企业不变。新总营业收入=1168.51亿元。重新计算各企业市场份额并求HHI：新HHI=(501.54/1168.51×100)²+(594.09/1168.51×100)²+(44.08/1168.51×100)²+(20.15/1168.51×100)²+(6.69/1168.51×100)²+(1.97/1168.51×100)²=4444.61。"
+  ],
+  "steps_num": 6,
+  "evidence": [
+    "从policy_release_status.csv中找到27条采矿业相关政策，其中6条部委政策、21条地方政策。",
+    "从policy_resource.csv中分析27条政策全文，筛选出3条国家层面碳达峰绿色转型政策和17个省份的21条地方采矿业政策。",
+    "从company_profile.csv中找到143家采矿业企业，关联company_operation_status.csv后筛选出4个符合条件省份共30家企业。",
+    "从company_operation_status.csv中获取这30家企业的营业收入数据，用于计算HHI指数。"
+  ],
+  "milestone": {
+    "采矿业相关政策总数（条）": 27,
+    "国家层面碳达峰绿色转型政策数（条）": 3,
+    "地方采矿业政策涉及省份数（个）": 17,
+    "符合条件省份数（>=6家企业且利润>0）": 4,
+    "新疆维吾尔自治区采矿业企业数（家）": 6,
+    "新疆维吾尔自治区总营业收入（亿元）": 1502.87,
+    "新疆维吾尔自治区原始HHI": 4666.82,
+    "排名第一企业营业收入（亿元）": 835.9,
+    "冲击后排名第一企业营业收入（亿元）": 501.54,
+    "冲击后总营业收入（亿元）": 1168.51,
+    "冲击后HHI": 4444.61
+  },
+  "answer": [
+    "新疆维吾尔自治区",
+    "4444.61"
+  ],
+  "metadata": {
+    "db": "bm_rag_qa",
+    "level": "medium",
+    "category": "risk_assessment"
+  }
+}
diff --git a/dataclaw/__init__.py b/dataclaw/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/dataclaw/__pycache__/__init__.cpython-312.pyc b/dataclaw/__pycache__/__init__.cpython-312.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..31ab3453b8a2aa414cbdcfc2819ee53058080fb1
Binary files /dev/null and b/dataclaw/__pycache__/__init__.cpython-312.pyc differ
diff --git a/dataclaw/__pycache__/lib_tasks.cpython-312.pyc b/dataclaw/__pycache__/lib_tasks.cpython-312.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..0484ec0ddd21b92d2336cbfdc9423c4a5d80ff76
Binary files /dev/null and b/dataclaw/__pycache__/lib_tasks.cpython-312.pyc differ
diff --git a/dataclaw/build_tasks.py b/dataclaw/build_tasks.py
new file mode 100644
index 0000000000000000000000000000000000000000..f85d5ce6d294ed191bda14ddd8769aee140be34b
--- /dev/null
+++ b/dataclaw/build_tasks.py
@@ -0,0 +1,209 @@
+#!/usr/bin/env python3
+"""Generate DataClaw/EIP-style benchmark tasks from QA JSON files."""
+
+from __future__ import annotations
+
+import argparse
+import json
+from pathlib import Path
+from typing import Any
+
+
+def _slug(text: str) -> str:
+    return "".join(ch if ch.isalnum() else "_" for ch in text.lower()).strip("_")
+
+
+def _discover_database_files(project_root: Path) -> list[dict[str, str]]:
+    database_root = project_root / "assets" / "database"
+    files: list[dict[str, str]] = []
+    for path in sorted(database_root.rglob("*")):
+        if not path.is_file():
+            continue
+        rel = path.relative_to(database_root).as_posix()
+        files.append({"source": f"database/{rel}", "dest": f"database/{rel}"})
+    return files
+
+
+def _answer_repr(answer: Any) -> str:
+    return json.dumps(answer, ensure_ascii=False)
+
+
+def _validate_payload(payload: dict[str, Any], qa_file: Path) -> None:
+    required = ("id", "question", "guidelines", "answer")
+    missing = [key for key in required if key not in payload]
+    if missing:
+        raise ValueError(f"{qa_file} missing required keys: {', '.join(missing)}")
+    if not isinstance(payload["question"], str) or not payload["question"].strip():
+        raise ValueError(f"{qa_file} has invalid question")
+    if not isinstance(payload["guidelines"], str) or not payload["guidelines"].strip():
+        raise ValueError(f"{qa_file} has invalid guidelines")
+
+
+def _mk_rubric(answer: Any, guidelines: str) -> str:
+    answer_json = _answer_repr(answer)
+    if isinstance(answer, list):
+        n = len(answer)
+        return f"""### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`{answer_json}`
+
+Scoring rules:
+- The gold answer is a list with N={n} parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_{n - 1}` each as 0 or 1.
+- Return `total = (sum(part_i)) / {n}` exactly.
+- If the model output is missing or cannot be parsed into {n} comparable parts, score all parts 0.
+"""
+
+    return f"""### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`{answer_json}`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+"""
+
+
+def _build_task_markdown(
+    *,
+    task_id: str,
+    task_name: str,
+    category: str,
+    item_id: str,
+    question: str,
+    guidelines: str,
+    answer: Any,
+    timeout_seconds: int,
+    workspace_files: list[dict[str, str]],
+) -> str:
+    ws_json = json.dumps(workspace_files, ensure_ascii=False, indent=2)
+    rubric = _mk_rubric(answer, guidelines)
+    expected = (
+        "Agent should read the provided `database/` files, compute the result, and return the final answer. "
+        "The final answer must follow the required output format."
+    )
+    criteria = (
+        "- [ ] Final answer semantically matches the gold `answer`.\n"
+        "- [ ] Output format follows `guidelines`."
+    )
+    data_sources = (
+        "You may use files under `./database/` and web search."
+        if category == "international_comparison"
+        else "Only use files under `./database/`."
+    )
+    gold_file = f"qa_gold/{category}/{item_id}.json"
+    return f"""---
+id: {task_id}
+name: {task_name}
+category: {category}
+grading_type: llm_judge
+timeout_seconds: {timeout_seconds}
+gold_file: {gold_file}
+workspace_files: {ws_json}
+---
+
+## Prompt
+
+{question}
+
+Output guidelines:
+{guidelines}
+
+{data_sources}
+
+## Expected Behavior
+
+{expected}
+
+## Grading Criteria
+
+{criteria}
+
+## LLM Judge Rubric
+
+{rubric}
+"""
+
+
+def build(project_root: Path) -> int:
+    qa_root = project_root / "assets" / "qa_raw"
+    gold_root = project_root / "assets" / "qa_gold"
+    tasks_root = project_root / "tasks"
+
+    gold_root.mkdir(parents=True, exist_ok=True)
+    tasks_root.mkdir(parents=True, exist_ok=True)
+
+    for stale in tasks_root.glob("task_*.md"):
+        stale.unlink()
+
+    workspace_files = _discover_database_files(project_root)
+    if not workspace_files:
+        raise RuntimeError("No files found under assets/database")
+    qa_files = sorted(qa_root.rglob("*_result.json"))
+    if not qa_files:
+        raise RuntimeError(f"No QA files found in {qa_root}")
+
+    for idx, qa_file in enumerate(qa_files, start=1):
+        payload = json.loads(qa_file.read_text(encoding="utf-8"))
+        _validate_payload(payload, qa_file)
+        category = str(payload.get("metadata", {}).get("category", qa_file.parent.name))
+        level = str(payload.get("metadata", {}).get("level", "unknown"))
+        item_id = str(payload.get("id", qa_file.stem.replace("_result", "")))
+        question = str(payload.get("question", "")).strip()
+        guidelines = str(payload.get("guidelines", "")).strip()
+        answer = payload.get("answer")
+        task_id = f"task_{idx:03d}_{_slug(category)}_{_slug(level)}_{_slug(item_id)}"
+        task_name = f"{category}-{level}-{item_id}"
+
+        gold_dir = gold_root / category
+        gold_dir.mkdir(parents=True, exist_ok=True)
+        gold_payload = {
+            "id": item_id,
+            "question": question,
+            "guidelines": guidelines,
+            "answer": answer,
+            "metadata": payload.get("metadata", {}),
+            "steps": payload.get("steps", []),
+            "steps_num": payload.get("steps_num", 0),
+            "milestone": payload.get("milestone", {}),
+        }
+        (gold_dir / f"{item_id}.json").write_text(
+            json.dumps(gold_payload, ensure_ascii=False, indent=2),
+            encoding="utf-8",
+        )
+
+        task_md = _build_task_markdown(
+            task_id=task_id,
+            task_name=task_name,
+            category=category,
+            item_id=item_id,
+            question=question,
+            guidelines=guidelines,
+            answer=answer,
+            timeout_seconds=1200,
+            workspace_files=workspace_files,
+        )
+        (tasks_root / f"{task_id}.md").write_text(task_md, encoding="utf-8")
+
+    return len(qa_files)
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description="Build EIP OpenClaw benchmark tasks")
+    parser.add_argument(
+        "--project-root",
+        default=str(Path(__file__).resolve().parent.parent),
+        help="Project root path",
+    )
+    args = parser.parse_args()
+    project_root = Path(args.project_root).resolve()
+    count = build(project_root)
+    print(f"Generated {count} tasks.")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/dataclaw/eval/__init__.py b/dataclaw/eval/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/dataclaw/eval/run_batch.py b/dataclaw/eval/run_batch.py
new file mode 100644
index 0000000000000000000000000000000000000000..32823f5a91e37d8b8c76196ef35319ff682ad930
--- /dev/null
+++ b/dataclaw/eval/run_batch.py
@@ -0,0 +1,1043 @@
+"""
+DataClaw — Host-side benchmark orchestrator.
+
+Runs each task in an isolated Docker container. The host manages container
+lifecycle, workspace injection, agent execution, LLM-judge grading, and
+result collection.
+
+Usage:
+    python dataclaw/eval/run_batch.py --model openrouter/anthropic/claude-sonnet-4.6
+    python dataclaw/eval/run_batch.py --model ... --suite task_001,task_002
+    python dataclaw/eval/run_batch.py --model ... --parallel 4
+    python dataclaw/eval/run_batch.py --task tasks/task_001_xxx.md
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import logging
+import os
+import re
+import subprocess
+import sys
+import tempfile
+import threading
+import time
+import uuid
+from concurrent.futures import ThreadPoolExecutor, as_completed
+from datetime import datetime
+from pathlib import Path
+from typing import Any, Dict, List, Optional
+
+try:
+    from dotenv import load_dotenv
+    load_dotenv()
+except ImportError:
+    pass
+
+# Allow imports from project root
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..", ".."))
+
+from dataclaw.lib_tasks import Task, TaskLoader
+from dataclaw.utils.docker_utils import (
+    JUDGE_CUSTOM_API_KEY,
+    JUDGE_CUSTOM_BASE_URL,
+    JUDGE_CUSTOM_MODEL_ID,
+    close_proc_log,
+    collect_output,
+    collect_transcript,
+    detect_transcript_errors,
+    extract_usage_from_jsonl,
+    onboard_openclaw,
+    register_custom_provider,
+    remove_container,
+    set_model,
+    setup_workspace,
+    start_container,
+    start_gateway,
+)
+from dataclaw.utils.grading import GradeResult, grade_task, _run_judge_in_container
+from dataclaw.utils.process_grading import (
+    parse_trajectory,
+    compute_efficiency,
+    build_gpr_judge_prompt,
+    parse_gpr_judge_response,
+    compute_tgpr,
+)
+
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s [%(levelname)s] %(message)s",
+    datefmt="%H:%M:%S",
+)
+logger = logging.getLogger(__name__)
+
+# ---------------------------------------------------------------------------
+# Configuration from environment
+# ---------------------------------------------------------------------------
+
+ROOT_DIR = Path(__file__).resolve().parent.parent.parent
+TASKS_DIR = ROOT_DIR / os.environ.get("TASKS_SUBDIR", "tasks")
+ASSETS_DIR = ROOT_DIR / "assets"
+OUTPUT_DIR = ROOT_DIR / os.environ.get("OUTPUT_SUBDIR", "output")
+
+DEFAULT_MODEL = os.environ.get("DEFAULT_MODEL", "")
+DEFAULT_PARALLEL = int(os.environ.get("DEFAULT_PARALLEL", "1"))
+DEFAULT_JUDGE_MODEL = os.environ.get("JUDGE_MODEL", "openrouter/anthropic/claude-opus-4.5")
+TIMEOUT_MULTIPLIER = float(os.environ.get("BENCHMARK_TIMEOUT_MULTIPLIER", "1.0"))
+BENCHMARK_RUNS = int(os.environ.get("BENCHMARK_RUNS", "1"))
+TMP_WORKSPACE = os.environ.get("TMP_WORKSPACE", "/tmp_workspace")
+
+OPENROUTER_API_KEY = os.environ.get("OPENROUTER_API_KEY", "")
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+def _slugify_model(model_id: str) -> str:
+    return re.sub(r"[^a-zA-Z0-9.\-_]", "_", model_id.rsplit("/", 1)[-1])
+
+
+def _get_git_version() -> str:
+    try:
+        r = subprocess.run(
+            ["git", "rev-parse", "--short", "HEAD"],
+            capture_output=True, text=True, encoding="utf-8", timeout=2, check=False, cwd=ROOT_DIR,
+        )
+        return r.stdout.strip() if r.returncode == 0 else ""
+    except (subprocess.SubprocessError, FileNotFoundError, OSError):
+        return ""
+
+
+def _validate_openrouter_model(model_id: str) -> bool:
+    """Basic check that the model exists on OpenRouter (skippable)."""
+    if os.environ.get("DATACLAW_SKIP_OPENROUTER_MODEL_VALIDATION", "").strip().lower() in (
+        "1", "true", "yes",
+    ):
+        return True
+    if os.environ.get("OPENCLAW_CUSTOM_BASE_URL", "").strip():
+        return True
+
+    bare = model_id
+    if bare.startswith("openrouter/"):
+        bare = bare[len("openrouter/"):]
+    if bare.startswith("bailian/") or "/" not in bare:
+        return True
+    if not OPENROUTER_API_KEY:
+        logger.warning("OPENROUTER_API_KEY not set, skipping model validation")
+        return True
+
+    from urllib import error, request as urlreq
+    encoded = bare.replace("/", "%2F")
+    url = f"https://openrouter.ai/api/v1/models/{encoded}"
+    req = urlreq.Request(url, headers={
+        "Authorization": f"Bearer {OPENROUTER_API_KEY}",
+        "HTTP-Referer": "https://github.com/GTMLLab/DataClaw",
+        "X-Title": "DataClaw",
+    }, method="GET")
+    try:
+        with urlreq.urlopen(req, timeout=10):
+            return True
+    except error.HTTPError as exc:
+        if exc.code == 404:
+            logger.error("Model '%s' not found on OpenRouter", bare)
+            return False
+        return True
+    except (error.URLError, OSError):
+        return True
+
+
+# ---------------------------------------------------------------------------
+# Progress file helpers (for --resume support)
+# ---------------------------------------------------------------------------
+
+_progress_lock = threading.Lock()
+
+
+def _progress_path(model_slug: str) -> Path:
+    return OUTPUT_DIR / f"progress_{model_slug}.json"
+
+
+def _load_progress(path: Path) -> Optional[Dict[str, Any]]:
+    """Load progress file, returning the parsed dict or None."""
+    if not path.exists():
+        return None
+    try:
+        return json.loads(path.read_text(encoding="utf-8"))
+    except (json.JSONDecodeError, OSError) as exc:
+        logger.warning("Failed to read progress file %s: %s", path, exc)
+        return None
+
+
+def _save_progress(
+    path: Path,
+    model: str,
+    suite: str,
+    runs: int,
+    completed: List[Dict[str, Any]],
+) -> None:
+    """Atomically write progress file (temp file + os.replace)."""
+    data = {
+        "model": model,
+        "suite": suite,
+        "runs": runs,
+        "completed": completed,
+    }
+    path.parent.mkdir(parents=True, exist_ok=True)
+    tmp_path = path.with_suffix(".tmp")
+    tmp_path.write_text(
+        json.dumps(data, indent=2, ensure_ascii=False), encoding="utf-8"
+    )
+    os.replace(str(tmp_path), str(path))
+
+
+# ---------------------------------------------------------------------------
+# Process grading helpers
+# ---------------------------------------------------------------------------
+
+PROCESS_JUDGE_TIMEOUT_SECONDS = 180
+
+
+def _load_gold_process_data(task: Task) -> Optional[Dict[str, Any]]:
+    """Load gold process fields (steps, milestone, steps_num) from qa_gold."""
+    gold_file = task.frontmatter.get("gold_file", "")
+    if not gold_file:
+        return None
+    gold_path = ASSETS_DIR / gold_file
+    if not gold_path.exists():
+        logger.warning("Gold file not found: %s", gold_path)
+        return None
+    try:
+        data = json.loads(gold_path.read_text(encoding="utf-8"))
+    except (json.JSONDecodeError, OSError) as exc:
+        logger.warning("Failed to read gold file %s: %s", gold_path, exc)
+        return None
+    if not data.get("milestone") or not data.get("steps"):
+        return None
+    return data
+
+
+GPR_MAX_ATTEMPTS = 5
+
+
+def _extract_assistant_text(transcript_path: Path) -> str:
+    """Concatenate all assistant text content from a transcript JSONL.
+
+    Returns empty string on missing file or read error (callers treat
+    empty result as an L2 failure).
+    """
+    if not transcript_path.exists():
+        return ""
+    try:
+        text = transcript_path.read_text(encoding="utf-8", errors="replace")
+    except OSError:
+        return ""
+    out = ""
+    for line in text.splitlines():
+        line = line.strip()
+        if not line:
+            continue
+        try:
+            entry = json.loads(line)
+        except json.JSONDecodeError:
+            continue
+        if entry.get("type") != "message":
+            continue
+        msg = entry.get("message", {})
+        if msg.get("role") != "assistant":
+            continue
+        for item in msg.get("content", []) or []:
+            if isinstance(item, dict) and item.get("type") == "text":
+                out += item.get("text", "") or ""
+    return out
+
+
+def _run_process_grading(
+    *,
+    container_id: str,
+    task_id: str,
+    transcript_path: Path,
+    gold_data: Dict[str, Any],
+    outcome_score: float,
+    output_dir: Path,
+) -> Dict[str, Any]:
+    """Run process grading conditional on outcome score.
+
+    Full-score tasks (score >= 1.0) get only Efficiency. Incorrect tasks
+    (score < 1.0) get only GPR (via LLM Judge, with GPR_MAX_ATTEMPTS retries
+    across three layers: runtime / transcript / parse), TGPR and TPE.
+    Raises RuntimeError if the GPR judge fails every attempt.
+    """
+    milestones = gold_data.get("milestone", {})
+    gold_steps = gold_data.get("steps", [])
+    steps_num = gold_data.get("steps_num", len(gold_steps))
+
+    if outcome_score >= 1.0:
+        # --- Full score: Efficiency only ---
+        steps = parse_trajectory(transcript_path)
+        eff = compute_efficiency(steps, steps_num)
+        process_scores: Dict[str, Any] = {"efficiency": eff.to_dict()}
+        (output_dir / "process_score.json").write_text(
+            json.dumps(process_scores, indent=2, ensure_ascii=False), encoding="utf-8"
+        )
+        logger.info(
+            "[%s] Process: efficiency=%.2f (full score, GPR/TGPR/TPE skipped)",
+            task_id,
+            eff.efficiency if eff.efficiency is not None else 0.0,
+        )
+        return process_scores
+
+    # --- Incorrect: GPR + TGPR + TPE only ---
+    steps = parse_trajectory(transcript_path)
+    gpr_prompt = build_gpr_judge_prompt(
+        steps=steps,
+        milestones=milestones,
+        gold_steps=gold_steps,
+        final_answer_correct=False,
+    )
+
+    gpr_result = None
+    last_err: Optional[str] = None
+
+    for attempt in range(1, GPR_MAX_ATTEMPTS + 1):
+        # Clear previous judge sessions before each attempt
+        subprocess.run(
+            ["docker", "exec", container_id, "/bin/bash", "-c",
+             "rm -rf /root/.openclaw/agents/judge/sessions/*"],
+            capture_output=True, text=True, encoding="utf-8",
+        )
+
+        # L1: runtime
+        try:
+            _run_judge_in_container(container_id, gpr_prompt)
+        except RuntimeError as exc:
+            last_err = f"runtime: {exc}"
+            logger.warning("[%s] GPR attempt %d/%d failed — %s",
+                           task_id, attempt, GPR_MAX_ATTEMPTS, last_err)
+            continue
+
+        gpr_transcript = collect_transcript(
+            container_id, output_dir, agent_id="judge",
+            output_filename="judge_process_chat.jsonl",
+        )
+
+        # L2: transcript validity
+        if not gpr_transcript.exists():
+            last_err = "transcript: file missing"
+            logger.warning("[%s] GPR attempt %d/%d failed — %s",
+                           task_id, attempt, GPR_MAX_ATTEMPTS, last_err)
+            continue
+
+        tr_err = detect_transcript_errors(gpr_transcript)
+        if tr_err:
+            last_err = f"transcript: {tr_err}"
+            logger.warning("[%s] GPR attempt %d/%d failed — %s",
+                           task_id, attempt, GPR_MAX_ATTEMPTS, last_err)
+            continue
+
+        gpr_raw_text = _extract_assistant_text(gpr_transcript)
+        if not gpr_raw_text.strip():
+            last_err = "transcript: empty assistant response"
+            logger.warning("[%s] GPR attempt %d/%d failed — %s",
+                           task_id, attempt, GPR_MAX_ATTEMPTS, last_err)
+            continue
+
+        # L3: format — parse_gpr_judge_response falls back to a sentinel
+        # GPRResult when required fields are missing; detect it explicitly.
+        try:
+            parsed_gpr = parse_gpr_judge_response(gpr_raw_text, milestones)
+        except Exception as exc:
+            last_err = f"format: parse failed: {exc}"
+            logger.warning("[%s] GPR attempt %d/%d failed — %s",
+                           task_id, attempt, GPR_MAX_ATTEMPTS, last_err)
+            continue
+
+        if "could not be parsed" in (parsed_gpr.chain_summary or ""):
+            last_err = "format: judge response missing required fields (milestones)"
+            logger.warning("[%s] GPR attempt %d/%d failed — %s",
+                           task_id, attempt, GPR_MAX_ATTEMPTS, last_err)
+            continue
+
+        gpr_result = parsed_gpr
+        logger.info("[%s] GPR judge succeeded on attempt %d", task_id, attempt)
+        break
+
+    if gpr_result is None:
+        raise RuntimeError(
+            f"GPR judge failed after {GPR_MAX_ATTEMPTS} attempts: {last_err}"
+        )
+
+    tgpr_result = compute_tgpr(gpr_result, s_gold=steps_num)
+
+    process_scores = {
+        "gpr": gpr_result.to_dict(),
+        "tgpr": tgpr_result.to_dict(),
+    }
+    (output_dir / "process_score.json").write_text(
+        json.dumps(process_scores, indent=2, ensure_ascii=False), encoding="utf-8"
+    )
+    logger.info(
+        "[%s] Process: GPR=%.2f, TGPR=%.2f, TPE=%.2f (incorrect, efficiency skipped)",
+        task_id,
+        gpr_result.gpr,
+        tgpr_result.tgpr,
+        tgpr_result.tpe,
+    )
+    return process_scores
+
+
+# ---------------------------------------------------------------------------
+# Single-task execution
+# ---------------------------------------------------------------------------
+
+def run_single_task(
+    task: Task,
+    model: str,
+    judge_model: str,
+    timeout_multiplier: float,
+) -> Dict[str, Any]:
+    """Execute a single task in an isolated container. Thread-safe."""
+    timestamp = datetime.now().strftime("%Y%m%d_%H%M")
+    run_id = uuid.uuid4().hex[:6]
+    short_model = _slugify_model(model)
+    suffix = f"{short_model}_{timestamp}_{run_id}"
+    container_id = f"{task.task_id}_{suffix}"
+
+    # Truncate container name to Docker's 128-char limit
+    if len(container_id) > 128:
+        container_id = container_id[:128]
+
+    output_dir = OUTPUT_DIR / task.task_id / suffix
+    output_dir.mkdir(parents=True, exist_ok=True)
+
+    result: Dict[str, Any] = {
+        "task_id": task.task_id,
+        "model": model,
+        "scores": {},
+        "grade": None,
+        "process_grade": None,
+        "usage": {},
+        "error": None,
+        "elapsed_time": 0.0,
+    }
+    errors: List[str] = []
+
+    gateway_proc = None
+    agent_proc = None
+    timeout_seconds = task.timeout_seconds * timeout_multiplier
+    start_time = time.perf_counter()
+
+    try:
+        # 1. Start container
+        start_container(container_id)
+
+        # 2. Inject workspace files
+        if task.workspace_files:
+            setup_workspace(container_id, task.workspace_files, ASSETS_DIR)
+
+        # 3. Onboard OpenClaw
+        onboard_openclaw(container_id)
+
+        # 3.5 Register judge provider if using a separate endpoint
+        if JUDGE_CUSTOM_BASE_URL and JUDGE_CUSTOM_API_KEY:
+            register_custom_provider(
+                container_id,
+                JUDGE_CUSTOM_BASE_URL,
+                JUDGE_CUSTOM_API_KEY,
+                JUDGE_CUSTOM_MODEL_ID or judge_model,
+            )
+
+        # 4. Start gateway
+        gateway_proc = start_gateway(container_id, output_dir / "gateway.log")
+
+        # 5. Set model
+        set_model(container_id, model)
+
+        # 6. Run agent — use docker cp to bypass Windows cmd-line limit
+        fd, tmp_path = tempfile.mkstemp(suffix=".txt")
+        with os.fdopen(fd, "w", encoding="utf-8") as f:
+            f.write(task.prompt)
+        subprocess.run(
+            ["docker", "cp", tmp_path,
+             f"{container_id}:/tmp/agent_prompt.txt"],
+            capture_output=True, text=True, encoding="utf-8",
+        )
+        os.unlink(tmp_path)
+
+        agent_bash = (
+            f"prompt=$(cat /tmp/agent_prompt.txt) && "
+            f"cd {TMP_WORKSPACE} && "
+            f"openclaw agent --session-id chat "
+            f"--timeout {int(timeout_seconds)} --message \"$prompt\""
+        )
+
+        log_file = (output_dir / "agent.log").open("w", encoding="utf-8")
+        agent_proc = subprocess.Popen(
+            ["docker", "exec", container_id, "/bin/bash", "-c", agent_bash],
+            stdout=log_file,
+            stderr=subprocess.STDOUT,
+            encoding="utf-8",
+        )
+        agent_proc._log_file = log_file  # type: ignore[attr-defined]
+
+        logger.info("[%s] Agent running (timeout=%ds)", container_id, int(timeout_seconds))
+        try:
+            agent_proc.wait(timeout=timeout_seconds)
+            elapsed = time.perf_counter() - start_time
+            logger.info("[%s] Agent finished (%.1fs, exit=%s)",
+                        container_id, elapsed, agent_proc.returncode)
+        except subprocess.TimeoutExpired:
+            elapsed = timeout_seconds
+            logger.info("[%s] Agent timed out", container_id)
+            agent_proc.kill()
+            agent_proc.wait()
+
+        # 7. Collect transcript and compute usage
+        transcript_path = collect_transcript(container_id, output_dir)
+        usage = extract_usage_from_jsonl(transcript_path)
+        usage["elapsed_time"] = round(time.perf_counter() - start_time, 2)
+        result["usage"] = usage
+        (output_dir / "usage.json").write_text(
+            json.dumps(usage, indent=2, ensure_ascii=False), encoding="utf-8"
+        )
+
+        # 7.5 Check for API errors in transcript BEFORE grading
+        transcript_err = detect_transcript_errors(transcript_path)
+        if transcript_err:
+            logger.warning("[%s] %s — skipping judge", container_id, transcript_err)
+            errors.append(transcript_err)
+        else:
+            # 8. LLM Judge grading
+            grade = None
+            try:
+                grade = grade_task(
+                    container_id=container_id,
+                    task_id=task.task_id,
+                    task_prompt=task.prompt,
+                    expected_behavior=task.expected_behavior,
+                    grading_criteria=task.grading_criteria,
+                    llm_judge_rubric=task.llm_judge_rubric,
+                    agent_transcript_path=transcript_path,
+                    output_dir=output_dir,
+                    judge_model=judge_model,
+                )
+                result["grade"] = grade.to_dict()
+                result["scores"] = grade.breakdown
+                score_path = output_dir / "score.json"
+                score_path.write_text(
+                    json.dumps(grade.to_dict(), indent=2, ensure_ascii=False), encoding="utf-8"
+                )
+                logger.info("[%s] Grade: %.2f/%.2f", container_id, grade.score, grade.max_score)
+            except Exception as exc:
+                logger.error("[%s] Grading failed: %s", container_id, exc)
+                errors.append(f"Grading failed: {exc}")
+
+            # 8.5 Process grading — only if step 8 succeeded
+            if not errors:
+                try:
+                    gold_data = _load_gold_process_data(task)
+                    if gold_data is not None:
+                        outcome_score = grade.score if grade else 0.0
+                        process_scores = _run_process_grading(
+                            container_id=container_id,
+                            task_id=task.task_id,
+                            transcript_path=transcript_path,
+                            gold_data=gold_data,
+                            outcome_score=outcome_score,
+                            output_dir=output_dir,
+                        )
+                        result["process_grade"] = process_scores
+                    else:
+                        logger.info("[%s] No process annotations, skipping process grading", container_id)
+                except Exception as exc:
+                    logger.error("[%s] Process grading failed: %s", container_id, exc)
+                    errors.append(f"Process grading failed: {exc}")
+
+        # 9. Collect task output (always attempt — preserve artifacts for debugging)
+        try:
+            collect_output(container_id, output_dir)
+        except Exception as exc:
+            logger.warning("[%s] Output collection failed: %s", container_id, exc)
+            errors.append(f"Output collection failed: {exc}")
+
+    except Exception as exc:
+        logger.error("[%s] Execution error: %s", container_id, exc)
+        errors.append(f"Execution error: {exc}")
+
+    finally:
+        result["elapsed_time"] = round(time.perf_counter() - start_time, 2)
+        if errors:
+            result["error"] = "; ".join(errors)
+
+        if gateway_proc is not None:
+            try:
+                gateway_proc.terminate()
+            except Exception:
+                pass
+        for proc in [gateway_proc, agent_proc]:
+            if proc is not None:
+                try:
+                    close_proc_log(proc)
+                except Exception:
+                    pass
+
+        remove_container(container_id)
+        logger.info("[%s] Container cleaned up", container_id)
+
+    return result
+
+
+# ---------------------------------------------------------------------------
+# Summary & reporting
+# ---------------------------------------------------------------------------
+
+def _print_summary(results: List[Dict[str, Any]], model_name: str) -> None:
+    print(f"\n{'#' * 60}")
+    print(f"  Summary Report — {model_name}")
+    print(f"{'#' * 60}")
+
+    scored = 0
+    total_score = 0.0
+    for r in results:
+        grade = r.get("grade")
+        if r.get("error") or not grade:
+            print(f"  X {r['task_id']}: {r.get('error', 'no grade')}")
+            continue
+        scored += 1
+        total_score += grade.get("score", 0.0)
+        pct = grade["score"] / grade["max_score"] * 100 if grade["max_score"] > 0 else 0
+        print(f"  + {r['task_id']}: {grade['score']:.2f}/{grade['max_score']:.2f} ({pct:.0f}%)")
+
+    error_count = sum(1 for r in results if r.get("error"))
+    if scored:
+        avg = total_score / scored
+        print(f"\n  Scored: {scored}/{len(results)}  Average: {avg:.4f}")
+        if error_count:
+            print(f"  Errors: {error_count} task(s) skipped due to API errors")
+    else:
+        print("\n  No tasks scored successfully")
+        if error_count:
+            print(f"  Errors: {error_count} task(s) skipped due to API errors")
+
+    valid = [r for r in results if not r.get("error")]
+    total_out_tok = sum(r.get("usage", {}).get("output_tokens", 0) or 0 for r in valid)
+    total_cost = sum(r.get("usage", {}).get("cost_usd", 0.0) or 0.0 for r in valid)
+    print(f"  Total output tokens: {total_out_tok}  Total cost: ${total_cost:.4f}")
+
+    # Process metrics summary — exclude error tasks
+    proc = [
+        r["process_grade"] for r in results
+        if r.get("process_grade") and not r.get("error")
+    ]
+    if proc:
+        eff_vals = [p["efficiency"]["efficiency"] for p in proc if p.get("efficiency", {}).get("efficiency") is not None]
+        gpr_vals = [p["gpr"]["gpr"] for p in proc if "gpr" in p]
+        tgpr_vals = [p["tgpr"]["tgpr"] for p in proc if "tgpr" in p]
+        # TPE is added to the same TGPRResult dict; older runs may lack the key.
+        tpe_vals = [p["tgpr"]["tpe"] for p in proc if "tgpr" in p and "tpe" in p["tgpr"]]
+        print(f"\n  Process metrics ({len(proc)} tasks):")
+        if eff_vals:
+            print(f"    Avg Efficiency: {sum(eff_vals)/len(eff_vals):.4f}")
+        if gpr_vals:
+            print(f"    Avg GPR:        {sum(gpr_vals)/len(gpr_vals):.4f}")
+        if tgpr_vals:
+            print(f"    Avg TGPR:       {sum(tgpr_vals)/len(tgpr_vals):.4f}")
+        if tpe_vals:
+            print(f"    Avg TPE:        {sum(tpe_vals)/len(tpe_vals):.4f}")
+
+    print("#" * 60)
+
+    # Errored tasks — listed together for easy review
+    errored = [r for r in results if r.get("error")]
+    if errored:
+        print(f"\n{'#' * 60}")
+        print(f"  Errored Tasks ({len(errored)})")
+        print(f"{'#' * 60}")
+        for r in errored:
+            print(f"  X {r['task_id']}: {r.get('error')}")
+        print()
+        print("  -> Run with --resume to retry these tasks")
+        print("#" * 60)
+
+
+def _write_global_summary(
+    results: List[Dict[str, Any]],
+    model: str,
+    model_slug: str,
+    suite: str,
+    runs_per_task: int,
+    task_objects: Optional[List[Task]] = None,
+) -> Path:
+    task_meta_map: Dict[str, Dict[str, Any]] = {}
+    if task_objects:
+        for t in task_objects:
+            task_meta_map[t.task_id] = t.frontmatter
+
+    aggregate: Dict[str, Any] = {
+        "model": model,
+        "benchmark_version": _get_git_version(),
+        "timestamp": time.time(),
+        "suite": suite,
+        "runs_per_task": runs_per_task,
+        "tasks": [],
+    }
+
+    agg_total_tokens = 0
+    agg_total_cost = 0.0
+    agg_total_requests = 0
+
+    for r in results:
+        grade = r.get("grade")
+        usage = r.get("usage", {})
+        score = grade.get("score", 0.0) if grade else 0.0
+        elapsed = r.get("elapsed_time", 0.0)
+
+        # Only count tokens/cost/requests from non-error tasks
+        if not r.get("error"):
+            agg_total_tokens += usage.get("total_tokens", 0) or 0
+            agg_total_cost += usage.get("cost_usd", 0.0) or 0.0
+            agg_total_requests += usage.get("request_count", 0) or 0
+
+        entry: Dict[str, Any] = {
+            "task_id": r["task_id"],
+            "frontmatter": task_meta_map.get(r["task_id"], {}),
+            "grade": grade,
+            "process_grade": r.get("process_grade"),
+            "grading": {"mean": score},
+            "usage": usage,
+            "error": r.get("error"),
+            "elapsed_time": elapsed,
+            "execution_time": elapsed,
+        }
+        aggregate["tasks"].append(entry)
+
+    # Compute global score — exclude error tasks
+    valid_results = [r for r in results if not r.get("error")]
+    error_count = sum(1 for r in results if r.get("error"))
+    grades = [r["grade"] for r in valid_results if r.get("grade")]
+    if grades:
+        total_score = sum(g["score"] for g in grades)
+        max_score = sum(g["max_score"] for g in grades)
+        overall = round(total_score / max_score, 4) if max_score > 0 else 0.0
+        aggregate["overall_score"] = overall
+        aggregate["total_score"] = total_score
+        aggregate["max_score"] = max_score
+    else:
+        overall = 0.0
+        aggregate["overall_score"] = 0
+    aggregate["error_count"] = error_count
+    aggregate["evaluated_count"] = len(results) - error_count
+
+    aggregate["efficiency"] = {
+        "total_tokens": agg_total_tokens,
+        "total_cost_usd": round(agg_total_cost, 6),
+        "total_requests": agg_total_requests,
+    }
+
+    # Aggregate process metrics — exclude error tasks
+    process_results = [
+        r["process_grade"] for r in results
+        if r.get("process_grade") is not None and not r.get("error")
+    ]
+    if process_results:
+        eff_values = [
+            p["efficiency"]["efficiency"]
+            for p in process_results
+            if p.get("efficiency", {}).get("efficiency") is not None
+        ]
+        gpr_values = [p["gpr"]["gpr"] for p in process_results if "gpr" in p]
+        tgpr_values = [p["tgpr"]["tgpr"] for p in process_results if "tgpr" in p]
+        aggregate["process_metrics"] = {
+            "tasks_with_process": len(process_results),
+            "avg_efficiency": round(sum(eff_values) / len(eff_values), 4) if eff_values else None,
+            "avg_gpr": round(sum(gpr_values) / len(gpr_values), 4) if gpr_values else None,
+            "avg_tgpr": round(sum(tgpr_values) / len(tgpr_values), 4) if tgpr_values else None,
+        }
+
+    summary_path = OUTPUT_DIR / f"summary_{model_slug}.json"
+    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
+    summary_path.write_text(
+        json.dumps(aggregate, indent=2, ensure_ascii=False), encoding="utf-8"
+    )
+    return summary_path
+
+
+# ---------------------------------------------------------------------------
+# CLI & main
+# ---------------------------------------------------------------------------
+
+def main() -> None:
+    parser = argparse.ArgumentParser(
+        description="DataClaw — per-task container benchmark orchestrator",
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+        epilog="""
+Examples:
+  # Run all tasks
+  python dataclaw/eval/run_batch.py --model openrouter/anthropic/claude-sonnet-4.6
+
+  # Run specific tasks
+  python dataclaw/eval/run_batch.py --model ... --suite task_001,task_002
+
+  # Run with parallelism
+  python dataclaw/eval/run_batch.py --model ... --parallel 4
+
+  # Run a single task file
+  python dataclaw/eval/run_batch.py --task tasks/task_001_xxx.md
+
+  # Resume an interrupted run
+  python dataclaw/eval/run_batch.py --model ... --suite all --resume
+""",
+    )
+
+    mode = parser.add_mutually_exclusive_group()
+    mode.add_argument("--task", "-t", help="Path to a single task.md file")
+    mode.add_argument(
+        "--suite", "-s",
+        default="all",
+        help='Tasks to run: "all" or comma-separated task IDs (default: all)',
+    )
+
+    parser.add_argument(
+        "--model", "-m",
+        default=DEFAULT_MODEL,
+        help=f"Model identifier (default: {DEFAULT_MODEL or 'from .env'})",
+    )
+    parser.add_argument(
+        "--judge",
+        default=DEFAULT_JUDGE_MODEL,
+        help=f"Judge model identifier (default: {DEFAULT_JUDGE_MODEL})",
+    )
+    parser.add_argument(
+        "--parallel", "-p",
+        type=int,
+        default=DEFAULT_PARALLEL,
+        metavar="N",
+        help=f"Number of parallel containers (default: {DEFAULT_PARALLEL})",
+    )
+    parser.add_argument(
+        "--timeout-multiplier",
+        type=float,
+        default=TIMEOUT_MULTIPLIER,
+        help=f"Scale all task timeouts (default: {TIMEOUT_MULTIPLIER})",
+    )
+    parser.add_argument(
+        "--runs",
+        type=int,
+        default=BENCHMARK_RUNS,
+        help=f"Number of runs per task (default: {BENCHMARK_RUNS})",
+    )
+    parser.add_argument(
+        "--resume",
+        action="store_true",
+        help="Resume from last interrupted run (skip completed tasks)",
+    )
+    parser.add_argument(
+        "--verbose", "-v",
+        action="store_true",
+        help="Enable verbose logging",
+    )
+
+    args = parser.parse_args()
+
+    if not args.model:
+        logger.error("--model is required (or set DEFAULT_MODEL in .env)")
+        sys.exit(1)
+
+    # Validate model
+    if not _validate_openrouter_model(args.model):
+        sys.exit(1)
+
+    logger.info("DataClaw — per-task container benchmark")
+    logger.info("Model: %s | Judge: %s | Parallel: %d", args.model, args.judge, args.parallel)
+
+    # Load tasks
+    if args.task:
+        task_file = Path(args.task)
+        if not task_file.exists():
+            logger.error("File not found: %s", task_file)
+            sys.exit(1)
+        loader = TaskLoader(task_file.parent)
+        task = loader.load_task(task_file)
+        tasks = [task]
+        suite_label = task.task_id
+    else:
+        if not TASKS_DIR.exists():
+            logger.error("Tasks directory not found: %s", TASKS_DIR)
+            sys.exit(1)
+        loader = TaskLoader(TASKS_DIR)
+        all_tasks = loader.load_all_tasks()
+        if args.suite == "all":
+            tasks = all_tasks
+            suite_label = "all"
+        else:
+            requested = {tid.strip() for tid in args.suite.split(",") if tid.strip()}
+            known = {t.task_id for t in all_tasks}
+            unknown = requested - known
+            if unknown:
+                logger.error("Unknown task IDs: %s", ", ".join(sorted(unknown)))
+                sys.exit(1)
+            tasks = [t for t in all_tasks if t.task_id in requested]
+            suite_label = args.suite
+
+    if not tasks:
+        logger.error("No tasks to run")
+        sys.exit(1)
+
+    logger.info("Tasks: %d | Runs per task: %d", len(tasks), args.runs)
+
+    # Execute
+    all_results: List[Dict[str, Any]] = []
+    model_slug = _slugify_model(args.model)
+    prog_path = _progress_path(model_slug)
+
+    # --resume: load previous progress and validate parameters
+    completed_keys: set = set()
+    progress_entries: List[Dict[str, Any]] = []
+
+    if args.resume:
+        prog_data = _load_progress(prog_path)
+        if prog_data is not None:
+            mismatches = []
+            if prog_data.get("model") != args.model:
+                mismatches.append(
+                    f"model (progress={prog_data.get('model')}, current={args.model})"
+                )
+            if prog_data.get("suite") != suite_label:
+                mismatches.append(
+                    f"suite (progress={prog_data.get('suite')}, current={suite_label})"
+                )
+            if prog_data.get("runs") != args.runs:
+                mismatches.append(
+                    f"runs (progress={prog_data.get('runs')}, current={args.runs})"
+                )
+            if mismatches:
+                logger.error(
+                    "Resume failed: parameter mismatch — %s. "
+                    "Use the same parameters as the original run, or remove %s to start fresh.",
+                    "; ".join(mismatches), prog_path,
+                )
+                sys.exit(1)
+
+            for entry in prog_data.get("completed", []):
+                completed_keys.add((entry["task_id"], entry["run_index"]))
+                all_results.append(entry["result"])
+                progress_entries.append(entry)
+            logger.info("Resuming: %d tasks already completed", len(completed_keys))
+        else:
+            logger.warning(
+                "No progress file found for model '%s', starting from scratch",
+                args.model,
+            )
+
+    for run_index in range(args.runs):
+        if args.runs > 1:
+            logger.info("=== Run %d/%d ===", run_index + 1, args.runs)
+
+        pending_tasks = [
+            t for t in tasks if (t.task_id, run_index) not in completed_keys
+        ]
+        if not pending_tasks:
+            logger.info("All tasks in run %d already completed, skipping", run_index + 1)
+            continue
+        if completed_keys:
+            skipped = len(tasks) - len(pending_tasks)
+            if skipped:
+                logger.info(
+                    "Skipping %d completed task(s), running %d remaining",
+                    skipped, len(pending_tasks),
+                )
+
+        if args.parallel <= 1:
+            for i, task in enumerate(pending_tasks, 1):
+                logger.info("--- Task %d/%d: %s ---", i, len(pending_tasks), task.task_id)
+                result = run_single_task(
+                    task, args.model, args.judge, args.timeout_multiplier,
+                )
+                all_results.append(result)
+                if result.get("error"):
+                    logger.warning(
+                        "[%s] Skipping progress save (error: %s) — will retry on resume",
+                        task.task_id, result["error"],
+                    )
+                else:
+                    with _progress_lock:
+                        progress_entries.append({
+                            "task_id": task.task_id,
+                            "run_index": run_index,
+                            "result": result,
+                        })
+                        _save_progress(
+                            prog_path, args.model, suite_label, args.runs,
+                            progress_entries,
+                        )
+        else:
+            with ThreadPoolExecutor(max_workers=args.parallel) as pool:
+                futures = {
+                    pool.submit(
+                        run_single_task,
+                        task, args.model, args.judge, args.timeout_multiplier,
+                    ): (task.task_id, run_index)
+                    for task in pending_tasks
+                }
+                for future in as_completed(futures):
+                    tid, ridx = futures[future]
+                    try:
+                        result = future.result()
+                    except Exception as exc:
+                        logger.error("[%s] Thread exception: %s", tid, exc)
+                        result = {
+                            "task_id": tid,
+                            "model": args.model,
+                            "scores": {},
+                            "grade": None,
+                            "process_grade": None,
+                            "usage": {},
+                            "error": str(exc),
+                            "elapsed_time": 0.0,
+                        }
+                    all_results.append(result)
+                    if result.get("error"):
+                        logger.warning(
+                            "[%s] Skipping progress save (error: %s) — will retry on resume",
+                            tid, result["error"],
+                        )
+                    else:
+                        with _progress_lock:
+                            progress_entries.append({
+                                "task_id": tid,
+                                "run_index": ridx,
+                                "result": result,
+                            })
+                            _save_progress(
+                                prog_path, args.model, suite_label, args.runs,
+                                progress_entries,
+                            )
+
+    # Summary
+    _print_summary(all_results, args.model)
+    summary_path = _write_global_summary(
+        all_results, args.model, model_slug, suite_label, args.runs,
+        task_objects=tasks,
+    )
+    logger.info("Summary written to: %s", summary_path)
+
+    # Clean up progress file only if all tasks succeeded (no errors).
+    # If there are errors, keep progress so --resume can retry just the
+    # errored tasks without re-running the ones that already succeeded.
+    error_count = sum(1 for r in all_results if r.get("error"))
+    if error_count == 0:
+        try:
+            prog_path.unlink(missing_ok=True)
+        except OSError:
+            pass
+    else:
+        logger.info(
+            "%d task(s) had errors — progress file kept for --resume",
+            error_count,
+        )
+
+
+if __name__ == "__main__":
+    main()
diff --git a/dataclaw/lib_tasks.py b/dataclaw/lib_tasks.py
new file mode 100644
index 0000000000000000000000000000000000000000..c5632e5c9b4a7a0d12f6a6f62b2b93cac97eb9ac
--- /dev/null
+++ b/dataclaw/lib_tasks.py
@@ -0,0 +1,173 @@
+"""
+DataClaw task library — load and parse benchmark tasks from markdown + YAML.
+"""
+
+import logging
+import re
+from pathlib import Path
+from typing import Dict, List, Optional, Any
+
+import yaml
+
+
+logger = logging.getLogger(__name__)
+
+
+class Task:
+    """Represents a single benchmark task."""
+    
+    def __init__(
+        self,
+        task_id: str,
+        name: str,
+        category: str,
+        grading_type: str,
+        timeout_seconds: int,
+        workspace_files: List[Dict[str, str]],
+        prompt: str,
+        expected_behavior: str,
+        grading_criteria: List[str],
+        llm_judge_rubric: Optional[str] = None,
+        file_path: Optional[Path] = None,
+        frontmatter: Optional[Dict[str, Any]] = None,
+    ):
+        self.task_id = task_id
+        self.name = name
+        self.category = category
+        self.grading_type = grading_type
+        self.timeout_seconds = timeout_seconds
+        self.workspace_files = workspace_files
+        self.prompt = prompt
+        self.expected_behavior = expected_behavior
+        self.grading_criteria = grading_criteria
+        self.llm_judge_rubric = llm_judge_rubric
+        self.file_path = file_path
+        self.frontmatter = frontmatter or {}
+    
+    def __repr__(self) -> str:
+        return f"Task(id={self.task_id}, name={self.name}, category={self.category})"
+    
+    def to_dict(self) -> Dict[str, Any]:
+        """Convert task to dictionary representation."""
+        return {
+            'task_id': self.task_id,
+            'name': self.name,
+            'category': self.category,
+            'grading_type': self.grading_type,
+            'timeout_seconds': self.timeout_seconds,
+            'workspace_files': self.workspace_files,
+            'prompt': self.prompt,
+            'expected_behavior': self.expected_behavior,
+            'grading_criteria': self.grading_criteria,
+            'has_llm_judge_rubric': self.llm_judge_rubric is not None,
+            'frontmatter': self.frontmatter,
+        }
+
+
+class TaskLoader:
+    """Loads and parses task files from the tasks directory."""
+    
+    def __init__(self, tasks_dir: Path):
+        self.tasks_dir = tasks_dir
+        logger.info("[tasks] loader · %s", tasks_dir)
+    
+    def load_all_tasks(self) -> List[Task]:
+        """Load all task files from the tasks directory."""
+        tasks = []
+        task_files = sorted(self.tasks_dir.glob("task_*.md"))
+        
+        logger.info("[tasks] found %d file(s)", len(task_files))
+        
+        for task_file in task_files:
+            try:
+                task = self.load_task(task_file)
+                tasks.append(task)
+                logger.info("[tasks] ok · %s", task.task_id)
+            except Exception as e:
+                logger.error("[tasks] failed · %s · %s", task_file, e, exc_info=True)
+        
+        logger.info("[tasks] loaded %d task(s)", len(tasks))
+        return tasks
+    
+    def load_task(self, task_file: Path) -> Task:
+        """Load and parse a single task file."""
+        logger.debug(f"Loading task from: {task_file}")
+        
+        content = task_file.read_text(encoding='utf-8')
+        
+        # Extract YAML frontmatter
+        frontmatter_match = re.match(r'^---\s*\n(.*?)\n---\s*\n(.*)$', content, re.DOTALL)
+        if not frontmatter_match:
+            raise ValueError(f"No YAML frontmatter found in {task_file}")
+        
+        frontmatter_text = frontmatter_match.group(1)
+        body_text = frontmatter_match.group(2)
+        
+        # Parse YAML frontmatter
+        try:
+            metadata = yaml.safe_load(frontmatter_text)
+        except yaml.YAMLError as e:
+            raise ValueError(f"Invalid YAML frontmatter in {task_file}: {e}")
+        
+        # Extract sections from body
+        sections = self._parse_sections(body_text)
+        
+        # Extract grading criteria
+        grading_criteria = self._extract_grading_criteria(
+            sections.get('Grading Criteria', '')
+        )
+        
+        # Create Task object
+        task = Task(
+            task_id=metadata.get('id', ''),
+            name=metadata.get('name', ''),
+            category=metadata.get('category', ''),
+            grading_type=metadata.get('grading_type', 'llm_judge'),
+            timeout_seconds=metadata.get('timeout_seconds', 120),
+            workspace_files=metadata.get('workspace_files', []),
+            prompt=sections.get('Prompt', '').strip(),
+            expected_behavior=sections.get('Expected Behavior', '').strip(),
+            grading_criteria=grading_criteria,
+            llm_judge_rubric=sections.get('LLM Judge Rubric', None),
+            file_path=task_file,
+            frontmatter=metadata,
+        )
+        
+        return task
+    
+    def _parse_sections(self, body: str) -> Dict[str, str]:
+        """Parse markdown sections from task body."""
+        sections = {}
+        current_section = None
+        current_content = []
+        
+        for line in body.split('\n'):
+            # Check for section headers (## Header)
+            header_match = re.match(r'^##\s+(.+)$', line)
+            if header_match:
+                # Save previous section
+                if current_section:
+                    sections[current_section] = '\n'.join(current_content).strip()
+                
+                # Start new section
+                current_section = header_match.group(1)
+                current_content = []
+            else:
+                if current_section:
+                    current_content.append(line)
+        
+        # Save last section
+        if current_section:
+            sections[current_section] = '\n'.join(current_content).strip()
+        
+        return sections
+    
+    def _extract_grading_criteria(self, criteria_text: str) -> List[str]:
+        """Extract grading criteria from checklist format."""
+        criteria = []
+        for line in criteria_text.split('\n'):
+            # Match checklist items: - [ ] or - [x]
+            match = re.match(r'^-\s+\[[ x]\]\s+(.+)$', line.strip())
+            if match:
+                criteria.append(match.group(1))
+        return criteria
diff --git a/dataclaw/utils/__init__.py b/dataclaw/utils/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/dataclaw/utils/__pycache__/__init__.cpython-312.pyc b/dataclaw/utils/__pycache__/__init__.cpython-312.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..0427d6fc27e2c260dee7965bc9043ce9928e214f
Binary files /dev/null and b/dataclaw/utils/__pycache__/__init__.cpython-312.pyc differ
diff --git a/dataclaw/utils/__pycache__/docker_utils.cpython-312.pyc b/dataclaw/utils/__pycache__/docker_utils.cpython-312.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..ccd5f84f775bcc840289d578a239710166022207
Binary files /dev/null and b/dataclaw/utils/__pycache__/docker_utils.cpython-312.pyc differ
diff --git a/dataclaw/utils/__pycache__/grading.cpython-312.pyc b/dataclaw/utils/__pycache__/grading.cpython-312.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..adea0cd6428fad419ac9bd876b517a94c22b69d5
Binary files /dev/null and b/dataclaw/utils/__pycache__/grading.cpython-312.pyc differ
diff --git a/dataclaw/utils/__pycache__/process_grading.cpython-312.pyc b/dataclaw/utils/__pycache__/process_grading.cpython-312.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..30da89c02e947400a881a52197b1ab4103644746
Binary files /dev/null and b/dataclaw/utils/__pycache__/process_grading.cpython-312.pyc differ
diff --git a/dataclaw/utils/docker_utils.py b/dataclaw/utils/docker_utils.py
new file mode 100644
index 0000000000000000000000000000000000000000..2ed4f33f678ee30fe5e5aed55efec0f2d4f761fd
--- /dev/null
+++ b/dataclaw/utils/docker_utils.py
@@ -0,0 +1,648 @@
+"""
+Container lifecycle management for per-task Docker isolation.
+
+Each benchmark task runs in its own Docker container. The host-side orchestrator
+(dataclaw/eval/run_batch.py) uses these helpers to start, configure, drive, and tear down
+containers via the Docker CLI.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+import os
+import shutil
+import subprocess
+import tempfile
+import time
+from pathlib import Path
+from typing import Any, Dict, List, Optional
+
+try:
+    from dotenv import load_dotenv
+    load_dotenv()
+except ImportError:
+    pass
+logger = logging.getLogger(__name__)
+
+DOCKER_IMAGE = os.environ.get("DOCKER_IMAGE", "dataclaw:0.1.0")
+TMP_WORKSPACE = os.environ.get("TMP_WORKSPACE", "/tmp_workspace")
+GATEWAY_PORT = int(os.environ.get("GATEWAY_PORT", "3333"))
+DEFAULT_MODEL = os.environ.get("DEFAULT_MODEL", "")
+
+OPENROUTER_API_KEY = os.environ.get("OPENROUTER_API_KEY", "")
+OPENCLAW_CUSTOM_BASE_URL = os.environ.get("OPENCLAW_CUSTOM_BASE_URL", "")
+OPENCLAW_CUSTOM_API_KEY = os.environ.get("OPENCLAW_CUSTOM_API_KEY", "")
+OPENCLAW_CUSTOM_MODEL_ID = os.environ.get("OPENCLAW_CUSTOM_MODEL_ID", "")
+JUDGE_CUSTOM_BASE_URL = os.environ.get("JUDGE_CUSTOM_BASE_URL", "")
+JUDGE_CUSTOM_API_KEY = os.environ.get("JUDGE_CUSTOM_API_KEY", "")
+JUDGE_CUSTOM_MODEL_ID = os.environ.get("JUDGE_CUSTOM_MODEL_ID", "")
+OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY", "")
+ANTHROPIC_API_KEY = os.environ.get("ANTHROPIC_API_KEY", "")
+GEMINI_API_KEY = os.environ.get("GEMINI_API_KEY", "")
+BRAVE_API_KEY = os.environ.get("BRAVE_API_KEY", "")
+
+OPENCLAW_MODEL_CONTEXT_WINDOW = int(os.environ.get("OPENCLAW_MODEL_CONTEXT_WINDOW", "128000"))
+OPENCLAW_MODEL_MAX_TOKENS = int(os.environ.get("OPENCLAW_MODEL_MAX_TOKENS", "16384"))
+OPENCLAW_MODEL_COST_INPUT = float(os.environ.get("OPENCLAW_MODEL_COST_INPUT", "0"))
+OPENCLAW_MODEL_COST_OUTPUT = float(os.environ.get("OPENCLAW_MODEL_COST_OUTPUT", "0"))
+OPENCLAW_MODEL_COST_CACHE_READ = float(os.environ.get("OPENCLAW_MODEL_COST_CACHE_READ", "0"))
+OPENCLAW_MODEL_COST_CACHE_WRITE = float(os.environ.get("OPENCLAW_MODEL_COST_CACHE_WRITE", "0"))
+
+JUDGE_MODEL_CONTEXT_WINDOW = int(os.environ.get("JUDGE_MODEL_CONTEXT_WINDOW", "128000"))
+JUDGE_MODEL_MAX_TOKENS = int(os.environ.get("JUDGE_MODEL_MAX_TOKENS", "16384"))
+JUDGE_MODEL_COST_INPUT = float(os.environ.get("JUDGE_MODEL_COST_INPUT", "0"))
+JUDGE_MODEL_COST_OUTPUT = float(os.environ.get("JUDGE_MODEL_COST_OUTPUT", "0"))
+JUDGE_MODEL_COST_CACHE_READ = float(os.environ.get("JUDGE_MODEL_COST_CACHE_READ", "0"))
+JUDGE_MODEL_COST_CACHE_WRITE = float(os.environ.get("JUDGE_MODEL_COST_CACHE_WRITE", "0"))
+
+
+def remove_container(name: str) -> None:
+    subprocess.run(["docker", "rm", "-f", name], capture_output=True, encoding="utf-8")
+
+
+def start_container(task_id: str, extra_env: Optional[Dict[str, str]] = None) -> None:
+    """Start a detached container for one task."""
+    env_args: List[str] = []
+
+    proxy_http = os.environ.get("HTTP_PROXY_INNER", "")
+    proxy_https = os.environ.get("HTTPS_PROXY_INNER", "")
+    if proxy_http or proxy_https:
+        env_args += [
+            "-e", f"http_proxy={proxy_http}",
+            "-e", f"https_proxy={proxy_https}",
+            "-e", f"HTTP_PROXY={proxy_http}",
+            "-e", f"HTTPS_PROXY={proxy_https}",
+        ]
+
+    for key, value in (extra_env or {}).items():
+        env_args += ["-e", f"{key}={value}"]
+
+    cmd = [
+        "docker", "run", "-d",
+        "--name", task_id,
+        *env_args,
+        DOCKER_IMAGE,
+        "/bin/bash", "-c", "tail -f /dev/null",
+    ]
+    logger.info("[%s] Starting container (image=%s)", task_id, DOCKER_IMAGE)
+    r = subprocess.run(cmd, capture_output=True, text=True, encoding="utf-8")
+    if r.returncode != 0:
+        raise RuntimeError(f"Container startup failed:\n{r.stderr}")
+    logger.info("[%s] Container started: %s", task_id, r.stdout.strip()[:12])
+
+
+def setup_workspace(
+    task_id: str,
+    workspace_files: List[Dict[str, str]],
+    assets_dir: Path,
+) -> None:
+    """Copy task workspace files into the container's /tmp_workspace."""
+    subprocess.run(
+        ["docker", "exec", task_id, "mkdir", "-p", TMP_WORKSPACE],
+        capture_output=True, text=True, encoding="utf-8",
+    )
+
+    with tempfile.TemporaryDirectory() as staging:
+        staging_path = Path(staging)
+        for file_spec in workspace_files:
+            source = assets_dir / file_spec["source"]
+            dest_rel = file_spec["dest"]
+            dest_local = staging_path / dest_rel
+            dest_local.parent.mkdir(parents=True, exist_ok=True)
+            if not source.exists():
+                logger.error("[%s] Workspace file not found: %s", task_id, source)
+                raise FileNotFoundError(f"Workspace file not found: {source}")
+            shutil.copy2(str(source), str(dest_local))
+
+        r = subprocess.run(
+            ["docker", "cp", f"{staging}/.", f"{task_id}:{TMP_WORKSPACE}/"],
+            capture_output=True, text=True, encoding="utf-8",
+        )
+        if r.returncode != 0:
+            raise RuntimeError(f"Workspace copy failed:\n{r.stderr}")
+    logger.info("[%s] Workspace files injected (%d files)", task_id, len(workspace_files))
+
+    # Symlink OpenClaw workspace to TMP_WORKSPACE so tools can access files
+    subprocess.run(
+        ["docker", "exec", task_id, "/bin/bash", "-c",
+         f"rm -rf /root/.openclaw/workspace && ln -s {TMP_WORKSPACE} /root/.openclaw/workspace"],
+        capture_output=True, text=True, encoding="utf-8",
+    )
+
+
+def onboard_openclaw(task_id: str) -> None:
+    """Run openclaw onboard inside the container with configured auth."""
+    onboard_args = [
+        "--non-interactive",
+        "--accept-risk",
+        "--skip-health",
+        "--workspace", "/root/.openclaw/workspace",
+        "--gateway-bind", "loopback",
+        "--gateway-port", str(GATEWAY_PORT),
+    ]
+
+    if OPENCLAW_CUSTOM_BASE_URL:
+        if not OPENCLAW_CUSTOM_API_KEY:
+            raise ValueError(
+                "OPENCLAW_CUSTOM_API_KEY is required when OPENCLAW_CUSTOM_BASE_URL is set"
+            )
+        onboard_args += [
+            "--auth-choice", "custom-api-key",
+            "--custom-base-url", OPENCLAW_CUSTOM_BASE_URL,
+            "--custom-api-key", OPENCLAW_CUSTOM_API_KEY,
+            "--custom-model-id", OPENCLAW_CUSTOM_MODEL_ID or DEFAULT_MODEL,
+        ]
+    elif OPENROUTER_API_KEY:
+        onboard_args += ["--openrouter-api-key", OPENROUTER_API_KEY]
+
+    if OPENAI_API_KEY:
+        onboard_args += ["--openai-api-key", OPENAI_API_KEY]
+    if ANTHROPIC_API_KEY:
+        onboard_args += ["--anthropic-api-key", ANTHROPIC_API_KEY]
+    if GEMINI_API_KEY:
+        onboard_args += ["--gemini-api-key", GEMINI_API_KEY]
+
+    cmd = ["docker", "exec", task_id, "openclaw", "onboard", *onboard_args]
+    logger.info("[%s] Running openclaw onboard", task_id)
+    r = subprocess.run(cmd, capture_output=True, text=True, encoding="utf-8")
+    if r.returncode != 0:
+        raise RuntimeError(f"OpenClaw onboard failed:\n{r.stderr}")
+    logger.info("[%s] Onboard complete", task_id)
+
+    _patch_streaming_usage_compat(task_id)
+    _patch_main_model_capabilities(task_id)
+    _patch_brave_web_search(task_id)
+
+
+def _patch_main_model_capabilities(task_id: str) -> None:
+    """Patch contextWindow, maxTokens, and cost for the main (agent) model.
+
+    OpenClaw onboard assigns conservative defaults for unknown custom providers.
+    This overwrites them with values from the OPENCLAW_MODEL_* environment variables.
+    """
+    config = _read_openclaw_config(task_id)
+    if not config:
+        return
+    patched = False
+    for _pname, pinfo in config.get("models", {}).get("providers", {}).items():
+        for model in pinfo.get("models", []):
+            model["contextWindow"] = OPENCLAW_MODEL_CONTEXT_WINDOW
+            model["maxTokens"] = OPENCLAW_MODEL_MAX_TOKENS
+            model["cost"] = {
+                "input": OPENCLAW_MODEL_COST_INPUT,
+                "output": OPENCLAW_MODEL_COST_OUTPUT,
+                "cacheRead": OPENCLAW_MODEL_COST_CACHE_READ,
+                "cacheWrite": OPENCLAW_MODEL_COST_CACHE_WRITE,
+            }
+            patched = True
+    if patched:
+        _write_openclaw_config(task_id, config)
+        logger.info(
+            "[%s] Patched main model capabilities (contextWindow=%d, maxTokens=%d)",
+            task_id, OPENCLAW_MODEL_CONTEXT_WINDOW, OPENCLAW_MODEL_MAX_TOKENS,
+        )
+
+
+def _patch_brave_web_search(task_id: str) -> None:
+    """Enable Brave as ``web_search`` provider when ``BRAVE_API_KEY`` is set.
+
+    Matches OpenClaw canonical config:
+    https://docs.openclaw.ai/tools/brave-search
+    """
+    api_key = (BRAVE_API_KEY or "").strip()
+    if not api_key:
+        return
+    config = _read_openclaw_config(task_id)
+    if not config:
+        logger.warning("[%s] Cannot read openclaw.json; skip Brave web_search patch", task_id)
+        return
+
+    plugins = config.setdefault("plugins", {})
+    entries = plugins.setdefault("entries", {})
+    brave_entry = entries.setdefault("brave", {})
+    brave_entry["enabled"] = True
+    brave_cfg = brave_entry.setdefault("config", {})
+    brave_cfg.setdefault("webSearch", {})["apiKey"] = api_key
+
+    tools = config.setdefault("tools", {})
+    web = tools.setdefault("web", {})
+    search = web.setdefault("search", {})
+    search["provider"] = "brave"
+    search["maxResults"] = int(os.environ.get("BRAVE_WEB_SEARCH_MAX_RESULTS", "5"))
+    search["timeoutSeconds"] = int(os.environ.get("BRAVE_WEB_SEARCH_TIMEOUT_SECONDS", "30"))
+
+    _write_openclaw_config(task_id, config)
+    logger.info("[%s] Patched OpenClaw config: web_search provider=brave", task_id)
+
+
+def _patch_streaming_usage_compat(task_id: str) -> None:
+    """Ensure all custom-provider models have ``compat.supportsUsageInStreaming: true``.
+
+    OpenClaw defaults this flag to ``false`` for unrecognised providers, which
+    prevents ``stream_options: {include_usage: true}`` from being sent in API
+    requests, resulting in zero-value token usage in transcripts.
+    """
+    config = _read_openclaw_config(task_id)
+    if not config:
+        return
+    patched = False
+    for _pname, pinfo in config.get("models", {}).get("providers", {}).items():
+        for model in pinfo.get("models", []):
+            compat = model.get("compat")
+            if compat is None:
+                model["compat"] = {"supportsUsageInStreaming": True}
+                patched = True
+            elif not compat.get("supportsUsageInStreaming"):
+                compat["supportsUsageInStreaming"] = True
+                patched = True
+    if patched:
+        _write_openclaw_config(task_id, config)
+        logger.info("[%s] Patched compat.supportsUsageInStreaming for custom providers", task_id)
+
+
+def start_gateway(task_id: str, log_path: Path) -> subprocess.Popen:
+    """Start the OpenClaw gateway in the background inside the container."""
+    log_path.parent.mkdir(parents=True, exist_ok=True)
+    log_file = log_path.open("w", encoding="utf-8")
+
+    gateway_cmd = f"openclaw gateway run --bind loopback --port {GATEWAY_PORT}"
+    exports: List[str] = []
+    if OPENROUTER_API_KEY:
+        exports.append(f"export OPENROUTER_API_KEY='{OPENROUTER_API_KEY}'")
+    brave = (BRAVE_API_KEY or "").strip()
+    if brave:
+        exports.append(f"export BRAVE_API_KEY='{brave}'")
+    if exports:
+        gateway_cmd = " && ".join(exports) + " && " + gateway_cmd
+
+    proc = subprocess.Popen(
+        ["docker", "exec", task_id, "/bin/bash", "-c", gateway_cmd],
+        stdout=log_file,
+        stderr=subprocess.STDOUT,
+        encoding="utf-8",
+    )
+    proc._log_file = log_file  # type: ignore[attr-defined]
+    logger.info("[%s] Gateway starting (PID=%s)", task_id, proc.pid)
+
+    time.sleep(2)
+    return proc
+
+
+def _read_openclaw_config(task_id: str) -> Dict[str, Any]:
+    """Read and parse openclaw.json from the container."""
+    r = subprocess.run(
+        ["docker", "exec", task_id, "cat", "/root/.openclaw/openclaw.json"],
+        capture_output=True, text=True, encoding="utf-8",
+    )
+    if r.returncode != 0:
+        return {}
+    try:
+        return json.loads(r.stdout)
+    except json.JSONDecodeError:
+        return {}
+
+
+def _write_openclaw_config(task_id: str, config: Dict[str, Any]) -> None:
+    """Write openclaw.json back into the container via docker cp."""
+    with tempfile.NamedTemporaryFile(
+        "w", suffix=".json", delete=False, encoding="utf-8",
+    ) as f:
+        json.dump(config, f, indent=2, ensure_ascii=False)
+        tmp_path = f.name
+    try:
+        subprocess.run(
+            ["docker", "cp", tmp_path, f"{task_id}:/root/.openclaw/openclaw.json"],
+            capture_output=True, text=True, encoding="utf-8", check=True,
+        )
+    finally:
+        os.unlink(tmp_path)
+
+
+def register_custom_provider(
+    task_id: str, base_url: str, api_key: str, model_id: str,
+) -> None:
+    """Register an additional custom model provider in the container."""
+    config = _read_openclaw_config(task_id)
+    if not config:
+        raise RuntimeError("Cannot read openclaw.json from container")
+
+    from urllib.parse import urlparse
+    hostname = urlparse(base_url).hostname or ""
+    slug = f"custom-{hostname.replace('.', '-')}"
+
+    providers = config.setdefault("models", {}).setdefault("providers", {})
+    if slug not in providers:
+        providers[slug] = {
+            "baseUrl": base_url,
+            "apiKey": api_key,
+            "api": "openai-completions",
+            "models": [],
+        }
+    existing_ids = {m["id"] for m in providers[slug].get("models", [])}
+    if model_id not in existing_ids:
+        providers[slug]["models"].append({
+            "id": model_id,
+            "name": f"{model_id} (Custom Provider)",
+            "reasoning": False,
+            "input": ["text"],
+            "cost": {
+                "input": JUDGE_MODEL_COST_INPUT,
+                "output": JUDGE_MODEL_COST_OUTPUT,
+                "cacheRead": JUDGE_MODEL_COST_CACHE_READ,
+                "cacheWrite": JUDGE_MODEL_COST_CACHE_WRITE,
+            },
+            "compat": {"supportsUsageInStreaming": True},
+            "contextWindow": JUDGE_MODEL_CONTEXT_WINDOW,
+            "maxTokens": JUDGE_MODEL_MAX_TOKENS,
+        })
+
+    _write_openclaw_config(task_id, config)
+    logger.info("[%s] Registered custom provider: %s/%s", task_id, slug, model_id)
+
+
+def resolve_qualified_model(task_id: str, model: str) -> str:
+    """Resolve a bare model name to its fully-qualified provider/model form
+    by reading the live openclaw.json inside the container."""
+    if "/" in model:
+        return model
+    config = _read_openclaw_config(task_id)
+    if not config:
+        logger.warning("[%s] Cannot read openclaw.json, using bare model name: %s",
+                       task_id, model)
+        return model
+    providers = config.get("models", {}).get("providers", {})
+    for provider_name, provider_info in providers.items():
+        for m in provider_info.get("models", []):
+            if m.get("id") == model:
+                return f"{provider_name}/{model}"
+    return model
+
+
+def set_model(task_id: str, model: str) -> None:
+    """Set the active model inside the container."""
+    qualified = resolve_qualified_model(task_id, model)
+    r = subprocess.run(
+        ["docker", "exec", task_id, "/bin/bash", "-c",
+         f"openclaw models set '{qualified}'"],
+        capture_output=True, text=True, encoding="utf-8",
+    )
+    if r.returncode != 0:
+        raise RuntimeError(f"Model setup failed:\n{r.stderr}")
+    logger.info("[%s] Model set: %s", task_id, qualified)
+
+
+def create_agent(task_id: str, agent_id: str, model: str) -> None:
+    """Create an OpenClaw agent inside the container."""
+    r = subprocess.run(
+        ["docker", "exec", task_id, "openclaw", "agents", "add", agent_id,
+         "--model", model, "--non-interactive",
+         "--workspace", "/root/.openclaw/workspace"],
+        capture_output=True, text=True, encoding="utf-8",
+    )
+    if r.returncode != 0:
+        logger.warning("[%s] Agent creation returned %s: %s", task_id, r.returncode, r.stderr)
+
+
+def run_agent_message(
+    task_id: str,
+    message: str,
+    timeout_seconds: float,
+    log_path: Path,
+    agent_id: str = "main",
+) -> subprocess.Popen:
+    """Send a message to an agent inside the container (background).
+
+    Uses docker cp + file read to bypass Windows command-line length limits.
+    """
+    log_path.parent.mkdir(parents=True, exist_ok=True)
+    log_file = log_path.open("w", encoding="utf-8")
+
+    fd, tmp_path = tempfile.mkstemp(suffix=".txt")
+    with os.fdopen(fd, "w", encoding="utf-8") as f:
+        f.write(message)
+
+    subprocess.run(
+        ["docker", "cp", tmp_path,
+         f"{task_id}:/tmp/agent_prompt.txt"],
+        capture_output=True, text=True, encoding="utf-8",
+    )
+    os.unlink(tmp_path)
+
+    bash_cmd = (
+        f"prompt=$(cat /tmp/agent_prompt.txt) && "
+        f"cd {TMP_WORKSPACE} && "
+        f"openclaw agent --agent {agent_id} --session-id chat "
+        f"--timeout {int(timeout_seconds)} --message \"$prompt\""
+    )
+
+    proc = subprocess.Popen(
+        ["docker", "exec", task_id, "/bin/bash", "-c", bash_cmd],
+        stdout=log_file,
+        stderr=subprocess.STDOUT,
+        encoding="utf-8",
+    )
+    proc._log_file = log_file  # type: ignore[attr-defined]
+    logger.info("[%s] Agent message sent (PID=%s, timeout=%ds)", task_id, proc.pid, int(timeout_seconds))
+    return proc
+
+
+def run_judge_message(
+    task_id: str,
+    message: str,
+    timeout_seconds: float = 180,
+    judge_model: Optional[str] = None,
+) -> Dict[str, Any]:
+    """Run the LLM judge inside the container synchronously. Returns transcript entries."""
+    if judge_model:
+        subprocess.run(
+            ["docker", "exec", task_id, "openclaw", "agents", "add", "judge",
+             "--model", judge_model, "--non-interactive",
+             "--workspace", "/root/.openclaw/workspace"],
+            capture_output=True, text=True, encoding="utf-8",
+        )
+
+    fd, tmp_path = tempfile.mkstemp(suffix=".txt")
+    try:
+        with os.fdopen(fd, "w", encoding="utf-8") as f:
+            f.write(message)
+
+        cp_result = subprocess.run(
+            ["docker", "cp", tmp_path,
+             f"{task_id}:/tmp/judge_prompt.txt"],
+            capture_output=True, text=True, encoding="utf-8",
+        )
+        if cp_result.returncode != 0:
+            return {
+                "stdout": "",
+                "stderr": f"docker cp failed: {cp_result.stderr}",
+                "exit_code": -1,
+                "timed_out": False,
+            }
+
+        bash_cmd = (
+            f"prompt=$(cat /tmp/judge_prompt.txt) && "
+            f"cd {TMP_WORKSPACE} && "
+            f"openclaw agent --agent judge --session-id judge_chat "
+            f"--message \"$prompt\""
+        )
+
+        r = subprocess.run(
+            ["docker", "exec", task_id, "/bin/bash", "-c", bash_cmd],
+            capture_output=True, text=True, encoding="utf-8",
+            timeout=timeout_seconds,
+        )
+        return {
+            "stdout": r.stdout,
+            "stderr": r.stderr,
+            "exit_code": r.returncode,
+            "timed_out": False,
+        }
+    except subprocess.TimeoutExpired:
+        return {
+            "stdout": "",
+            "stderr": "Judge timed out",
+            "exit_code": -1,
+            "timed_out": True,
+        }
+    finally:
+        try:
+            os.unlink(tmp_path)
+        except OSError:
+            pass
+
+
+def collect_transcript(task_id: str, output_dir: Path, agent_id: str = "main", output_filename: str = "") -> Path:
+    """Copy the agent transcript from the container to the host."""
+    output_dir.mkdir(parents=True, exist_ok=True)
+    if output_filename:
+        filename = output_filename
+    else:
+        filename = "chat.jsonl" if agent_id == "main" else f"{agent_id}_chat.jsonl"
+    transcript_host = output_dir / filename
+
+    transcript_container = f"/root/.openclaw/agents/{agent_id}/sessions/chat.jsonl"
+    r = subprocess.run(
+        ["docker", "cp", f"{task_id}:{transcript_container}", str(transcript_host)],
+        capture_output=True, text=True, encoding="utf-8",
+    )
+
+    if r.returncode != 0:
+        # Try to find transcript within the specific agent's directory only
+        agent_dir = f"/root/.openclaw/agents/{agent_id}"
+        find_cmd = f"find {agent_dir} -name '*.jsonl' -type f 2>/dev/null | head -5"
+        find_r = subprocess.run(
+            ["docker", "exec", task_id, "/bin/bash", "-c", find_cmd],
+            capture_output=True, text=True, encoding="utf-8",
+        )
+        if find_r.stdout.strip():
+            first_jsonl = find_r.stdout.strip().splitlines()[0]
+            r2 = subprocess.run(
+                ["docker", "cp", f"{task_id}:{first_jsonl}", str(transcript_host)],
+                capture_output=True, text=True, encoding="utf-8",
+            )
+            if r2.returncode == 0:
+                logger.info("[%s] Transcript found via fallback: %s", task_id, first_jsonl)
+                return transcript_host
+
+        logger.warning("[%s] Transcript not found for agent '%s': %s",
+                       task_id, agent_id, r.stderr.strip())
+
+    return transcript_host
+
+
+def collect_output(task_id: str, output_dir: Path) -> None:
+    """Collect all task output files from the container.
+
+    Note: workspace and OpenClaw session data are NOT collected because they
+    consist entirely of static input files (database/, md files) or duplicates
+    of files already saved at the output_dir level (chat.jsonl, judge_chat.jsonl).
+    """
+    output_dir.mkdir(parents=True, exist_ok=True)
+    logger.info("[%s] Task output collected to %s", task_id, output_dir)
+
+
+def close_proc_log(proc: subprocess.Popen) -> None:
+    """Close the log file handle created by run_agent_message / start_gateway."""
+    log_file = getattr(proc, "_log_file", None)
+    if log_file and not log_file.closed:
+        log_file.close()
+
+
+def detect_transcript_errors(transcript_path: Path) -> Optional[str]:
+    """Check an OpenClaw transcript for API / provider errors.
+
+    OpenClaw reports errors as assistant messages with stopReason="error"
+    and an errorMessage field containing the details (e.g. 429, 504, etc.).
+    This is the framework's own structured error reporting — no guessing.
+
+    Returns an error description string if errors found, else None.
+    """
+    if not transcript_path or not transcript_path.exists():
+        return None
+    try:
+        text = transcript_path.read_text(encoding="utf-8", errors="replace")
+    except OSError:
+        return None
+
+    error_messages: List[str] = []
+    for line in text.splitlines():
+        line = line.strip()
+        if not line:
+            continue
+        try:
+            entry = json.loads(line)
+        except json.JSONDecodeError:
+            continue
+        if entry.get("type") != "message":
+            continue
+        msg = entry.get("message", {})
+        if msg.get("role") == "assistant" and msg.get("stopReason") == "error":
+            em = msg.get("errorMessage", "unknown error")
+            # Keep first line only (some errors contain full HTML pages)
+            first_line = em.split("\n", 1)[0].strip()[:200]
+            error_messages.append(first_line)
+
+    if error_messages:
+        sample = error_messages[0]
+        return f"API error ({len(error_messages)} occurrence(s)): {sample}"
+    return None
+
+
+def extract_usage_from_jsonl(jsonl_path: Path) -> Dict[str, Any]:
+    """Sum token usage and cost from all assistant messages in a transcript JSONL."""
+    totals: Dict[str, Any] = {
+        "input_tokens": 0,
+        "output_tokens": 0,
+        "cache_read_tokens": 0,
+        "cache_write_tokens": 0,
+        "total_tokens": 0,
+        "cost_usd": 0.0,
+        "request_count": 0,
+    }
+    if not jsonl_path.exists():
+        return totals
+    for line in jsonl_path.read_text(encoding="utf-8").splitlines():
+        line = line.strip()
+        if not line:
+            continue
+        try:
+            entry = json.loads(line)
+        except json.JSONDecodeError:
+            continue
+        if entry.get("type") != "message":
+            continue
+        msg = entry.get("message", {})
+        if msg.get("role") != "assistant":
+            continue
+        totals["request_count"] += 1
+        usage = msg.get("usage", {})
+        totals["input_tokens"] += usage.get("input", 0)
+        totals["output_tokens"] += usage.get("output", 0)
+        totals["cache_read_tokens"] += usage.get("cacheRead", 0)
+        totals["cache_write_tokens"] += usage.get("cacheWrite", 0)
+        totals["total_tokens"] += usage.get("totalTokens", 0)
+        cost = usage.get("cost", {})
+        totals["cost_usd"] += cost.get("total", 0.0)
+    totals["cost_usd"] = round(totals["cost_usd"], 6)
+    return totals
diff --git a/dataclaw/utils/grading.py b/dataclaw/utils/grading.py
new file mode 100644
index 0000000000000000000000000000000000000000..15b4c9add1ed256bb054b2c3de5d0fb10cdbef1b
--- /dev/null
+++ b/dataclaw/utils/grading.py
@@ -0,0 +1,504 @@
+"""
+LLM Judge grading engine for per-task container mode.
+
+Runs the judge agent inside the same container where the task executed,
+reusing the already-running OpenClaw gateway.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+import os
+import re
+import subprocess
+import tempfile
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any, Dict, List, Optional
+
+from dataclaw.utils.docker_utils import (
+    TMP_WORKSPACE,
+    collect_transcript,
+    detect_transcript_errors,
+    resolve_qualified_model,
+)
+
+logger = logging.getLogger(__name__)
+
+DEFAULT_JUDGE_MODEL = os.environ.get(
+    "JUDGE_MODEL", "openrouter/anthropic/claude-opus-4.5"
+)
+MAX_JUDGE_PARSE_ATTEMPTS = 5
+JUDGE_TIMEOUT_SECONDS = 180
+
+
+@dataclass
+class GradeResult:
+    task_id: str
+    score: float
+    max_score: float
+    grading_type: str
+    breakdown: Dict[str, float]
+    notes: str
+
+    def to_dict(self) -> Dict[str, Any]:
+        return {
+            "task_id": self.task_id,
+            "score": self.score,
+            "max_score": self.max_score,
+            "grading_type": self.grading_type,
+            "breakdown": self.breakdown,
+            "notes": self.notes,
+        }
+
+
+# ---------------------------------------------------------------------------
+# Public API
+# ---------------------------------------------------------------------------
+
+def grade_task(
+    *,
+    container_id: str,
+    task_id: str,
+    task_prompt: str,
+    expected_behavior: str,
+    grading_criteria: List[str],
+    llm_judge_rubric: Optional[str],
+    agent_transcript_path: Path,
+    output_dir: Path,
+    judge_model: str = DEFAULT_JUDGE_MODEL,
+) -> GradeResult:
+    """Grade a task using the LLM judge inside the task's container."""
+    final_text = _final_assistant_text(agent_transcript_path)
+    rubric = llm_judge_rubric or _format_grading_criteria(grading_criteria)
+    base_prompt = _build_judge_prompt(task_prompt, expected_behavior, final_text, rubric)
+
+    # Create judge agent inside the container
+    qualified_judge = resolve_qualified_model(container_id, judge_model)
+    agent_create = subprocess.run(
+        ["docker", "exec", container_id, "openclaw", "agents", "add", "judge",
+         "--model", qualified_judge, "--non-interactive",
+         "--workspace", "/root/.openclaw/workspace"],
+        capture_output=True, text=True, encoding="utf-8",
+    )
+    if agent_create.returncode != 0:
+        logger.warning(
+            "[%s] Judge agent creation failed (exit=%s): %s",
+            task_id, agent_create.returncode,
+            (agent_create.stderr or agent_create.stdout or "").strip()[:500],
+        )
+
+    parsed: Dict[str, Any] = {"scores": {}, "total": None, "notes": ""}
+    last_err: Optional[str] = None
+    success = False
+
+    for attempt in range(1, MAX_JUDGE_PARSE_ATTEMPTS + 1):
+        if attempt > 1:
+            # Clear previous judge sessions before retry
+            subprocess.run(
+                ["docker", "exec", container_id, "/bin/bash", "-c",
+                 "rm -rf /root/.openclaw/agents/judge/sessions/*"],
+                capture_output=True, text=True, encoding="utf-8",
+            )
+
+        # Enhance prompt only when the previous failure was a format issue
+        if attempt > 1 and last_err and last_err.startswith("format"):
+            prompt = (
+                base_prompt
+                + "\n\n---\nYour previous reply was not a valid scoring JSON with a numeric "
+                "`total`. Reply with ONLY a single JSON object, no markdown fences, no other "
+                "text, exactly this shape:\n"
+                '{"scores": {"criterion_name": 0.0}, "total": 0.0, "notes": "brief justification"}\n'
+            )
+        else:
+            prompt = base_prompt
+
+        # L1: runtime (subprocess / timeout)
+        try:
+            _run_judge_in_container(container_id, prompt)
+        except RuntimeError as exc:
+            last_err = f"runtime: {exc}"
+            logger.warning("[%s] Judge attempt %d/%d failed — %s",
+                           task_id, attempt, MAX_JUDGE_PARSE_ATTEMPTS, last_err)
+            continue
+
+        transcript_path = collect_transcript(container_id, output_dir, agent_id="judge")
+
+        # L2: transcript validity
+        if not transcript_path.exists():
+            last_err = "transcript: file missing"
+            logger.warning("[%s] Judge attempt %d/%d failed — %s",
+                           task_id, attempt, MAX_JUDGE_PARSE_ATTEMPTS, last_err)
+            continue
+
+        tr_err = detect_transcript_errors(transcript_path)
+        if tr_err:
+            last_err = f"transcript: {tr_err}"
+            logger.warning("[%s] Judge attempt %d/%d failed — %s",
+                           task_id, attempt, MAX_JUDGE_PARSE_ATTEMPTS, last_err)
+            continue
+
+        try:
+            final_text = _final_assistant_text(transcript_path)
+        except OSError as exc:
+            last_err = f"transcript: read failed: {exc}"
+            logger.warning("[%s] Judge attempt %d/%d failed — %s",
+                           task_id, attempt, MAX_JUDGE_PARSE_ATTEMPTS, last_err)
+            continue
+        if not final_text:
+            last_err = "transcript: empty assistant response"
+            logger.warning("[%s] Judge attempt %d/%d failed — %s",
+                           task_id, attempt, MAX_JUDGE_PARSE_ATTEMPTS, last_err)
+            continue
+
+        # L3: format (required fields present)
+        try:
+            raw_parsed = _parse_judge_response_from_file(transcript_path)
+        except OSError as exc:
+            last_err = f"transcript: read failed: {exc}"
+            logger.warning("[%s] Judge attempt %d/%d failed — %s",
+                           task_id, attempt, MAX_JUDGE_PARSE_ATTEMPTS, last_err)
+            continue
+        parsed = _normalize_judge_response(raw_parsed)
+
+        if _has_total_field(parsed):
+            logger.info("[%s] Judge returned valid score on attempt %d", task_id, attempt)
+            success = True
+            break
+
+        last_err = "format: missing or non-numeric total field"
+        logger.warning("[%s] Judge attempt %d/%d failed — %s",
+                       task_id, attempt, MAX_JUDGE_PARSE_ATTEMPTS, last_err)
+
+    if not success:
+        raise RuntimeError(
+            f"Judge failed after {MAX_JUDGE_PARSE_ATTEMPTS} attempts: {last_err}"
+        )
+
+    breakdown = parsed.get("scores", {})
+    total = parsed.get("total")
+    notes = parsed.get("notes", "")
+
+    return GradeResult(
+        task_id=task_id,
+        score=float(total) if total is not None else 0.0,
+        max_score=1.0,
+        grading_type="llm_judge",
+        breakdown=_normalize_score_dict(breakdown),
+        notes=str(notes) if notes is not None else "",
+    )
+
+
+# ---------------------------------------------------------------------------
+# Container interaction
+# ---------------------------------------------------------------------------
+
+def _run_judge_in_container(container_id: str, message: str) -> None:
+    """Send a message to the judge agent inside the container.
+
+    Uses docker cp + file read to bypass Windows command-line length limits.
+    Raises RuntimeError on docker cp failure, judge subprocess non-zero exit,
+    or timeout — so callers can detect and retry.
+    """
+    fd, tmp_path = tempfile.mkstemp(suffix=".txt")
+    try:
+        with os.fdopen(fd, "w", encoding="utf-8") as f:
+            f.write(message)
+
+        cp_result = subprocess.run(
+            ["docker", "cp", tmp_path,
+             f"{container_id}:/tmp/judge_prompt.txt"],
+            capture_output=True, text=True, encoding="utf-8",
+        )
+        if cp_result.returncode != 0:
+            raise RuntimeError(
+                f"docker cp failed (exit={cp_result.returncode}): "
+                f"{(cp_result.stderr or '').strip()[:300]}"
+            )
+
+        bash_cmd = (
+            f"prompt=$(cat /tmp/judge_prompt.txt) && "
+            f"cd {TMP_WORKSPACE} && "
+            f"openclaw agent --agent judge --session-id judge_chat "
+            f"--message \"$prompt\""
+        )
+        try:
+            result = subprocess.run(
+                ["docker", "exec", container_id, "/bin/bash", "-c", bash_cmd],
+                capture_output=True, text=True, encoding="utf-8",
+                timeout=JUDGE_TIMEOUT_SECONDS,
+            )
+        except subprocess.TimeoutExpired as exc:
+            raise RuntimeError(
+                f"judge timed out after {JUDGE_TIMEOUT_SECONDS}s"
+            ) from exc
+
+        if result.returncode != 0:
+            stderr_snippet = (result.stderr or result.stdout or "").strip()[:300]
+            raise RuntimeError(
+                f"judge subprocess exit={result.returncode}: {stderr_snippet}"
+            )
+    finally:
+        try:
+            os.unlink(tmp_path)
+        except OSError:
+            pass
+
+
+# ---------------------------------------------------------------------------
+# Transcript parsing
+# ---------------------------------------------------------------------------
+
+def _final_assistant_text(transcript_path: Path) -> str:
+    """Extract text from the most recent assistant message that contains
+    text content.
+
+    Agents (especially tool-heavy ones) may have their last message be a
+    pure toolCall with no text — particularly when interrupted mid-call by
+    a timeout. In that case, fall back to the most recent assistant
+    message that actually has text content so the judge still sees the
+    agent's reasoning.
+    """
+    if not transcript_path.exists():
+        return ""
+
+    last_text_message: Optional[Dict[str, Any]] = None
+    for line in transcript_path.read_text(encoding="utf-8").splitlines():
+        line = line.strip()
+        if not line:
+            continue
+        try:
+            entry = json.loads(line)
+        except json.JSONDecodeError:
+            continue
+        if entry.get("type") != "message":
+            continue
+        msg = entry.get("message", {})
+        if msg.get("role") != "assistant":
+            continue
+        content = msg.get("content", []) or []
+        has_text = any(
+            isinstance(item, dict)
+            and item.get("type") == "text"
+            and item.get("text")
+            for item in content
+        )
+        if has_text:
+            last_text_message = msg
+
+    if last_text_message is None:
+        return ""
+
+    parts: List[str] = []
+    for item in last_text_message.get("content", []) or []:
+        if isinstance(item, dict) and item.get("type") == "text" and item.get("text"):
+            parts.append(item["text"])
+    return "\n".join(parts).strip()
+
+
+def _parse_judge_response_from_file(transcript_path: Path) -> Dict[str, Any]:
+    """Parse the judge's JSON response from transcript JSONL."""
+    if not transcript_path.exists():
+        return {}
+
+    content_chunks: List[str] = []
+    for line in transcript_path.read_text(encoding="utf-8").splitlines():
+        line = line.strip()
+        if not line:
+            continue
+        try:
+            entry = json.loads(line)
+        except json.JSONDecodeError:
+            continue
+        if entry.get("type") != "message":
+            continue
+        msg = entry.get("message", {})
+        if msg.get("role") != "assistant":
+            continue
+        for item in msg.get("content", []):
+            if item.get("type") == "text":
+                content_chunks.append(item.get("text", ""))
+
+    raw_text = "\n".join(content_chunks).strip()
+    if not raw_text:
+        return {}
+
+    return _parse_json_from_text(raw_text)
+
+
+def _parse_json_from_text(raw_text: str) -> Dict[str, Any]:
+    """Extract a JSON object from free-form text."""
+    # Try code blocks first
+    code_block_match = re.search(r"```json\s*(.*?)\s*```", raw_text, re.DOTALL)
+    if code_block_match:
+        try:
+            parsed = json.loads(code_block_match.group(1))
+            if isinstance(parsed, dict):
+                return parsed
+        except json.JSONDecodeError:
+            pass
+
+    # Find balanced-brace JSON candidates
+    json_candidates: List[str] = []
+    brace_depth = 0
+    current_json: List[str] = []
+    for char in raw_text:
+        if char == "{":
+            if brace_depth == 0:
+                current_json = []
+            brace_depth += 1
+        if brace_depth > 0:
+            current_json.append(char)
+        if char == "}":
+            brace_depth -= 1
+            if brace_depth == 0 and current_json:
+                json_candidates.append("".join(current_json))
+
+    for candidate in reversed(json_candidates):
+        try:
+            parsed = json.loads(candidate)
+            if isinstance(parsed, dict) and "scores" in parsed:
+                return parsed
+        except json.JSONDecodeError:
+            continue
+    for candidate in reversed(json_candidates):
+        try:
+            parsed = json.loads(candidate)
+            if isinstance(parsed, dict):
+                return parsed
+        except json.JSONDecodeError:
+            continue
+
+    # Fallback: regex for "total: 0.XX"
+    score_pattern = re.search(
+        r"(?:total|overall|final)\s*(?:score)?[:\s]*(0\.\d+|1\.0+)",
+        raw_text,
+        re.IGNORECASE,
+    )
+    if score_pattern:
+        try:
+            total = float(score_pattern.group(1))
+            if 0.0 <= total <= 1.0:
+                return {"scores": {}, "total": total, "notes": "Score extracted from prose"}
+        except ValueError:
+            pass
+
+    logger.warning("Failed to parse judge JSON response")
+    return {}
+
+
+# ---------------------------------------------------------------------------
+# Normalization helpers
+# ---------------------------------------------------------------------------
+
+def _has_total_field(normalized: Dict[str, Any]) -> bool:
+    """Format check: total field exists and is a numeric value (range not enforced)."""
+    if not isinstance(normalized, dict):
+        return False
+    total = normalized.get("total")
+    if total is None:
+        return False
+    try:
+        t = float(total)
+    except (TypeError, ValueError):
+        return False
+    if t != t:  # NaN
+        return False
+    return True
+
+
+def _normalize_judge_response(parsed: Dict[str, Any]) -> Dict[str, Any]:
+    """Normalize judge response to {scores, total, notes}."""
+    result: Dict[str, Any] = {"scores": {}, "total": None, "notes": ""}
+
+    if "scores" in parsed:
+        scores_data = parsed["scores"]
+        if isinstance(scores_data, dict):
+            for key, value in scores_data.items():
+                if isinstance(value, dict) and "score" in value:
+                    try:
+                        result["scores"][key] = float(value["score"])
+                    except (TypeError, ValueError):
+                        pass
+                elif isinstance(value, (int, float)):
+                    result["scores"][key] = value
+    elif "criteria_scores" in parsed:
+        criteria = parsed["criteria_scores"]
+        if isinstance(criteria, dict):
+            for key, value in criteria.items():
+                if isinstance(value, dict) and "score" in value:
+                    result["scores"][key] = value["score"]
+                elif isinstance(value, (int, float)):
+                    result["scores"][key] = value
+
+    if "total" in parsed and parsed["total"] is not None:
+        try:
+            result["total"] = float(parsed["total"])
+        except (TypeError, ValueError):
+            pass
+    elif "score" in parsed and isinstance(parsed["score"], (int, float)):
+        result["total"] = float(parsed["score"])
+    elif "overall_score" in parsed and isinstance(parsed["overall_score"], (int, float)):
+        result["total"] = float(parsed["overall_score"])
+    elif result["scores"]:
+        values = [v for v in result["scores"].values() if isinstance(v, (int, float))]
+        if values:
+            result["total"] = sum(values) / len(values)
+
+    if "notes" in parsed:
+        result["notes"] = str(parsed["notes"])
+    elif "justification" in parsed:
+        result["notes"] = str(parsed["justification"])
+    elif "reasoning" in parsed:
+        result["notes"] = str(parsed["reasoning"])
+
+    return result
+
+
+def _normalize_score_dict(scores: Dict[str, Any]) -> Dict[str, float]:
+    normalized: Dict[str, float] = {}
+    for key, value in scores.items():
+        try:
+            normalized[str(key)] = float(value)
+        except (TypeError, ValueError):
+            continue
+    return normalized
+
+
+# ---------------------------------------------------------------------------
+# Prompt building
+# ---------------------------------------------------------------------------
+
+def _format_grading_criteria(criteria: List[str]) -> str:
+    if not criteria:
+        return ""
+    return "\n".join(f"- {c}" for c in criteria)
+
+
+def _build_judge_prompt(
+    task_prompt: str,
+    expected_behavior: str,
+    agent_final_text: str,
+    rubric: str,
+) -> str:
+    return (
+        "You are a grading function. Your ONLY job is to output a single JSON object.\n\n"
+        "CRITICAL RULES:\n"
+        "- Do NOT use any tools (no Read, Write, exec, or any other tool calls)\n"
+        "- Do NOT create files or run commands\n"
+        "- Do NOT write any prose, explanation, or commentary outside the JSON\n"
+        "- Respond with ONLY a JSON object — nothing else\n\n"
+        "Be a strict evaluator. Judge the final assistant message against the task and rubric.\n\n"
+        "## Task\n"
+        f"{task_prompt}\n\n"
+        "## Expected Behavior\n"
+        f"{expected_behavior}\n\n"
+        "## Agent final answer\n"
+        f"{agent_final_text}\n\n"
+        "## Grading Rubric\n"
+        f"{rubric}\n\n"
+        "Score each criterion from 0.0 to 1.0.\n\n"
+        "Respond with ONLY this JSON structure (no markdown, no code fences, no extra text):\n"
+        '{"scores": {"criterion_name": 0.0}, "total": 0.0, "notes": "brief justification"}'
+    )
diff --git a/dataclaw/utils/process_grading.py b/dataclaw/utils/process_grading.py
new file mode 100644
index 0000000000000000000000000000000000000000..3135dda262d30da5c40e4878254bc2a0a5f03c5b
--- /dev/null
+++ b/dataclaw/utils/process_grading.py
@@ -0,0 +1,761 @@
+"""
+Process-oriented grading metrics for DataClaw.
+
+Provides fine-grained evaluation of agent execution trajectories
+beyond outcome-only scoring.
+
+Metrics implemented:
+- Execution Efficiency: S_gold / S_agent
+- Goal Progress Rate (GPR): LLM-Judge-based milestone evaluation
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+import math
+import random
+import re
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any, Dict, List, Optional, Tuple
+
+logger = logging.getLogger(__name__)
+
+# ---------------------------------------------------------------------------
+# Constants
+# ---------------------------------------------------------------------------
+
+NUMERIC_REL_TOL = 0.01      # 1% relative tolerance for candidate extraction
+NUMERIC_ABS_TOL = 1e-6      # absolute tolerance for values near zero
+SNIPPET_CONTEXT_CHARS = 300  # ~100 tokens context window around matches
+SHORT_RESULT_TOKEN_THRESHOLD = 1000  # tool results <= this are included whole
+HIGH_FREQ_THRESHOLD = 5             # value occurrences above this trigger key-proximity filter
+MAX_SNIPPETS_WHEN_HIGH_FREQ = 5      # max snippets to randomly sample when high-freq filtered
+CHARS_PER_TOKEN = 3                  # rough chars-per-token estimate for token counting
+ARG_MAX_BYTES = 2 * 1024 * 1024     # Linux ARG_MAX ~2MB — prompt size guard before subprocess
+
+
+# ---------------------------------------------------------------------------
+# Numeric helpers
+# ---------------------------------------------------------------------------
+
+def _is_numeric(value: Any) -> bool:
+    if isinstance(value, bool):
+        return False
+    return isinstance(value, (int, float))
+
+
+def _numbers_match(expected: float, actual: float) -> bool:
+    if math.isnan(expected) or math.isnan(actual):
+        return False
+    if expected == actual:
+        return True
+    if abs(expected) < NUMERIC_ABS_TOL:
+        return abs(actual - expected) < NUMERIC_ABS_TOL
+    return abs(actual - expected) / abs(expected) <= NUMERIC_REL_TOL
+
+
+# ---------------------------------------------------------------------------
+# Trajectory parsing
+# ---------------------------------------------------------------------------
+
+@dataclass
+class AgentStep:
+    """A single agent step, corresponding to one assistant turn."""
+    index: int
+    tool_calls: List[Dict[str, Any]]
+    text: str
+    tool_results: List[str] = field(default_factory=list)
+    timestamp: Optional[str] = None
+
+    @property
+    def has_tool_call(self) -> bool:
+        return len(self.tool_calls) > 0
+
+    @property
+    def tool_call_count(self) -> int:
+        return len(self.tool_calls)
+
+    @property
+    def full_text(self) -> str:
+        parts = []
+        if self.text:
+            parts.append(self.text)
+        parts.extend(self.tool_results)
+        return "\n".join(parts)
+
+
+def parse_trajectory(transcript_path: Path) -> List[AgentStep]:
+    """Parse chat.jsonl into a list of AgentSteps.
+
+    Each assistant message is one step. Tool results following an
+    assistant turn are attached to that step.
+    """
+    if not transcript_path.exists():
+        logger.warning("Transcript not found: %s", transcript_path)
+        return []
+
+    messages: List[Dict[str, Any]] = []
+    for line in transcript_path.read_text(encoding="utf-8").splitlines():
+        line = line.strip()
+        if not line:
+            continue
+        try:
+            entry = json.loads(line)
+        except json.JSONDecodeError:
+            continue
+        if entry.get("type") != "message":
+            continue
+        messages.append(entry)
+
+    steps: List[AgentStep] = []
+    current_step: Optional[AgentStep] = None
+    step_index = 0
+
+    for entry in messages:
+        msg = entry.get("message", {})
+        role = msg.get("role", "")
+
+        if role == "assistant":
+            content = msg.get("content", []) or []
+            text_parts: List[str] = []
+            tool_calls: List[Dict[str, Any]] = []
+
+            for item in content:
+                if item.get("type") == "text" and item.get("text"):
+                    text_parts.append(item["text"])
+                elif item.get("type") == "toolCall":
+                    tool_calls.append({
+                        "id": item.get("id", ""),
+                        "name": item.get("name", ""),
+                        "arguments": item.get("arguments", {}),
+                    })
+
+            step_index += 1
+            current_step = AgentStep(
+                index=step_index,
+                tool_calls=tool_calls,
+                text="\n".join(text_parts).strip(),
+                timestamp=msg.get("timestamp") or entry.get("timestamp"),
+            )
+            steps.append(current_step)
+
+        elif role == "toolResult" and current_step is not None:
+            content = msg.get("content", []) or []
+            for item in content:
+                if item.get("type") == "text" and item.get("text"):
+                    current_step.tool_results.append(item["text"])
+
+        else:
+            current_step = None
+
+    return steps
+
+
+# ---------------------------------------------------------------------------
+# Execution Efficiency
+# ---------------------------------------------------------------------------
+
+@dataclass
+class EfficiencyResult:
+    s_agent: int
+    s_gold: int
+    efficiency: Optional[float]
+
+    def to_dict(self) -> Dict[str, Any]:
+        return {
+            "s_agent": self.s_agent,
+            "s_gold": self.s_gold,
+            "efficiency": self.efficiency,
+        }
+
+
+def compute_efficiency(steps: List[AgentStep], s_gold: int) -> EfficiencyResult:
+    """Compute Execution Efficiency = S_gold / S_agent."""
+    s_agent = len(steps)
+    if s_agent == 0:
+        return EfficiencyResult(s_agent=0, s_gold=s_gold, efficiency=None)
+    return EfficiencyResult(
+        s_agent=s_agent,
+        s_gold=s_gold,
+        efficiency=round(s_gold / s_agent, 4),
+    )
+
+
+# ---------------------------------------------------------------------------
+# GPR: Candidate snippet extraction
+# ---------------------------------------------------------------------------
+
+@dataclass
+class CandidateSnippet:
+    """A context snippet from a tool result containing milestone candidate values."""
+    snippet: str
+    milestone_matches: List[Tuple[str, float, float]]  # (key, expected, matched)
+    tool_result_index: int
+
+
+def _extract_candidates_for_step(
+    step: AgentStep,
+    numeric_milestones: Dict[str, float],
+    context_chars: int = SNIPPET_CONTEXT_CHARS,
+) -> List[CandidateSnippet]:
+    """Extract context snippets around numeric milestone matches in tool results.
+
+    Short tool results (≤ SHORT_RESULT_TOKEN_THRESHOLD tokens) are included whole.
+
+    For long tool results, windows are extracted per milestone.  If a milestone
+    value appears more than HIGH_FREQ_THRESHOLD times (likely a raw data column),
+    only snippets that contain at least one component of the milestone key name
+    are kept, and at most MAX_SNIPPETS_WHEN_HIGH_FREQ are randomly sampled.
+    If none survive the key-proximity filter, the milestone is skipped for that
+    tool result entirely.
+    """
+    snippets: List[CandidateSnippet] = []
+
+    for tr_idx, tr_text in enumerate(step.tool_results):
+        approx_tokens = len(tr_text) // CHARS_PER_TOKEN
+
+        if approx_tokens <= SHORT_RESULT_TOKEN_THRESHOLD:
+            # Short result: include whole text, annotate with all matching milestones
+            matched: List[Tuple[str, float, float]] = []
+            seen_keys: set = set()
+            for num_match in re.finditer(r'-?\d+(?:\.\d+)?', tr_text):
+                try:
+                    num_val = float(num_match.group())
+                except ValueError:
+                    continue
+                for key, expected in numeric_milestones.items():
+                    if key not in seen_keys and _numbers_match(expected, num_val):
+                        matched.append((key, expected, num_val))
+                        seen_keys.add(key)
+            if matched:
+                snippets.append(CandidateSnippet(
+                    snippet=tr_text,
+                    milestone_matches=matched,
+                    tool_result_index=tr_idx,
+                ))
+            continue
+
+        # Long result: per-milestone window extraction + frequency-based filtering
+        for key, expected in numeric_milestones.items():
+            # Collect all positions where this milestone value appears
+            match_positions: List[Tuple[int, int, float]] = []
+            for num_match in re.finditer(r'-?\d+(?:\.\d+)?', tr_text):
+                try:
+                    num_val = float(num_match.group())
+                except ValueError:
+                    continue
+                if _numbers_match(expected, num_val):
+                    match_positions.append((num_match.start(), num_match.end(), num_val))
+
+            if not match_positions:
+                continue
+
+            value_freq = len(match_positions)
+
+            # Build and merge overlapping windows
+            raw_windows = [
+                (max(0, start - context_chars), min(len(tr_text), end + context_chars), val)
+                for start, end, val in match_positions
+            ]
+            raw_windows.sort()
+            cur_start, cur_end, cur_val = raw_windows[0]
+            merged_windows: List[Tuple[int, int, float]] = []
+            for start, end, val in raw_windows[1:]:
+                if start <= cur_end:
+                    cur_end = max(cur_end, end)
+                else:
+                    merged_windows.append((cur_start, cur_end, cur_val))
+                    cur_start, cur_end, cur_val = start, end, val
+            merged_windows.append((cur_start, cur_end, cur_val))
+
+            candidate_texts: List[Tuple[str, float]] = [
+                (tr_text[s:e], v) for s, e, v in merged_windows
+            ]
+
+            if value_freq > HIGH_FREQ_THRESHOLD:
+                # High-frequency value: require key component in snippet
+                key_parts = [p.lower() for p in key.split('_') if p]
+                filtered = [
+                    (s, v) for s, v in candidate_texts
+                    if any(part in s.lower() for part in key_parts)
+                ]
+                if not filtered:
+                    continue  # no reliable evidence for this milestone in this tool result
+                candidate_texts = random.sample(
+                    filtered, min(MAX_SNIPPETS_WHEN_HIGH_FREQ, len(filtered))
+                )
+
+            for snippet_text, matched_val in candidate_texts:
+                snippets.append(CandidateSnippet(
+                    snippet=snippet_text,
+                    milestone_matches=[(key, expected, matched_val)],
+                    tool_result_index=tr_idx,
+                ))
+
+    return snippets
+
+
+# ---------------------------------------------------------------------------
+# GPR: Judge prompt construction
+# ---------------------------------------------------------------------------
+
+def build_gpr_judge_prompt(
+    steps: List[AgentStep],
+    milestones: Dict[str, Any],
+    gold_steps: List[str],
+    final_answer_correct: bool,
+) -> str:
+    """Build the LLM Judge prompt for GPR milestone evaluation.
+
+    Constructs the prompt with:
+    1. Gold reference steps and milestones
+    2. Final answer correctness status
+    3. Agent trajectory with assistant text + candidate snippets
+
+    If the assembled prompt exceeds ARG_MAX_BYTES, falls back to a text-only
+    trajectory (no tool-result snippets) to stay within the Linux ARG_MAX limit.
+    """
+    # --- Format gold steps ---
+    steps_formatted = "\n".join(
+        f"{i + 1}. {s}" for i, s in enumerate(gold_steps)
+    )
+
+    # --- Format milestones ---
+    milestone_lines = []
+    for i, (key, value) in enumerate(milestones.items()):
+        milestone_lines.append(f"M{i + 1}. {key}: {value}")
+    milestones_formatted = "\n".join(milestone_lines)
+
+    # --- Extract numeric milestones for candidate search ---
+    numeric_milestones = {
+        k: float(v) for k, v in milestones.items() if _is_numeric(v)
+    }
+
+    # --- Final answer status ---
+    final_answer_status = (
+        "The agent's final answer is CORRECT."
+        if final_answer_correct
+        else "The agent's final answer is INCORRECT."
+    )
+
+    # --- Inner helpers ---
+    def _format_trajectory(include_candidates: bool) -> str:
+        parts: List[str] = []
+        for step in steps:
+            lines: List[str] = [f"### Step {step.index}"]
+
+            if step.text:
+                lines.append(f"Assistant: {step.text}")
+            else:
+                lines.append("Assistant: (no text)")
+
+            if include_candidates:
+                candidates = _extract_candidates_for_step(step, numeric_milestones)
+                if candidates:
+                    for c in candidates:
+                        keys_str = ", ".join(
+                            f"{k}≈{matched}" for k, _exp, matched in c.milestone_matches
+                        )
+                        lines.append(f"[Candidate: {keys_str}]")
+                        lines.append(c.snippet)
+                elif step.tool_results:
+                    lines.append("(tool executed, no milestone candidates detected)")
+            else:
+                if step.tool_results:
+                    lines.append("(tool executed, results omitted — prompt size limit)")
+
+            parts.append("\n".join(lines))
+        return "\n\n".join(parts)
+
+    def _assemble(trajectory: str) -> str:
+        return f"""\
+You are a milestone evaluator for an AI agent benchmark. Your task is to determine \
+which milestones an agent achieved during its execution trajectory.
+
+## Reference Solution Path
+
+The following ordered steps describe the gold-standard approach to solving this task. \
+Milestones are key intermediate results that should be produced along this path.
+
+### Gold Steps (in order)
+{steps_formatted}
+
+### Milestones to Evaluate
+{milestones_formatted}
+
+## Agent Execution Trajectory
+
+Below is a reconstruction of the agent's work, organized by step (assistant turn). \
+For each step, "Assistant" is the agent's own text output. "Candidate" sections are \
+excerpts from tool execution outputs where milestone-relevant values were detected \
+by numeric matching. These candidates may or may not be true milestone achievements — \
+you must verify the semantic context.
+
+{trajectory}
+
+## Evaluation Rules
+
+1. **Direct evidence**: A milestone is achieved if the trajectory clearly shows the \
+agent computed or obtained the expected value (or a value within 1% relative error) \
+in the CORRECT semantic context. A number appearing in an unrelated context \
+(e.g., a value for Province A matching a milestone defined for Province B) does NOT count.
+
+2. **Temporal coupling inference**: Milestones follow a logical dependency chain as \
+defined in the Gold Steps. If a downstream milestone is correctly achieved, its \
+upstream dependencies can be inferred as achieved — even if they were not explicitly \
+output. For example, if the agent correctly computed "ratio = A/B", then milestones \
+for computing A and B can be inferred as achieved.
+
+3. **Chain-break identification**: If the final answer is INCORRECT, identify the \
+earliest milestone in the logical chain that was NOT achieved — this is the \
+"break point" where the agent's reasoning diverged from the correct path.
+
+4. **Different-but-valid paths**: The agent may use a different method than the gold \
+steps. Judge milestone achievement based on whether the agent obtained the correct \
+intermediate values, regardless of method.
+
+## Output Format
+
+Respond with ONLY a JSON object. No markdown fences, no extra text.
+
+{{"milestones": [{{"key": "milestone key", "achieved": true, "evidence_type": "direct", "first_step": 4, "reason": "brief justification"}}, ...], "break_point": null, "chain_summary": "one-sentence summary"}}
+
+Where:
+- "achieved": true or false
+- "evidence_type": "direct" (found in trajectory), "inferred" (implied by downstream achievement), or "final_answer" (implied by correct final answer)
+- "first_step": the agent Step number where this milestone was first achieved (integer), or null if not achieved. For "inferred" milestones, use the step of the downstream milestone that implies it. For "final_answer" milestones, use the step of the agent's final answer.
+- "break_point": null if all achieved, otherwise 0-based index of first unachieved milestone
+- "chain_summary": where and why the chain broke, or "All milestones achieved" if none broke"""
+
+    # --- Assemble with candidates; fall back to text-only if prompt is too large ---
+    prompt = _assemble(_format_trajectory(include_candidates=True))
+    prompt_bytes = len(prompt.encode("utf-8"))
+    if prompt_bytes > ARG_MAX_BYTES:
+        logger.warning(
+            "GPR judge prompt is %d bytes, exceeds ARG_MAX (%d bytes); "
+            "falling back to text-only trajectory",
+            prompt_bytes, ARG_MAX_BYTES,
+        )
+        prompt = _assemble(_format_trajectory(include_candidates=False))
+
+    return prompt
+
+
+# ---------------------------------------------------------------------------
+# GPR: Judge response parsing
+# ---------------------------------------------------------------------------
+
+@dataclass
+class MilestoneResult:
+    """Evaluation result for a single milestone."""
+    key: str
+    expected: Any
+    achieved: bool
+    evidence_type: str = ""   # "direct" | "inferred" | "final_answer"
+    reason: str = ""
+    first_step: Optional[int] = None
+
+    def to_dict(self) -> Dict[str, Any]:
+        d: Dict[str, Any] = {
+            "key": self.key,
+            "expected": self.expected,
+            "achieved": self.achieved,
+            "evidence_type": self.evidence_type,
+            "reason": self.reason,
+        }
+        if self.first_step is not None:
+            d["first_step"] = self.first_step
+        return d
+
+
+@dataclass
+class GPRResult:
+    """Result of the Goal Progress Rate metric."""
+    gpr: float
+    milestones_total: int
+    milestones_achieved: int
+    break_point: Optional[int]
+    chain_summary: str
+    details: List[MilestoneResult]
+
+    def to_dict(self) -> Dict[str, Any]:
+        return {
+            "gpr": self.gpr,
+            "milestones_total": self.milestones_total,
+            "milestones_achieved": self.milestones_achieved,
+            "break_point": self.break_point,
+            "chain_summary": self.chain_summary,
+            "details": [d.to_dict() for d in self.details],
+        }
+
+
+def _parse_json_from_text(raw_text: str) -> Dict[str, Any]:
+    """Extract a JSON object from free-form text (handles markdown fences, etc.)."""
+    # Try code blocks first
+    code_block = re.search(r"```(?:json)?\s*(.*?)\s*```", raw_text, re.DOTALL)
+    if code_block:
+        try:
+            parsed = json.loads(code_block.group(1))
+            if isinstance(parsed, dict):
+                return parsed
+        except json.JSONDecodeError:
+            pass
+
+    # Find balanced-brace JSON candidates
+    candidates: List[str] = []
+    depth = 0
+    current: List[str] = []
+    for char in raw_text:
+        if char == "{":
+            if depth == 0:
+                current = []
+            depth += 1
+        if depth > 0:
+            current.append(char)
+        if char == "}":
+            depth -= 1
+            if depth == 0 and current:
+                candidates.append("".join(current))
+
+    # Prefer the one with "milestones" key
+    for c in reversed(candidates):
+        try:
+            parsed = json.loads(c)
+            if isinstance(parsed, dict) and "milestones" in parsed:
+                return parsed
+        except json.JSONDecodeError:
+            continue
+    # Fallback: any valid dict
+    for c in reversed(candidates):
+        try:
+            parsed = json.loads(c)
+            if isinstance(parsed, dict):
+                return parsed
+        except json.JSONDecodeError:
+            continue
+
+    return {}
+
+
+def parse_gpr_judge_response(
+    raw_text: str,
+    milestones: Dict[str, Any],
+) -> GPRResult:
+    """Parse the LLM Judge response into a GPRResult.
+
+    If parsing fails, returns a GPRResult with all milestones marked unachieved.
+    """
+    milestone_keys = list(milestones.keys())
+    n = len(milestone_keys)
+
+    if not raw_text.strip():
+        logger.warning("Empty judge response for GPR")
+        return _fallback_gpr_result(milestones)
+
+    parsed = _parse_json_from_text(raw_text)
+    if not parsed or "milestones" not in parsed:
+        logger.warning("Failed to parse GPR judge response")
+        return _fallback_gpr_result(milestones)
+
+    judge_milestones = parsed["milestones"]
+    if not isinstance(judge_milestones, list):
+        logger.warning("Judge 'milestones' is not a list")
+        return _fallback_gpr_result(milestones)
+
+    # Build a lookup from judge response by key
+    judge_lookup: Dict[str, Dict[str, Any]] = {}
+    for item in judge_milestones:
+        if isinstance(item, dict) and "key" in item:
+            judge_lookup[item["key"]] = item
+
+    # Match judge results to our milestones (by key or by position)
+    details: List[MilestoneResult] = []
+    for i, key in enumerate(milestone_keys):
+        expected = milestones[key]
+
+        # Try key-based match first, then positional fallback
+        judge_item = judge_lookup.get(key)
+        if judge_item is None and i < len(judge_milestones):
+            judge_item = judge_milestones[i]
+
+        if judge_item and isinstance(judge_item, dict):
+            achieved = bool(judge_item.get("achieved", False))
+            evidence_type = str(judge_item.get("evidence_type", ""))
+            reason = str(judge_item.get("reason", ""))
+            raw_step = judge_item.get("first_step")
+            try:
+                first_step = int(raw_step) if raw_step is not None else None
+            except (TypeError, ValueError):
+                first_step = None
+        else:
+            achieved = False
+            evidence_type = ""
+            reason = "Not found in judge response"
+            first_step = None
+
+        details.append(MilestoneResult(
+            key=key,
+            expected=expected,
+            achieved=achieved,
+            evidence_type=evidence_type,
+            reason=reason,
+            first_step=first_step,
+        ))
+
+    achieved_count = sum(1 for d in details if d.achieved)
+    gpr = achieved_count / n if n > 0 else 0.0
+
+    break_point = parsed.get("break_point")
+    if break_point is not None:
+        try:
+            break_point = int(break_point)
+        except (TypeError, ValueError):
+            break_point = None
+    chain_summary = str(parsed.get("chain_summary", ""))
+
+    return GPRResult(
+        gpr=round(gpr, 4),
+        milestones_total=n,
+        milestones_achieved=achieved_count,
+        break_point=break_point,
+        chain_summary=chain_summary,
+        details=details,
+    )
+
+
+def _fallback_gpr_result(milestones: Dict[str, Any]) -> GPRResult:
+    """Return a GPRResult with all milestones marked as unachieved."""
+    details = [
+        MilestoneResult(
+            key=key,
+            expected=value,
+            achieved=False,
+            reason="Judge response could not be parsed",
+        )
+        for key, value in milestones.items()
+    ]
+    return GPRResult(
+        gpr=0.0,
+        milestones_total=len(milestones),
+        milestones_achieved=0,
+        break_point=0,
+        chain_summary="Judge response could not be parsed",
+        details=details,
+    )
+
+
+# ---------------------------------------------------------------------------
+# Temporal Goal Progress Rate (TGPR) and Temporal Progress Efficiency (TPE)
+# ---------------------------------------------------------------------------
+
+DEFAULT_GAMMA = 0.9  # exponential decay factor
+
+
+@dataclass
+class TGPRResult:
+    """Result of the Temporal Goal Progress Rate (TGPR) and Temporal Progress
+    Efficiency (TPE) metrics. Both share the same per-milestone decay table;
+    TGPR normalises by total milestones, TPE by achieved milestones (eval.tex)."""
+    tgpr: float
+    tpe: float
+    gpr: float
+    gamma: float
+    s_gold: int
+    milestones_total: int
+    milestones_achieved: int
+    details: List[Dict[str, Any]]
+
+    def to_dict(self) -> Dict[str, Any]:
+        return {
+            "tgpr": self.tgpr,
+            "tpe": self.tpe,
+            "gpr": self.gpr,
+            "gamma": self.gamma,
+            "s_gold": self.s_gold,
+            "milestones_total": self.milestones_total,
+            "milestones_achieved": self.milestones_achieved,
+            "details": self.details,
+        }
+
+
+def compute_tgpr(
+    gpr_result: GPRResult,
+    s_gold: int,
+    gamma: float = DEFAULT_GAMMA,
+) -> TGPRResult:
+    """Compute Temporal Goal Progress Rate (TGPR) and Temporal Progress
+    Efficiency (TPE).
+
+    TGPR = (1/n_total)    * sum_i  I(m_i) * gamma^max(t_i - S_gold, 0)
+    TPE  = (1/n_achieved) * sum_i  I(m_i) * gamma^max(t_i - S_gold, 0)
+                                                   (per eval.tex)
+
+    Uses S_gold (total gold-path steps) as a global baseline:
+    - If the agent achieves a milestone within S_gold steps, no penalty.
+    - If it takes more than S_gold steps, exponential decay kicks in.
+
+    For unachieved milestones, the contribution is 0.
+    Raises ValueError when the Judge marks a milestone achieved but
+    omits first_step — TPE has no defensible fallback for that case
+    (a decay=1 default would silently inflate the score).
+
+    TPE differs from TGPR only by the denominator: it averages decay over
+    *achieved* milestones, isolating timing from coverage. TPE = 0 when no
+    milestones were achieved.
+    """
+    n = gpr_result.milestones_total
+    if n == 0:
+        return TGPRResult(
+            tgpr=0.0, tpe=0.0, gpr=0.0, gamma=gamma, s_gold=s_gold,
+            milestones_total=0, milestones_achieved=0, details=[],
+        )
+
+    details: List[Dict[str, Any]] = []
+    tgpr_sum = 0.0
+    decay_sum = 0.0
+    achieved_count = 0
+
+    for milestone in gpr_result.details:
+        entry: Dict[str, Any] = {
+            "key": milestone.key,
+            "achieved": milestone.achieved,
+            "t_agent": milestone.first_step,
+            "s_gold": s_gold,
+            "delay": None,
+            "decay": None,
+            "contribution": 0.0,
+        }
+
+        if milestone.achieved and milestone.first_step is not None:
+            delay = max(milestone.first_step - s_gold, 0)
+            decay = gamma ** delay
+            contribution = (1.0 / n) * decay
+            entry["delay"] = delay
+            entry["decay"] = round(decay, 6)
+            entry["contribution"] = round(contribution, 6)
+            tgpr_sum += contribution
+            decay_sum += decay
+            achieved_count += 1
+        elif milestone.achieved and milestone.first_step is None:
+            raise ValueError(
+                f"Milestone {milestone.key!r} marked achieved but the judge "
+                "did not report first_step; refusing to compute TPE/TGPR "
+                "with a fabricated decay value."
+            )
+
+        details.append(entry)
+
+    tpe_value = decay_sum / achieved_count if achieved_count > 0 else 0.0
+
+    return TGPRResult(
+        tgpr=round(tgpr_sum, 4),
+        tpe=round(tpe_value, 4),
+        gpr=gpr_result.gpr,
+        gamma=gamma,
+        s_gold=s_gold,
+        milestones_total=n,
+        milestones_achieved=achieved_count,
+        details=details,
+    )
diff --git a/docs/index.html b/docs/index.html
new file mode 100644
index 0000000000000000000000000000000000000000..8c0a0fbdf92098ee7b2c4c0b43dd2442edfc2494
--- /dev/null
+++ b/docs/index.html
@@ -0,0 +1,811 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+<meta charset="UTF-8"/>
+<meta name="viewport" content="width=device-width,initial-scale=1.0"/>
+<title>DataClaw Leaderboard</title>
+<style>
+:root {
+  --bg: #0f172a;
+  --bg2: #1e293b;
+  --bg3: #334155;
+  --glass: rgba(30,41,59,0.65);
+  --glass-border: rgba(255,255,255,0.08);
+  --text: #f1f5f9;
+  --text2: #94a3b8;
+  --text3: #64748b;
+  --cyan: #06b6d4;
+  --blue: #3b82f6;
+  --emerald: #10b981;
+  --amber: #f59e0b;
+  --rose: #f43f5e;
+  --violet: #8b5cf6;
+  --grad: linear-gradient(135deg, #06b6d4, #3b82f6);
+  --grad2: linear-gradient(135deg, #06b6d4 0%, #8b5cf6 100%);
+  --r: 12px;
+  --r2: 8px;
+  --ff: system-ui, -apple-system, "Segoe UI", Roboto, "Helvetica Neue", Arial, sans-serif;
+  --fm: "SF Mono", "Cascadia Code", "Fira Code", Consolas, monospace;
+}
+*, *::before, *::after { box-sizing: border-box; margin: 0; padding: 0; }
+html { scroll-behavior: smooth; }
+body { background: var(--bg); color: var(--text); font-family: var(--ff); font-size: 14px; line-height: 1.6; min-height: 100vh; overflow-x: hidden; }
+a { color: var(--cyan); text-decoration: none; }
+a:hover { color: var(--blue); }
+
+/* ── NAV ── */
+.nav {
+  position: sticky; top: 0; z-index: 100;
+  background: rgba(15,23,42,0.85);
+  backdrop-filter: blur(16px); -webkit-backdrop-filter: blur(16px);
+  border-bottom: 1px solid var(--glass-border);
+  padding: 0 32px; height: 56px;
+  display: flex; align-items: center; justify-content: space-between;
+}
+.nav-brand { font-size: 18px; font-weight: 700; letter-spacing: -0.5px; }
+.nav-brand span { background: var(--grad); -webkit-background-clip: text; -webkit-text-fill-color: transparent; background-clip: text; }
+.nav-links { display: flex; gap: 4px; }
+.nav-links a {
+  font-size: 12px; padding: 6px 14px; border-radius: 6px;
+  color: var(--text2); letter-spacing: 0.5px; transition: all 0.15s;
+}
+.nav-links a:hover, .nav-links a.active { color: var(--text); background: var(--bg2); }
+.nav-right { display: flex; align-items: center; gap: 12px; }
+.lang-btn {
+  font-family: var(--ff); font-size: 11px; color: var(--text3); background: var(--bg2);
+  border: 1px solid var(--glass-border); border-radius: 6px; padding: 4px 10px;
+  cursor: pointer; transition: all 0.15s;
+}
+.lang-btn:hover { color: var(--text); border-color: var(--cyan); }
+
+/* ── HERO ── */
+.hero {
+  position: relative; overflow: hidden;
+  padding: 80px 32px 64px; text-align: center;
+}
+.hero::before {
+  content: ''; position: absolute; top: -40%; left: 50%; transform: translateX(-50%);
+  width: 600px; height: 600px; border-radius: 50%;
+  background: radial-gradient(circle, rgba(6,182,212,0.12) 0%, rgba(59,130,246,0.06) 40%, transparent 70%);
+  animation: pulse 6s ease-in-out infinite alternate;
+}
+@keyframes pulse { from { transform: translateX(-50%) scale(1); opacity: 0.7; } to { transform: translateX(-50%) scale(1.15); opacity: 1; } }
+.hero-title { position: relative; font-size: 56px; font-weight: 800; letter-spacing: -2px; margin-bottom: 12px; }
+.hero-title span { background: var(--grad); -webkit-background-clip: text; -webkit-text-fill-color: transparent; background-clip: text; }
+.hero-sub { position: relative; font-size: 16px; color: var(--text2); max-width: 600px; margin: 0 auto 32px; }
+.hero-pills { position: relative; display: flex; justify-content: center; gap: 16px; flex-wrap: wrap; }
+.pill {
+  display: inline-flex; align-items: center; gap: 8px;
+  background: var(--glass); border: 1px solid var(--glass-border);
+  backdrop-filter: blur(8px); -webkit-backdrop-filter: blur(8px);
+  border-radius: 20px; padding: 8px 20px; font-size: 13px; color: var(--text2);
+}
+.pill strong { color: var(--text); font-weight: 700; font-size: 16px; }
+.pill strong.accent { background: var(--grad); -webkit-background-clip: text; -webkit-text-fill-color: transparent; background-clip: text; }
+
+/* ── SECTION ── */
+.section { max-width: 1200px; margin: 0 auto; padding: 0 32px; }
+.sect-header { display: flex; align-items: center; gap: 16px; margin: 56px 0 24px; }
+.sect-header h2 { font-size: 22px; font-weight: 700; white-space: nowrap; }
+.sect-header h2 span { background: var(--grad); -webkit-background-clip: text; -webkit-text-fill-color: transparent; background-clip: text; }
+.sect-header::after { content: ''; flex: 1; height: 1px; background: linear-gradient(90deg, var(--glass-border), transparent); }
+
+/* ── OVERVIEW CARDS ── */
+.overview-row { display: grid; grid-template-columns: repeat(auto-fit, minmax(280px, 1fr)); gap: 16px; margin-bottom: 24px; }
+.ov-card {
+  position: relative; overflow: hidden;
+  background: var(--glass); border: 1px solid var(--glass-border);
+  backdrop-filter: blur(12px); -webkit-backdrop-filter: blur(12px);
+  border-radius: var(--r); padding: 24px;
+  transition: border-color 0.2s, box-shadow 0.2s;
+}
+.ov-card:hover { border-color: rgba(6,182,212,0.3); box-shadow: 0 4px 24px rgba(6,182,212,0.08); }
+.ov-card.gold { border-color: rgba(6,182,212,0.4); }
+.ov-card.gold::before {
+  content: ''; position: absolute; top: 0; left: 0; right: 0; height: 3px;
+  background: var(--grad);
+}
+.ov-rank {
+  position: absolute; top: 16px; right: 16px;
+  width: 32px; height: 32px; border-radius: 50%;
+  display: flex; align-items: center; justify-content: center;
+  font-size: 14px; font-weight: 700;
+}
+.rank-1 { background: linear-gradient(135deg, #06b6d4, #3b82f6); color: #fff; }
+.rank-2 { background: rgba(148,163,184,0.2); color: var(--text2); border: 1px solid rgba(148,163,184,0.3); }
+.rank-3 { background: rgba(245,158,11,0.15); color: var(--amber); border: 1px solid rgba(245,158,11,0.3); }
+.ov-model { font-size: 18px; font-weight: 700; margin-bottom: 16px; padding-right: 40px; }
+.ov-score { font-size: 36px; font-weight: 800; margin-bottom: 4px; }
+.ov-score span { background: var(--grad); -webkit-background-clip: text; -webkit-text-fill-color: transparent; background-clip: text; }
+.ov-label { font-size: 11px; color: var(--text3); text-transform: uppercase; letter-spacing: 1.5px; margin-bottom: 16px; }
+.ov-stats { display: grid; grid-template-columns: repeat(3, 1fr); gap: 12px; }
+.ov-stat-val { font-size: 14px; font-weight: 600; color: var(--text); font-family: var(--fm); }
+.ov-stat-key { font-size: 10px; color: var(--text3); text-transform: uppercase; letter-spacing: 1px; margin-top: 2px; }
+
+/* ── TABLE ── */
+.table-wrap {
+  background: var(--glass); border: 1px solid var(--glass-border);
+  backdrop-filter: blur(12px); -webkit-backdrop-filter: blur(12px);
+  border-radius: var(--r); overflow: hidden;
+}
+.table-wrap table { width: 100%; border-collapse: collapse; font-size: 13px; }
+.table-wrap thead th {
+  padding: 14px 16px; text-align: left;
+  font-size: 10px; letter-spacing: 1.5px; text-transform: uppercase;
+  color: var(--text3); background: rgba(15,23,42,0.5);
+  border-bottom: 1px solid var(--glass-border);
+  cursor: pointer; user-select: none; white-space: nowrap;
+  transition: color 0.15s;
+}
+.table-wrap thead th:hover { color: var(--cyan); }
+.table-wrap thead th.sorted { color: var(--cyan); }
+.sort-arrow { margin-left: 4px; font-size: 9px; opacity: 0.4; }
+.sorted .sort-arrow { opacity: 1; }
+.table-wrap tbody tr {
+  border-bottom: 1px solid rgba(255,255,255,0.03);
+  transition: background 0.1s;
+}
+.table-wrap tbody tr:hover { background: rgba(6,182,212,0.04); }
+.table-wrap tbody td { padding: 12px 16px; vertical-align: middle; }
+.table-wrap tbody tr:last-child { border-bottom: none; }
+td.rank-cell { text-align: center; font-weight: 700; font-size: 15px; width: 48px; }
+td.rank-cell.r1 { color: var(--cyan); }
+td.rank-cell.r2 { color: var(--text2); }
+td.rank-cell.r3 { color: var(--amber); }
+td.model-cell { font-weight: 600; }
+.score-bar { display: flex; align-items: center; gap: 10px; min-width: 180px; }
+.score-bar-bg { flex: 1; height: 6px; background: var(--bg); border-radius: 3px; overflow: hidden; }
+.score-bar-fill { height: 100%; border-radius: 3px; background: var(--grad); transition: width 0.6s cubic-bezier(0.16,1,0.3,1); }
+.score-bar-val { font-family: var(--fm); font-weight: 600; min-width: 52px; text-align: right; font-size: 14px; }
+td .mono { font-family: var(--fm); font-size: 12px; color: var(--text2); }
+
+/* ── CATEGORY BARS ── */
+.cat-group { margin-bottom: 24px; }
+.cat-group-title { font-size: 13px; font-weight: 600; color: var(--text); margin-bottom: 10px; display: flex; align-items: center; gap: 8px; }
+.cat-group-title .cat-badge { font-size: 10px; color: var(--text3); background: var(--bg2); padding: 2px 8px; border-radius: 4px; font-weight: 400; }
+.bar-row { display: flex; align-items: center; gap: 12px; margin-bottom: 6px; }
+.bar-model { font-size: 11px; color: var(--text2); width: 140px; text-align: right; flex-shrink: 0; overflow: hidden; text-overflow: ellipsis; white-space: nowrap; }
+.bar-track { flex: 1; height: 22px; background: var(--bg); border-radius: 4px; overflow: hidden; position: relative; }
+.bar-fill { height: 100%; border-radius: 4px; transition: width 0.8s cubic-bezier(0.16,1,0.3,1); display: flex; align-items: center; padding-left: 8px; }
+.bar-fill span { font-size: 10px; font-weight: 600; color: rgba(255,255,255,0.9); white-space: nowrap; }
+.bar-pct { font-family: var(--fm); font-size: 12px; font-weight: 600; min-width: 52px; text-align: right; flex-shrink: 0; color: var(--text); }
+.m-color-0 { background: linear-gradient(90deg, #06b6d4, #0ea5e9); color: var(--cyan); }
+.m-color-1 { background: linear-gradient(90deg, #3b82f6, #6366f1); color: var(--blue); }
+.m-color-2 { background: linear-gradient(90deg, #8b5cf6, #a78bfa); color: var(--violet); }
+.m-color-3 { background: linear-gradient(90deg, #10b981, #34d399); color: var(--emerald); }
+.m-color-4 { background: linear-gradient(90deg, #f59e0b, #fbbf24); color: var(--amber); }
+.m-color-5 { background: linear-gradient(90deg, #f43f5e, #fb7185); color: var(--rose); }
+
+/* ── FILTERS ── */
+.filter-row { display: flex; gap: 12px; margin-bottom: 16px; flex-wrap: wrap; }
+.filter-select {
+  font-family: var(--ff); font-size: 12px; color: var(--text);
+  background: var(--bg2); border: 1px solid var(--glass-border);
+  border-radius: 6px; padding: 6px 12px; cursor: pointer;
+  transition: border-color 0.15s;
+}
+.filter-select:hover, .filter-select:focus { border-color: var(--cyan); outline: none; }
+.filter-select option { background: var(--bg2); color: var(--text); }
+.load-more-wrap { display: flex; align-items: center; justify-content: center; gap: 16px; margin-top: 16px; }
+.load-more-info { font-family: var(--ff); font-size: 12px; color: var(--text3); }
+.load-more-btn {
+  font-family: var(--ff); font-size: 12px; color: var(--text);
+  background: var(--bg2); border: 1px solid var(--glass-border);
+  border-radius: 6px; padding: 6px 16px; cursor: pointer;
+  transition: border-color 0.15s, color 0.15s;
+}
+.load-more-btn:hover { border-color: var(--cyan); color: var(--cyan); }
+
+/* ── METRICS ── */
+.metrics-row { display: grid; grid-template-columns: repeat(auto-fit, minmax(240px, 1fr)); gap: 14px; }
+.sect-action {
+  display: inline-flex; align-items: center; gap: 6px;
+  font-family: var(--ff); font-size: 12px; font-weight: 500; color: var(--text2);
+  background: var(--bg2); border: 1px solid var(--glass-border);
+  border-radius: 6px; padding: 5px 12px; cursor: pointer;
+  white-space: nowrap; flex-shrink: 0;
+  transition: color 0.15s, border-color 0.15s, background 0.15s;
+}
+.sect-action:hover { color: var(--cyan); border-color: rgba(6,182,212,0.4); background: rgba(6,182,212,0.08); }
+.metric-card {
+  position: relative; overflow: hidden;
+  background: var(--glass); border: 1px solid var(--glass-border);
+  backdrop-filter: blur(12px); -webkit-backdrop-filter: blur(12px);
+  border-radius: var(--r); padding: 20px;
+  transition: border-color 0.2s;
+}
+.metric-card:hover { border-color: rgba(6,182,212,0.3); }
+.metric-head { display: flex; align-items: baseline; gap: 8px; margin-bottom: 4px; }
+.metric-abbr { font-size: 20px; font-weight: 800; }
+.metric-abbr span { background: var(--grad); -webkit-background-clip: text; -webkit-text-fill-color: transparent; background-clip: text; }
+.metric-dir { font-size: 12px; color: var(--emerald); font-weight: 600; }
+.metric-name { font-size: 12px; color: var(--text2); margin-bottom: 10px; }
+.metric-desc { font-size: 12px; color: var(--text2); line-height: 1.55; margin-bottom: 10px; }
+.metric-scope {
+  display: inline-block; font-size: 10px; letter-spacing: 1px; text-transform: uppercase;
+  padding: 3px 8px; border-radius: 4px; font-weight: 600;
+}
+.scope-all { background: rgba(59,130,246,0.15); color: #93c5fd; border: 1px solid rgba(59,130,246,0.3); }
+.scope-correct { background: rgba(16,185,129,0.15); color: #6ee7b7; border: 1px solid rgba(16,185,129,0.3); }
+.scope-incorrect { background: rgba(244,63,94,0.15); color: #fda4af; border: 1px solid rgba(244,63,94,0.3); }
+
+/* ── HEADER TOOLTIPS ── */
+.th-with-info { cursor: help; }
+.th-info { font-size: 9px; opacity: 0.5; margin-left: 4px; }
+.th-with-info:hover .th-info { opacity: 1; color: var(--cyan); }
+.th-scope { display: block; font-size: 8px; color: var(--text3); letter-spacing: 0.5px; text-transform: none; margin-top: 2px; font-weight: 400; }
+.th-clickable { cursor: pointer; }
+.th-clickable:hover .th-info { opacity: 1; color: var(--cyan); }
+
+/* ── MODAL ── */
+.modal-overlay {
+  position: fixed; inset: 0; z-index: 1000;
+  background: rgba(15,23,42,0.75); backdrop-filter: blur(6px);
+  display: none; align-items: center; justify-content: center;
+  padding: 24px;
+}
+.modal-overlay.show { display: flex; animation: fadeIn 0.15s ease; }
+@keyframes fadeIn { from { opacity: 0; } to { opacity: 1; } }
+.modal {
+  background: var(--bg2); border: 1px solid var(--glass-border);
+  border-radius: var(--r); max-width: 720px; width: 100%; max-height: 86vh;
+  overflow: auto; box-shadow: 0 20px 60px rgba(0,0,0,0.5);
+}
+.modal-head {
+  display: flex; align-items: center; justify-content: space-between;
+  padding: 18px 20px; border-bottom: 1px solid var(--glass-border);
+}
+.modal-title { font-size: 15px; font-weight: 700; }
+.modal-title .modal-sub { display: block; font-size: 11px; color: var(--text3); font-weight: 400; margin-top: 4px; letter-spacing: 0.3px; }
+.modal-close {
+  background: none; border: none; color: var(--text2); font-size: 22px;
+  cursor: pointer; padding: 0 4px; line-height: 1; transition: color 0.15s;
+}
+.modal-close:hover { color: var(--rose); }
+.modal-body { padding: 16px 20px 20px; }
+.modal-body table { width: 100%; border-collapse: collapse; font-size: 12px; }
+.modal-body thead th {
+  text-align: right; padding: 8px 10px; font-size: 10px; letter-spacing: 1px;
+  text-transform: uppercase; color: var(--text3);
+  border-bottom: 1px solid var(--glass-border);
+}
+.modal-body thead th:first-child { text-align: left; }
+.modal-body tbody td {
+  padding: 8px 10px; border-bottom: 1px solid rgba(255,255,255,0.04);
+  font-family: var(--fm); text-align: right;
+}
+.modal-body tbody td:first-child { font-family: var(--ff); text-align: left; font-weight: 600; }
+.modal-body tbody tr:last-child td { border-bottom: none; }
+.modal-body tbody td.na { color: var(--text3); }
+.modal-foot { padding: 0 20px 16px; font-size: 11px; color: var(--text3); }
+
+/* ── FOOTER ── */
+.footer {
+  max-width: 1200px; margin: 64px auto 0; padding: 24px 32px;
+  border-top: 1px solid var(--glass-border);
+  font-size: 11px; color: var(--text3); text-align: center;
+  display: flex; justify-content: center; gap: 24px; flex-wrap: wrap;
+}
+
+/* ── ANIMATIONS ── */
+@keyframes fadeUp { from { opacity: 0; transform: translateY(12px); } to { opacity: 1; transform: translateY(0); } }
+.fade-in { animation: fadeUp 0.5s ease both; }
+.fade-in-1 { animation-delay: 0.08s; }
+.fade-in-2 { animation-delay: 0.16s; }
+.fade-in-3 { animation-delay: 0.24s; }
+
+/* ── RESPONSIVE ── */
+@media (max-width: 768px) {
+  .nav { padding: 0 16px; }
+  .nav-links { display: none; }
+  .hero { padding: 48px 16px 40px; }
+  .hero-title { font-size: 36px; }
+  .section { padding: 0 16px; }
+  .overview-row { grid-template-columns: 1fr; }
+  .table-wrap { overflow-x: auto; }
+  .bar-model { width: 90px; font-size: 10px; }
+
+}
+</style>
+</head>
+<body>
+
+<!-- NAV -->
+<nav class="nav">
+  <div class="nav-brand"><span>DataClaw</span></div>
+  <div class="nav-links">
+    <a href="#overview" class="active" data-en="Overview" data-zh="总览">Overview</a>
+    <a href="#leaderboard" data-en="Leaderboard" data-zh="排行榜">Leaderboard</a>
+    <a href="#categories" data-en="Categories" data-zh="分类详情">Categories</a>
+    <a href="#tasks" data-en="Task Explorer" data-zh="任务浏览">Task Explorer</a>
+  </div>
+  <div class="nav-right">
+    <button class="lang-btn" onclick="toggleLang()" id="lang-btn">中文</button>
+    <a href="https://github.com/GTMLLab/DataClaw" target="_blank" style="color:var(--text3);" title="GitHub">
+      <svg width="18" height="18" viewBox="0 0 24 24" fill="currentColor"><path d="M12 0C5.37 0 0 5.37 0 12c0 5.31 3.435 9.795 8.205 11.385.6.105.825-.255.825-.57 0-.285-.015-1.23-.015-2.235-3.015.555-3.795-.735-4.035-1.41-.135-.345-.72-1.41-1.23-1.695-.42-.225-1.02-.78-.015-.795.945-.015 1.62.87 1.845 1.23 1.08 1.815 2.805 1.305 3.495.99.105-.78.42-1.305.765-1.605-2.67-.3-5.46-1.335-5.46-5.925 0-1.305.465-2.385 1.23-3.225-.12-.3-.54-1.53.12-3.18 0 0 1.005-.315 3.3 1.23.96-.27 1.98-.405 3-.405s2.04.135 3 .405c2.295-1.56 3.3-1.23 3.3-1.23.66 1.65.24 2.88.12 3.18.765.84 1.23 1.905 1.23 3.225 0 4.605-2.805 5.625-5.475 5.925.435.375.81 1.095.81 2.22 0 1.605-.015 2.895-.015 3.3 0 .315.225.69.825.57A12.02 12.02 0 0024 12c0-6.63-5.37-12-12-12z"/></svg>
+    </a>
+  </div>
+</nav>
+
+<!-- HERO -->
+<header class="hero" id="overview">
+  <h1 class="hero-title fade-in"><span>DataClaw</span></h1>
+  <p class="hero-sub fade-in fade-in-1" data-en="Data-Analysis Benchmark for OpenClaw-Style End-to-End Agents" data-zh="面向 OpenClaw 类端到端智能体的数据分析评测基准">Data-Analysis Benchmark for OpenClaw-Style End-to-End Agents</p>
+  <div class="hero-pills fade-in fade-in-2">
+    <div class="pill"><strong class="accent" id="pill-tasks">492</strong> <span data-en="Tasks" data-zh="任务">Tasks</span></div>
+    <div class="pill"><strong class="accent" id="pill-cats">7</strong> <span data-en="Categories" data-zh="分类">Categories</span></div>
+    <div class="pill"><strong class="accent" id="pill-models">0</strong> <span data-en="Models Evaluated" data-zh="已评测模型">Models Evaluated</span></div>
+  </div>
+</header>
+
+<!-- OVERVIEW CARDS -->
+<div class="section">
+  <div class="sect-header fade-in">
+    <h2><span data-en="Top Models" data-zh="模型排名">Top Models</span></h2>
+    <button class="sect-action" onclick="openMetricsModal()" title="View metric definitions"><span>📐</span> <span data-en="Metric Details" data-zh="Metrics 细节">Metric Details</span></button>
+  </div>
+  <div class="overview-row" id="overview-cards"></div>
+</div>
+
+<!-- LEADERBOARD -->
+<div class="section" id="leaderboard">
+  <div class="sect-header fade-in"><h2><span data-en="Leaderboard" data-zh="排行榜">Leaderboard</span></h2></div>
+  <div class="table-wrap fade-in">
+    <table>
+      <thead><tr>
+        <th data-col="rank"># <span class="sort-arrow">↕</span></th>
+        <th data-col="model" data-en="Model" data-zh="模型">Model <span class="sort-arrow">↕</span></th>
+        <th data-col="acc" class="sorted th-with-info" title="Acc ↑ — LLM-judged final-answer correctness, normalized over L subquestions. Range [0,1]."><span data-en="Acc." data-zh="准确率">Acc.</span> ↑ <span class="th-info">ⓘ</span> <span class="sort-arrow">↓</span><span class="th-scope" data-en="all tasks" data-zh="全部任务">all tasks</span></th>
+        <th data-col="ee" class="th-with-info" title="EE ↑ — Execution Efficiency = N / T (gold steps / agent steps). Computed only on correct tasks."><span>EE</span> ↑ <span class="th-info">ⓘ</span> <span class="sort-arrow">↕</span><span class="th-scope" data-en="correct only" data-zh="仅 correct">correct only</span></th>
+        <th data-col="gpr" class="th-with-info" title="GPR ↑ — Goal Progress Rate = (1/M) Σ 𝕀(mⱼ achieved). Computed only on incorrect tasks."><span>GPR</span> ↑ <span class="th-info">ⓘ</span> <span class="sort-arrow">↕</span><span class="th-scope" data-en="incorrect only" data-zh="仅 incorrect">incorrect only</span></th>
+        <th data-col="tpe" class="th-with-info" title="TPE ↑ — Temporal Progress Efficiency = (Σ 𝕀(mⱼ)·γ^max(tⱼ−N,0)) / Σ 𝕀(mⱼ), γ=0.9. Computed only on incorrect tasks."><span>TPE</span> ↑ <span class="th-info">ⓘ</span> <span class="sort-arrow">↕</span><span class="th-scope" data-en="incorrect only" data-zh="仅 incorrect">incorrect only</span></th>
+        <th data-col="total_tokens" data-en="Tokens (M)" data-zh="Token (M)">Tokens (M) <span class="sort-arrow">↕</span></th>
+        <th data-col="total_cost_usd" class="th-clickable" onclick="openPricingModal(event)" title="Click to view per-million-token unit prices used to compute these costs."><span data-en="Cost ($)" data-zh="费用 ($)">Cost ($)</span> <span class="th-info">ⓘ</span> <span class="sort-arrow">↕</span></th>
+        <th data-col="total_requests" data-en="Requests" data-zh="请求数">Requests <span class="sort-arrow">↕</span></th>
+      </tr></thead>
+      <tbody id="lb-body"></tbody>
+    </table>
+  </div>
+</div>
+
+<!-- CATEGORY BREAKDOWN -->
+<div class="section" id="categories">
+  <div class="sect-header fade-in"><h2><span data-en="Category Breakdown" data-zh="分类详情">Category Breakdown</span></h2></div>
+  <div id="cat-bars" class="fade-in"></div>
+</div>
+
+<!-- TASK EXPLORER -->
+<div class="section" id="tasks">
+  <div class="sect-header fade-in"><h2><span data-en="Task Explorer" data-zh="任务浏览">Task Explorer</span></h2></div>
+  <div class="filter-row fade-in">
+    <select class="filter-select" id="filter-cat" onchange="resetTaskTable()">
+      <option value="all" data-en="All Categories" data-zh="全部分类">All Categories</option>
+    </select>
+    <select class="filter-select" id="filter-model" onchange="resetTaskTable()">
+      <option value="all" data-en="All Models" data-zh="全部模型">All Models</option>
+    </select>
+  </div>
+  <div class="table-wrap fade-in">
+    <table>
+      <thead><tr>
+        <th data-en="Task ID" data-zh="任务ID">Task ID</th>
+        <th data-en="Category" data-zh="分类">Category</th>
+        <th data-en="Model" data-zh="模型">Model</th>
+        <th data-tcol="score" data-en="Score" data-zh="得分">Score</th>
+        <th data-tcol="ee" class="th-with-info" title="EE ↑ — N / T. Shown only when this task was solved correctly (score = 1)."><span>EE</span> ↑ <span class="th-info">ⓘ</span></th>
+        <th data-tcol="gpr" class="th-with-info" title="GPR ↑ — milestone fraction. Shown only when this task was answered incorrectly."><span>GPR</span> ↑ <span class="th-info">ⓘ</span></th>
+        <th data-tcol="tpe" class="th-with-info" title="TPE ↑ — temporal-decay average over achieved milestones (γ=0.9). Shown only when this task was answered incorrectly."><span>TPE</span> ↑ <span class="th-info">ⓘ</span></th>
+        <th data-tcol="tokens" data-en="Tokens" data-zh="Token">Tokens</th>
+        <th data-tcol="cost_usd" class="th-clickable" onclick="openPricingModal(event)" title="Click to view per-million-token unit prices used to compute these costs."><span data-en="Cost" data-zh="费用">Cost</span> <span class="th-info">ⓘ</span></th>
+        <th data-tcol="elapsed_time" data-en="Time (s)" data-zh="耗时 (s)">Time (s)</th>
+      </tr></thead>
+      <tbody id="task-body"></tbody>
+    </table>
+  </div>
+  <div class="load-more-wrap fade-in" id="load-more-wrap" style="display:none">
+    <span class="load-more-info" id="load-more-info"></span>
+    <button class="load-more-btn" id="load-more-btn" onclick="loadMoreTasks()" data-en="Load more" data-zh="加载更多">Load more</button>
+  </div>
+</div>
+
+<!-- METRICS MODAL -->
+<div class="modal-overlay" id="metrics-modal" onclick="closeMetricsModalOverlay(event)">
+  <div class="modal" onclick="event.stopPropagation()">
+    <div class="modal-head">
+      <div class="modal-title">
+        <span data-en="Evaluation Metrics" data-zh="评测指标">Evaluation Metrics</span>
+        <span class="modal-sub" data-en="Four metrics from the DataClaw protocol (paper §4)." data-zh="DataClaw 评测协议中的四个指标（论文 §4）。">Four metrics from the DataClaw protocol (paper §4).</span>
+      </div>
+      <button class="modal-close" onclick="closeMetricsModal()" aria-label="Close">×</button>
+    </div>
+    <div class="modal-body">
+      <div class="metrics-row">
+        <div class="metric-card">
+          <div class="metric-head"><div class="metric-abbr"><span>Acc</span></div><div class="metric-dir">↑</div></div>
+          <div class="metric-name" data-en="Accuracy" data-zh="准确率">Accuracy</div>
+          <div class="metric-desc" data-en="LLM-judge semantic match between predicted answer â and gold answer a; multi-subquestion tasks return the normalized mean over L subquestions." data-zh="LLM-judge 评判预测答案 â 与标准答案 a 的语义一致性，含 L 个子问题时取归一化平均。">LLM-judge semantic match between predicted answer â and gold answer a; multi-subquestion tasks return the normalized mean over L subquestions.</div>
+          <div class="metric-scope scope-all" data-en="All tasks" data-zh="全部任务">All tasks</div>
+        </div>
+        <div class="metric-card">
+          <div class="metric-head"><div class="metric-abbr"><span>EE</span></div><div class="metric-dir">↑</div></div>
+          <div class="metric-name" data-en="Execution Efficiency" data-zh="执行效率">Execution Efficiency</div>
+          <div class="metric-desc" data-en="Ratio of gold reference step count N to the agent's actual step count T. EE > 1 means the agent solved it in fewer steps than the gold trajectory; EE < 1 means it took more." data-zh="gold 参考步数 N 与 agent 实际步数 T 之比。EE>1 表示比 gold 更省步，EE<1 表示更费步。">Ratio of gold reference step count N to the agent's actual step count T. EE > 1 means the agent solved it in fewer steps than the gold trajectory; EE < 1 means it took more.</div>
+          <div class="metric-scope scope-correct" data-en="Correct tasks only" data-zh="仅 correct 任务">Correct tasks only</div>
+        </div>
+        <div class="metric-card">
+          <div class="metric-head"><div class="metric-abbr"><span>GPR</span></div><div class="metric-dir">↑</div></div>
+          <div class="metric-name" data-en="Goal Progress Rate" data-zh="目标进展率">Goal Progress Rate</div>
+          <div class="metric-desc" data-en="Fraction of M annotated milestones that the agent achieves anywhere in its trajectory. Captures partial process credit when the final answer is wrong." data-zh="agent 在轨迹中达成的里程碑占 M 个标注里程碑的比例，用于在答错时捕捉过程得分。">Fraction of M annotated milestones that the agent achieves anywhere in its trajectory. Captures partial process credit when the final answer is wrong.</div>
+          <div class="metric-scope scope-incorrect" data-en="Incorrect tasks only" data-zh="仅 incorrect 任务">Incorrect tasks only</div>
+        </div>
+        <div class="metric-card">
+          <div class="metric-head"><div class="metric-abbr"><span>TPE</span></div><div class="metric-dir">↑</div></div>
+          <div class="metric-name" data-en="Temporal Progress Efficiency" data-zh="时序进展效率">Temporal Progress Efficiency</div>
+          <div class="metric-desc" data-en="Averages an exponential temporal-decay factor γ^max(tⱼ−N,0) over the milestones the agent achieved. Milestones reached by step N contribute 1; later ones decay (γ = 0.9). TPE = 1 means all achieved milestones were on time; lower values mean later concentration." data-zh="对 agent 已达成的里程碑做时序衰减平均：第 tⱼ 步达成的里程碑按 γ^max(tⱼ−N,0) 加权（γ=0.9），N 步内达成贡献 1，越晚衰减越多。所有已达成里程碑均按时则 TPE=1，越靠后越低。">Averages an exponential temporal-decay factor γ^max(tⱼ−N,0) over the milestones the agent achieved. Milestones reached by step N contribute 1; later ones decay (γ = 0.9). TPE = 1 means all achieved milestones were on time; lower values mean later concentration.</div>
+          <div class="metric-scope scope-incorrect" data-en="Incorrect tasks only" data-zh="仅 incorrect 任务">Incorrect tasks only</div>
+        </div>
+      </div>
+    </div>
+  </div>
+</div>
+
+<!-- PRICING MODAL -->
+<div class="modal-overlay" id="pricing-modal" onclick="closePricingModalOverlay(event)">
+  <div class="modal" onclick="event.stopPropagation()">
+    <div class="modal-head">
+      <div class="modal-title">
+        <span data-en="Token Pricing" data-zh="Token 单价">Token Pricing</span>
+        <span class="modal-sub" id="pricing-source"></span>
+      </div>
+      <button class="modal-close" onclick="closePricingModal()" aria-label="Close">×</button>
+    </div>
+    <div class="modal-body">
+      <table>
+        <thead><tr>
+          <th data-en="Model" data-zh="模型">Model</th>
+          <th data-en="Input" data-zh="输入">Input</th>
+          <th data-en="Output" data-zh="输出">Output</th>
+          <th data-en="Cache Read" data-zh="缓存读">Cache Read</th>
+          <th data-en="Cache Write" data-zh="缓存写">Cache Write</th>
+        </tr></thead>
+        <tbody id="pricing-body"></tbody>
+      </table>
+    </div>
+    <div class="modal-foot">
+      <span data-en="Cost = (input·rᵢ + output·rₒ + cache_read·r_cr + cache_write·r_cw) / 10⁶. When a provider does not advertise a cache rate (—) but tokens are billed under that tier, cache_read falls back to the input rate and cache_write to the output rate."
+            data-zh="费用 = (input·rᵢ + output·rₒ + cache_read·r_cr + cache_write·r_cw) / 10⁶。若价格表中为 —，但实际计费存在该项 tokens，则 cache_read 用 input 单价、cache_write 用 output 单价兜底。">Cost = (input·rᵢ + output·rₒ + cache_read·r_cr + cache_write·r_cw) / 10⁶. When a provider does not advertise a cache rate (—) but tokens are billed under that tier, cache_read falls back to the input rate and cache_write to the output rate.</span>
+    </div>
+  </div>
+</div>
+
+<!-- FOOTER -->
+<footer class="footer" id="footer-info"></footer>
+
+<script>
+const D = {"generated_at":"2026-04-27 17:42 UTC","benchmark_version":"e9fe78b","total_tasks":492,"total_categories":7,"evaluated_tasks":492,"evaluated_categories":7,"category_names":{"comprehensive_decision":"Comprehensive Decision","enterprise_industry_analysis":"Enterprise-Industry Analysis","enterprise_industry_policy_analysis":"Enterprise-Industry-Policy Analysis","hypothesis_verification":"Hypothesis Verification","industry_planning":"Industry Planning","international_comparison":"International Comparison","risk_assessment":"Risk Assessment"},"category_names_zh":{"comprehensive_decision":"综合决策","enterprise_industry_analysis":"企业产业分析","enterprise_industry_policy_analysis":"企业产业政策分析","hypothesis_verification":"假设验证","industry_planning":"产业规划","international_comparison":"国际比较","risk_assessment":"风险评估"},"pricing":[{"model":"claude-opus-4-6-20260205","input":5.0,"output":25.0,"cache_read":0.5,"cache_write":6.25},{"model":"gemini-3.1-pro-preview","input":2.0,"output":12.0,"cache_read":0.2,"cache_write":0.375},{"model":"minimax-m2.7","input":0.3,"output":1.2,"cache_read":0.059,"cache_write":null},{"model":"qwen3.5-plus","input":0.26,"output":1.56,"cache_read":null,"cache_write":0.325},{"model":"deepseek-v3.2","input":0.252,"output":0.378,"cache_read":0.0252,"cache_write":null},{"model":"glm-5","input":0.6,"output":2.08,"cache_read":0.12,"cache_write":null},{"model":"kimi-k2.5","input":0.44,"output":2.0,"cache_read":0.22,"cache_write":null},{"model":"gpt-5.4-2026-03-05","input":2.5,"output":15.0,"cache_read":0.25,"cache_write":null}],"pricing_source":"OpenRouter list rates as of 2026-04-25 (USD per 1M tokens).","pricing_source_zh":"OpenRouter 公示单价，截止 2026-04-25（美元/百万 tokens）。","models":[{"model":"claude-opus-4-6-20260205","acc":0.6343,"ee":0.4221,"gpr":0.4508,"tpe":0.5206,"efficiency":{"total_tokens":282070496,"total_cost_usd":1454.26,"total_requests":7358},"category_scores":{"comprehensive_decision":{"acc":0.5871,"ee":0.4876,"gpr":0.4776,"tpe":0.5274,"count":70},"enterprise_industry_analysis":{"acc":0.7434,"ee":0.4196,"gpr":0.4624,"tpe":0.4936,"count":226},"enterprise_industry_policy_analysis":{"acc":0.5132,"ee":0.4113,"gpr":0.5099,"tpe":0.6731,"count":76},"hypothesis_verification":{"acc":0.3764,"ee":0.4357,"gpr":0.2587,"tpe":0.5862,"count":29},"industry_planning":{"acc":0.6311,"ee":0.4035,"gpr":0.3706,"tpe":0.3425,"count":28},"international_comparison":{"acc":0.5149,"ee":0.31,"gpr":0.5324,"tpe":0.3335,"count":39},"risk_assessment":{"acc":0.6389,"ee":0.4259,"gpr":0.3379,"tpe":0.6425,"count":24}},"tasks":[{"task_id":"task_001_comprehensive_decision_easy_easy001","category":"comprehensive_decision","score":0.0,"tokens":94300,"cost_usd":0.49836,"elapsed_time":117.77,"gpr":0.6667,"tpe":0.8145,"ee":0.5},{"task_id":"task_002_comprehensive_decision_easy_easy002","category":"comprehensive_decision","score":1.0,"tokens":99342,"cost_usd":0.51455,"elapsed_time":47.04,"gpr":1.0,"tpe":0.7059,"ee":0.2857},{"task_id":"task_003_comprehensive_decision_easy_easy003","category":"comprehensive_decision","score":1.0,"tokens":3885564,"cost_usd":19.60582,"elapsed_time":659.6,"gpr":1.0,"tpe":0.0053,"ee":0.0566},{"task_id":"task_004_comprehensive_decision_easy_easy004","category":"comprehensive_decision","score":1.0,"tokens":102257,"cost_usd":0.534225,"elapsed_time":70.66,"gpr":1.0,"tpe":0.7221,"ee":0.2857},{"task_id":"task_005_comprehensive_decision_easy_easy005","category":"comprehensive_decision","score":0.6,"tokens":292020,"cost_usd":1.51404,"elapsed_time":131.26,"gpr":0.6667,"tpe":0.5214,"ee":0.1765},{"task_id":"task_006_comprehensive_decision_easy_easy006","category":"comprehensive_decision","score":1.0,"tokens":416698,"cost_usd":2.15571,"elapsed_time":217.64,"gpr":1.0,"tpe":0.3138,"ee":0.1765},{"task_id":"task_007_comprehensive_decision_hard_hard001","category":"comprehensive_decision","score":0.0,"tokens":1275499,"cost_usd":6.781035,"elapsed_time":647.89,"gpr":0.625,"tpe":0.9063,"ee":0.3889},{"task_id":"task_008_comprehensive_decision_hard_hard002","category":"comprehensive_decision","score":0.0,"tokens":432420,"cost_usd":2.34454,"elapsed_time":235.33,"gpr":0.4,"tpe":0.4305,"ee":0.4375},{"task_id":"task_009_comprehensive_decision_hard_hard003","category":"comprehensive_decision","score":0.0,"tokens":1351376,"cost_usd":6.97706,"elapsed_time":335.21,"gpr":0.7143,"tpe":0.5624,"ee":0.28},{"task_id":"task_010_comprehensive_decision_hard_hard004","category":"comprehensive_decision","score":0.0,"tokens":558997,"cost_usd":3.054845,"elapsed_time":293.49,"gpr":0.875,"tpe":0.4371,"ee":0.3182},{"task_id":"task_011_comprehensive_decision_hard_hard005","category":"comprehensive_decision","score":0.0,"tokens":2040497,"cost_usd":10.536865,"elapsed_time":521.62,"gpr":0.8571,"tpe":0.3313,"ee":0.2258},{"task_id":"task_012_comprehensive_decision_hard_hard006","category":"comprehensive_decision","score":0.0,"tokens":458923,"cost_usd":2.561995,"elapsed_time":260.58,"gpr":0.3333,"tpe":0.5314,"ee":0.3684},{"task_id":"task_013_comprehensive_decision_hard_hard007","category":"comprehensive_decision","score":0.0,"tokens":3385048,"cost_usd":17.3661,"elapsed_time":858.7,"gpr":0.6667,"tpe":0.2786,"ee":0.1429},{"task_id":"task_014_comprehensive_decision_hard_hard008","category":"comprehensive_decision","score":1.0,"tokens":769797,"cost_usd":4.123825,"elapsed_time":349.84,"gpr":1.0,"tpe":0.4251,"ee":0.35},{"task_id":"task_015_comprehensive_decision_hard_hard009","category":"comprehensive_decision","score":1.0,"tokens":425376,"cost_usd":2.25914,"elapsed_time":275.82,"gpr":1.0,"tpe":0.8357,"ee":0.8},{"task_id":"task_016_comprehensive_decision_hard_hard010","category":"comprehensive_decision","score":0.0,"tokens":510284,"cost_usd":2.7935,"elapsed_time":263.01,"gpr":0.4286,"tpe":0.4512,"ee":0.3889},{"task_id":"task_017_comprehensive_decision_hard_hard011","category":"comprehensive_decision","score":0.0,"tokens":939097,"cost_usd":4.818265,"elapsed_time":183.9,"gpr":0.7778,"tpe":1.0,"ee":0.7},{"task_id":"task_018_comprehensive_decision_hard_hard012","category":"comprehensive_decision","score":1.0,"tokens":355558,"cost_usd":1.90837,"elapsed_time":282.93,"gpr":1.0,"tpe":0.5966,"ee":0.4667},{"task_id":"task_019_comprehensive_decision_hard_hard013","category":"comprehensive_decision","score":1.0,"tokens":174611,"cost_usd":0.953155,"elapsed_time":175.87,"gpr":1.0,"tpe":0.9789,"ee":0.7778},{"task_id":"task_020_comprehensive_decision_hard_hard014","category":"comprehensive_decision","score":0.0,"tokens":830873,"cost_usd":4.339305,"elapsed_time":400.49,"gpr":0.5714,"tpe":0.3538,"ee":0.2692},{"task_id":"task_021_comprehensive_decision_hard_hard015","category":"comprehensive_decision","score":1.0,"tokens":320044,"cost_usd":1.77172,"elapsed_time":297.02,"gpr":1.0,"tpe":0.8064,"ee":0.5714},{"task_id":"task_022_comprehensive_decision_hard_hard016","category":"comprehensive_decision","score":0.0,"tokens":366769,"cost_usd":1.946125,"elapsed_time":143.09,"gpr":1.0,"tpe":0.579,"ee":0.4286},{"task_id":"task_023_comprehensive_decision_hard_hard017","category":"comprehensive_decision","score":0.0,"tokens":1336045,"cost_usd":6.742245,"elapsed_time":187.62,"gpr":0.5,"tpe":0.3874,"ee":0.375},{"task_id":"task_024_comprehensive_decision_hard_hard018","category":"comprehensive_decision","score":1.0,"tokens":523837,"cost_usd":2.694985,"elapsed_time":193.3,"gpr":1.0,"tpe":0.6577,"ee":0.5},{"task_id":"task_025_comprehensive_decision_hard_hard019","category":"comprehensive_decision","score":1.0,"tokens":108327,"cost_usd":0.614555,"elapsed_time":90.26,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_026_comprehensive_decision_medium_medium001","category":"comprehensive_decision","score":1.0,"tokens":579256,"cost_usd":3.05484,"elapsed_time":290.38,"gpr":1.0,"tpe":0.4168,"ee":0.2381},{"task_id":"task_027_comprehensive_decision_medium_medium002","category":"comprehensive_decision","score":1.0,"tokens":255457,"cost_usd":1.335665,"elapsed_time":102.82,"gpr":1.0,"tpe":0.8526,"ee":0.4167},{"task_id":"task_028_comprehensive_decision_medium_medium003","category":"comprehensive_decision","score":0.5,"tokens":1002857,"cost_usd":5.236585,"elapsed_time":369.03,"gpr":0.4286,"tpe":0.3466,"ee":0.1613},{"task_id":"task_029_comprehensive_decision_medium_medium004","category":"comprehensive_decision","score":0.0,"tokens":446596,"cost_usd":2.37478,"elapsed_time":326.13,"gpr":0.875,"tpe":0.7273,"ee":0.3125},{"task_id":"task_030_comprehensive_decision_medium_medium005","category":"comprehensive_decision","score":1.0,"tokens":180703,"cost_usd":0.959995,"elapsed_time":229.34,"gpr":1.0,"tpe":1.0,"ee":0.5556},{"task_id":"task_031_comprehensive_decision_medium_medium006","category":"comprehensive_decision","score":1.0,"tokens":88321,"cost_usd":0.467605,"elapsed_time":80.25,"gpr":1.0,"tpe":1.0,"ee":0.8},{"task_id":"task_032_comprehensive_decision_medium_medium007","category":"comprehensive_decision","score":0.0,"tokens":385568,"cost_usd":1.99804,"elapsed_time":225.56,"gpr":0.375,"tpe":0.9,"ee":0.625},{"task_id":"task_033_comprehensive_decision_medium_medium008","category":"comprehensive_decision","score":1.0,"tokens":266220,"cost_usd":1.4311,"elapsed_time":117.02,"gpr":1.0,"tpe":0.8463,"ee":0.5556},{"task_id":"task_034_comprehensive_decision_medium_medium009","category":"comprehensive_decision","score":1.0,"tokens":114341,"cost_usd":0.603925,"elapsed_time":72.01,"gpr":1.0,"tpe":0.81,"ee":0.625},{"task_id":"task_035_comprehensive_decision_medium_medium010","category":"comprehensive_decision","score":1.0,"tokens":208081,"cost_usd":1.124585,"elapsed_time":178.14,"gpr":1.0,"tpe":0.6597,"ee":0.4167},{"task_id":"task_036_comprehensive_decision_medium_medium011","category":"comprehensive_decision","score":1.0,"tokens":119034,"cost_usd":0.64959,"elapsed_time":66.32,"gpr":1.0,"tpe":0.9,"ee":0.7143},{"task_id":"task_037_comprehensive_decision_medium_medium012","category":"comprehensive_decision","score":1.0,"tokens":446671,"cost_usd":2.358495,"elapsed_time":253.69,"gpr":1.0,"tpe":0.2542,"ee":0.2632},{"task_id":"task_038_comprehensive_decision_medium_medium013","category":"comprehensive_decision","score":0.0,"tokens":208492,"cost_usd":1.12088,"elapsed_time":173.07,"gpr":0.4,"tpe":0.729,"ee":0.4545},{"task_id":"task_039_comprehensive_decision_medium_medium014","category":"comprehensive_decision","score":0.0,"tokens":327261,"cost_usd":1.729525,"elapsed_time":169.86,"gpr":0.5,"tpe":0.4305,"ee":0.3333},{"task_id":"task_040_comprehensive_decision_medium_medium015","category":"comprehensive_decision","score":0.0,"tokens":134201,"cost_usd":0.720825,"elapsed_time":111.37,"gpr":0.25,"tpe":1.0,"ee":0.625},{"task_id":"task_041_comprehensive_decision_medium_medium016","category":"comprehensive_decision","score":0.0,"tokens":5422002,"cost_usd":27.57605,"elapsed_time":1207.22,"gpr":1.0,"tpe":0.3487,"ee":0.0641},{"task_id":"task_042_comprehensive_decision_medium_medium017","category":"comprehensive_decision","score":1.0,"tokens":294012,"cost_usd":1.52766,"elapsed_time":199.99,"gpr":1.0,"tpe":0.412,"ee":0.3333},{"task_id":"task_043_comprehensive_decision_medium_medium018","category":"comprehensive_decision","score":1.0,"tokens":227427,"cost_usd":1.211975,"elapsed_time":167.71,"gpr":0.7143,"tpe":0.6606,"ee":0.4545},{"task_id":"task_044_comprehensive_decision_medium_medium019","category":"comprehensive_decision","score":0.0,"tokens":660864,"cost_usd":3.36486,"elapsed_time":228.21,"gpr":0.3333,"tpe":0.6561,"ee":0.2941},{"task_id":"task_045_comprehensive_decision_medium_medium020","category":"comprehensive_decision","score":1.0,"tokens":215750,"cost_usd":1.15169,"elapsed_time":115.26,"gpr":1.0,"tpe":0.5595,"ee":0.4167},{"task_id":"task_046_comprehensive_decision_medium_medium021","category":"comprehensive_decision","score":0.0,"tokens":332309,"cost_usd":1.735645,"elapsed_time":159.12,"gpr":0.25,"tpe":0.4783,"ee":0.2778},{"task_id":"task_047_comprehensive_decision_medium_medium022","category":"comprehensive_decision","score":1.0,"tokens":264881,"cost_usd":1.378745,"elapsed_time":204.19,"gpr":1.0,"tpe":0.4305,"ee":0.3846},{"task_id":"task_048_comprehensive_decision_medium_medium023","category":"comprehensive_decision","score":1.0,"tokens":124358,"cost_usd":0.64875,"elapsed_time":70.57,"gpr":1.0,"tpe":0.7535,"ee":0.5},{"task_id":"task_049_comprehensive_decision_medium_medium024","category":"comprehensive_decision","score":1.0,"tokens":86745,"cost_usd":0.464845,"elapsed_time":109.94,"gpr":1.0,"tpe":0.9,"ee":0.8333},{"task_id":"task_050_comprehensive_decision_medium_medium025","category":"comprehensive_decision","score":1.0,"tokens":242609,"cost_usd":1.257605,"elapsed_time":103.15,"gpr":1.0,"tpe":0.4783,"ee":0.3846},{"task_id":"task_051_comprehensive_decision_medium_medium026","category":"comprehensive_decision","score":1.0,"tokens":91397,"cost_usd":0.475365,"elapsed_time":166.17,"gpr":1.0,"tpe":0.9,"ee":0.8333},{"task_id":"task_052_comprehensive_decision_medium_medium027","category":"comprehensive_decision","score":1.0,"tokens":71967,"cost_usd":0.394235,"elapsed_time":52.62,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_053_comprehensive_decision_medium_medium028","category":"comprehensive_decision","score":1.0,"tokens":571688,"cost_usd":2.97108,"elapsed_time":289.84,"gpr":1.0,"tpe":0.3628,"ee":0.2083},{"task_id":"task_054_comprehensive_decision_medium_medium029","category":"comprehensive_decision","score":1.0,"tokens":805438,"cost_usd":4.08137,"elapsed_time":187.52,"gpr":1.0,"tpe":0.6381,"ee":0.2778},{"task_id":"task_055_comprehensive_decision_medium_medium030","category":"comprehensive_decision","score":0.0,"tokens":189607,"cost_usd":0.980855,"elapsed_time":71.61,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_056_comprehensive_decision_medium_medium031","category":"comprehensive_decision","score":1.0,"tokens":133543,"cost_usd":0.701075,"elapsed_time":84.3,"gpr":1.0,"tpe":0.9093,"ee":0.5},{"task_id":"task_057_comprehensive_decision_medium_medium032","category":"comprehensive_decision","score":1.0,"tokens":143069,"cost_usd":0.746685,"elapsed_time":76.62,"gpr":0.7143,"tpe":0.92,"ee":0.5556},{"task_id":"task_058_comprehensive_decision_medium_medium033","category":"comprehensive_decision","score":1.0,"tokens":97410,"cost_usd":0.51161,"elapsed_time":54.5,"gpr":1.0,"tpe":0.975,"ee":0.8333},{"task_id":"task_059_comprehensive_decision_medium_medium034","category":"comprehensive_decision","score":0.0,"tokens":1061164,"cost_usd":5.43392,"elapsed_time":361.44,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_060_comprehensive_decision_medium_medium035","category":"comprehensive_decision","score":1.0,"tokens":889351,"cost_usd":4.539555,"elapsed_time":250.09,"gpr":1.0,"tpe":0.4843,"ee":0.1667},{"task_id":"task_061_comprehensive_decision_medium_medium036","category":"comprehensive_decision","score":1.0,"tokens":285934,"cost_usd":1.51371,"elapsed_time":198.2,"gpr":1.0,"tpe":0.4783,"ee":0.3333},{"task_id":"task_062_comprehensive_decision_medium_medium037","category":"comprehensive_decision","score":1.0,"tokens":279454,"cost_usd":1.46661,"elapsed_time":161.19,"gpr":1.0,"tpe":0.6561,"ee":0.3846},{"task_id":"task_063_comprehensive_decision_medium_medium038","category":"comprehensive_decision","score":0.0,"tokens":138752,"cost_usd":0.72384,"elapsed_time":90.01,"gpr":0.3333,"tpe":0.81,"ee":0.5},{"task_id":"task_064_comprehensive_decision_medium_medium039","category":"comprehensive_decision","score":0.0,"tokens":135792,"cost_usd":0.76548,"elapsed_time":135.89,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_065_comprehensive_decision_medium_medium040","category":"comprehensive_decision","score":1.0,"tokens":357738,"cost_usd":1.89091,"elapsed_time":160.33,"gpr":1.0,"tpe":0.4354,"ee":0.3333},{"task_id":"task_066_comprehensive_decision_medium_medium041","category":"comprehensive_decision","score":0.0,"tokens":206810,"cost_usd":1.07297,"elapsed_time":90.31,"gpr":0.0,"tpe":0.0,"ee":0.625},{"task_id":"task_067_comprehensive_decision_medium_medium042","category":"comprehensive_decision","score":0.0,"tokens":1482778,"cost_usd":7.45271,"elapsed_time":138.36,"gpr":0.25,"tpe":1.0,"ee":0.5556},{"task_id":"task_068_comprehensive_decision_medium_medium043","category":"comprehensive_decision","score":1.0,"tokens":140385,"cost_usd":0.736885,"elapsed_time":104.49,"gpr":1.0,"tpe":0.6561,"ee":0.5},{"task_id":"task_069_comprehensive_decision_medium_medium044","category":"comprehensive_decision","score":1.0,"tokens":218294,"cost_usd":1.13719,"elapsed_time":133.79,"gpr":1.0,"tpe":0.541,"ee":0.4167},{"task_id":"task_070_comprehensive_decision_medium_medium045","category":"comprehensive_decision","score":0.0,"tokens":183389,"cost_usd":0.967305,"elapsed_time":101.01,"gpr":0.25,"tpe":0.81,"ee":0.4167},{"task_id":"task_071_enterprise_industry_analysis_easy_easy001","category":"enterprise_industry_analysis","score":0.0,"tokens":1258948,"cost_usd":6.37428,"elapsed_time":381.62,"gpr":0.0,"tpe":0.0,"ee":0.0811},{"task_id":"task_072_enterprise_industry_analysis_easy_easy002","category":"enterprise_industry_analysis","score":0.0,"tokens":219805,"cost_usd":1.130725,"elapsed_time":109.71,"gpr":0.75,"tpe":0.6211,"ee":0.25},{"task_id":"task_073_enterprise_industry_analysis_easy_easy003","category":"enterprise_industry_analysis","score":1.0,"tokens":268562,"cost_usd":1.38775,"elapsed_time":125.57,"gpr":1.0,"tpe":0.3804,"ee":0.2},{"task_id":"task_074_enterprise_industry_analysis_easy_easy004","category":"enterprise_industry_analysis","score":1.0,"tokens":125498,"cost_usd":0.65541,"elapsed_time":78.05,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_075_enterprise_industry_analysis_easy_easy005","category":"enterprise_industry_analysis","score":1.0,"tokens":216089,"cost_usd":1.113665,"elapsed_time":96.34,"gpr":1.0,"tpe":0.4161,"ee":0.25},{"task_id":"task_076_enterprise_industry_analysis_easy_easy006","category":"enterprise_industry_analysis","score":1.0,"tokens":208974,"cost_usd":1.07991,"elapsed_time":92.32,"gpr":1.0,"tpe":0.4624,"ee":0.2727},{"task_id":"task_077_enterprise_industry_analysis_easy_easy007","category":"enterprise_industry_analysis","score":1.0,"tokens":376268,"cost_usd":1.92162,"elapsed_time":177.87,"gpr":1.0,"tpe":0.3371,"ee":0.2143},{"task_id":"task_078_enterprise_industry_analysis_easy_easy008","category":"enterprise_industry_analysis","score":1.0,"tokens":494138,"cost_usd":2.54443,"elapsed_time":313.29,"gpr":1.0,"tpe":0.1791,"ee":0.15},{"task_id":"task_079_enterprise_industry_analysis_easy_easy009","category":"enterprise_industry_analysis","score":1.0,"tokens":87484,"cost_usd":0.45826,"elapsed_time":88.0,"gpr":1.0,"tpe":0.783,"ee":0.5},{"task_id":"task_080_enterprise_industry_analysis_easy_easy010","category":"enterprise_industry_analysis","score":1.0,"tokens":160302,"cost_usd":0.83097,"elapsed_time":72.59,"gpr":1.0,"tpe":0.6342,"ee":0.375},{"task_id":"task_081_enterprise_industry_analysis_easy_easy011","category":"enterprise_industry_analysis","score":1.0,"tokens":203069,"cost_usd":1.052485,"elapsed_time":148.26,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_082_enterprise_industry_analysis_easy_easy012","category":"enterprise_industry_analysis","score":1.0,"tokens":164516,"cost_usd":0.86134,"elapsed_time":94.09,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_083_enterprise_industry_analysis_easy_easy013","category":"enterprise_industry_analysis","score":0.0,"tokens":473040,"cost_usd":2.43016,"elapsed_time":175.55,"gpr":0.3333,"tpe":0.1853,"ee":0.1304},{"task_id":"task_084_enterprise_industry_analysis_easy_easy014","category":"enterprise_industry_analysis","score":0.0,"tokens":263466,"cost_usd":1.35117,"elapsed_time":103.82,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_085_enterprise_industry_analysis_easy_easy015","category":"enterprise_industry_analysis","score":1.0,"tokens":284616,"cost_usd":1.46834,"elapsed_time":129.57,"gpr":1.0,"tpe":0.4161,"ee":0.25},{"task_id":"task_086_enterprise_industry_analysis_easy_easy016","category":"enterprise_industry_analysis","score":1.0,"tokens":244648,"cost_usd":1.2651,"elapsed_time":125.56,"gpr":1.0,"tpe":0.4032,"ee":0.2308},{"task_id":"task_087_enterprise_industry_analysis_easy_easy017","category":"enterprise_industry_analysis","score":0.0,"tokens":163687,"cost_usd":0.843695,"elapsed_time":140.06,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_088_enterprise_industry_analysis_easy_easy018","category":"enterprise_industry_analysis","score":0.0,"tokens":1264056,"cost_usd":6.48678,"elapsed_time":578.84,"gpr":1.0,"tpe":0.0269,"ee":0.075},{"task_id":"task_089_enterprise_industry_analysis_easy_easy019","category":"enterprise_industry_analysis","score":1.0,"tokens":188458,"cost_usd":0.98243,"elapsed_time":144.71,"gpr":1.0,"tpe":0.4464,"ee":0.2727},{"task_id":"task_090_enterprise_industry_analysis_easy_easy020","category":"enterprise_industry_analysis","score":1.0,"tokens":156212,"cost_usd":0.81036,"elapsed_time":87.49,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_091_enterprise_industry_analysis_easy_easy021","category":"enterprise_industry_analysis","score":1.0,"tokens":360594,"cost_usd":1.83901,"elapsed_time":158.41,"gpr":1.0,"tpe":0.2551,"ee":0.1765},{"task_id":"task_092_enterprise_industry_analysis_easy_easy022","category":"enterprise_industry_analysis","score":1.0,"tokens":190481,"cost_usd":0.993005,"elapsed_time":107.91,"gpr":1.0,"tpe":0.4624,"ee":0.2727},{"task_id":"task_093_enterprise_industry_analysis_easy_easy023","category":"enterprise_industry_analysis","score":1.0,"tokens":566055,"cost_usd":2.911115,"elapsed_time":353.38,"gpr":1.0,"tpe":0.1618,"ee":0.1429},{"task_id":"task_094_enterprise_industry_analysis_easy_easy024","category":"enterprise_industry_analysis","score":1.0,"tokens":198416,"cost_usd":1.02052,"elapsed_time":93.29,"gpr":1.0,"tpe":0.5334,"ee":0.3},{"task_id":"task_095_enterprise_industry_analysis_easy_easy025","category":"enterprise_industry_analysis","score":1.0,"tokens":520968,"cost_usd":2.6498,"elapsed_time":148.35,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_096_enterprise_industry_analysis_easy_easy026","category":"enterprise_industry_analysis","score":0.0,"tokens":1234757,"cost_usd":6.224905,"elapsed_time":205.61,"gpr":0.3333,"tpe":0.3874,"ee":0.1667},{"task_id":"task_097_enterprise_industry_analysis_easy_easy027","category":"enterprise_industry_analysis","score":1.0,"tokens":250694,"cost_usd":1.29531,"elapsed_time":103.15,"gpr":1.0,"tpe":0.4624,"ee":0.2727},{"task_id":"task_098_enterprise_industry_analysis_easy_easy028","category":"enterprise_industry_analysis","score":1.0,"tokens":99660,"cost_usd":0.52378,"elapsed_time":82.92,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_099_enterprise_industry_analysis_easy_easy029","category":"enterprise_industry_analysis","score":0.0,"tokens":178275,"cost_usd":0.926095,"elapsed_time":92.46,"gpr":0.6667,"tpe":0.481,"ee":0.2727},{"task_id":"task_100_enterprise_industry_analysis_easy_easy030","category":"enterprise_industry_analysis","score":1.0,"tokens":188243,"cost_usd":0.984215,"elapsed_time":118.27,"gpr":1.0,"tpe":0.3874,"ee":0.25},{"task_id":"task_101_enterprise_industry_analysis_easy_easy031","category":"enterprise_industry_analysis","score":1.0,"tokens":824562,"cost_usd":4.21757,"elapsed_time":314.38,"gpr":1.0,"tpe":0.032,"ee":0.0833},{"task_id":"task_102_enterprise_industry_analysis_easy_easy032","category":"enterprise_industry_analysis","score":0.0,"tokens":1356660,"cost_usd":6.93174,"elapsed_time":476.66,"gpr":0.3333,"tpe":0.3138,"ee":0.0714},{"task_id":"task_103_enterprise_industry_analysis_easy_easy033","category":"enterprise_industry_analysis","score":0.0,"tokens":706369,"cost_usd":3.655905,"elapsed_time":377.09,"gpr":0.3333,"tpe":0.0798,"ee":0.1},{"task_id":"task_104_enterprise_industry_analysis_easy_easy034","category":"enterprise_industry_analysis","score":1.0,"tokens":287655,"cost_usd":1.484695,"elapsed_time":144.5,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_105_enterprise_industry_analysis_easy_easy035","category":"enterprise_industry_analysis","score":0.0,"tokens":84513,"cost_usd":0.440505,"elapsed_time":63.65,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_106_enterprise_industry_analysis_easy_easy036","category":"enterprise_industry_analysis","score":1.0,"tokens":116572,"cost_usd":0.60476,"elapsed_time":77.1,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_107_enterprise_industry_analysis_easy_easy037","category":"enterprise_industry_analysis","score":1.0,"tokens":81387,"cost_usd":0.421975,"elapsed_time":40.5,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_108_enterprise_industry_analysis_easy_easy038","category":"enterprise_industry_analysis","score":0.0,"tokens":182163,"cost_usd":0.933095,"elapsed_time":73.45,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_109_enterprise_industry_analysis_easy_easy039","category":"enterprise_industry_analysis","score":1.0,"tokens":55692,"cost_usd":0.29588,"elapsed_time":36.1,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_110_enterprise_industry_analysis_easy_easy040","category":"enterprise_industry_analysis","score":1.0,"tokens":84840,"cost_usd":0.44244,"elapsed_time":55.5,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_111_enterprise_industry_analysis_easy_easy041","category":"enterprise_industry_analysis","score":0.0,"tokens":434870,"cost_usd":2.20847,"elapsed_time":114.54,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_112_enterprise_industry_analysis_easy_easy042","category":"enterprise_industry_analysis","score":0.0,"tokens":536413,"cost_usd":2.723565,"elapsed_time":138.08,"gpr":0.0,"tpe":0.0,"ee":0.1579},{"task_id":"task_113_enterprise_industry_analysis_easy_easy043","category":"enterprise_industry_analysis","score":1.0,"tokens":250881,"cost_usd":1.286865,"elapsed_time":149.23,"gpr":1.0,"tpe":0.3874,"ee":0.25},{"task_id":"task_114_enterprise_industry_analysis_easy_easy044","category":"enterprise_industry_analysis","score":1.0,"tokens":383121,"cost_usd":1.962445,"elapsed_time":144.25,"gpr":1.0,"tpe":0.1922,"ee":0.1579},{"task_id":"task_115_enterprise_industry_analysis_easy_easy045","category":"enterprise_industry_analysis","score":1.0,"tokens":295143,"cost_usd":1.514035,"elapsed_time":130.97,"gpr":1.0,"tpe":0.496,"ee":0.3},{"task_id":"task_116_enterprise_industry_analysis_easy_easy046","category":"enterprise_industry_analysis","score":0.0,"tokens":336740,"cost_usd":1.75316,"elapsed_time":179.22,"gpr":0.6667,"tpe":0.3487,"ee":0.2},{"task_id":"task_117_enterprise_industry_analysis_easy_easy047","category":"enterprise_industry_analysis","score":1.0,"tokens":87629,"cost_usd":0.455025,"elapsed_time":65.14,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_118_enterprise_industry_analysis_easy_easy048","category":"enterprise_industry_analysis","score":1.0,"tokens":327941,"cost_usd":1.688485,"elapsed_time":130.08,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_119_enterprise_industry_analysis_easy_easy049","category":"enterprise_industry_analysis","score":0.0,"tokens":621755,"cost_usd":3.168535,"elapsed_time":180.4,"gpr":0.0,"tpe":0.0,"ee":0.15},{"task_id":"task_120_enterprise_industry_analysis_easy_easy050","category":"enterprise_industry_analysis","score":1.0,"tokens":488599,"cost_usd":2.497535,"elapsed_time":176.46,"gpr":1.0,"tpe":0.2288,"ee":0.1765},{"task_id":"task_121_enterprise_industry_analysis_easy_easy051","category":"enterprise_industry_analysis","score":1.0,"tokens":69877,"cost_usd":0.362025,"elapsed_time":38.03,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_122_enterprise_industry_analysis_easy_easy052","category":"enterprise_industry_analysis","score":1.0,"tokens":97401,"cost_usd":0.501765,"elapsed_time":54.33,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_123_enterprise_industry_analysis_easy_easy053","category":"enterprise_industry_analysis","score":1.0,"tokens":101164,"cost_usd":0.52602,"elapsed_time":76.01,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_124_enterprise_industry_analysis_easy_easy054","category":"enterprise_industry_analysis","score":1.0,"tokens":225829,"cost_usd":1.160605,"elapsed_time":92.95,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_125_enterprise_industry_analysis_easy_easy055","category":"enterprise_industry_analysis","score":1.0,"tokens":269175,"cost_usd":1.379455,"elapsed_time":93.09,"gpr":1.0,"tpe":0.2824,"ee":0.2},{"task_id":"task_126_enterprise_industry_analysis_easy_easy056","category":"enterprise_industry_analysis","score":1.0,"tokens":166501,"cost_usd":0.863705,"elapsed_time":129.73,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_127_enterprise_industry_analysis_easy_easy057","category":"enterprise_industry_analysis","score":1.0,"tokens":392825,"cost_usd":2.007365,"elapsed_time":97.88,"gpr":1.0,"tpe":0.3874,"ee":0.25},{"task_id":"task_128_enterprise_industry_analysis_easy_easy058","category":"enterprise_industry_analysis","score":1.0,"tokens":304912,"cost_usd":1.56264,"elapsed_time":98.84,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_129_enterprise_industry_analysis_easy_easy059","category":"enterprise_industry_analysis","score":1.0,"tokens":149148,"cost_usd":0.76418,"elapsed_time":53.92,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_130_enterprise_industry_analysis_easy_easy060","category":"enterprise_industry_analysis","score":1.0,"tokens":135052,"cost_usd":0.69174,"elapsed_time":55.07,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_131_enterprise_industry_analysis_easy_easy061","category":"enterprise_industry_analysis","score":0.0,"tokens":576875,"cost_usd":2.912955,"elapsed_time":132.33,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_132_enterprise_industry_analysis_easy_easy062","category":"enterprise_industry_analysis","score":1.0,"tokens":80921,"cost_usd":0.422885,"elapsed_time":50.66,"gpr":1.0,"tpe":0.9033,"ee":0.6},{"task_id":"task_133_enterprise_industry_analysis_easy_easy063","category":"enterprise_industry_analysis","score":1.0,"tokens":104245,"cost_usd":0.538925,"elapsed_time":88.3,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_134_enterprise_industry_analysis_easy_easy064","category":"enterprise_industry_analysis","score":1.0,"tokens":166184,"cost_usd":0.85962,"elapsed_time":74.35,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_135_enterprise_industry_analysis_easy_easy065","category":"enterprise_industry_analysis","score":1.0,"tokens":81878,"cost_usd":0.42881,"elapsed_time":81.38,"gpr":1.0,"tpe":0.9,"ee":0.5},{"task_id":"task_136_enterprise_industry_analysis_easy_easy066","category":"enterprise_industry_analysis","score":1.0,"tokens":271992,"cost_usd":1.40122,"elapsed_time":126.88,"gpr":1.0,"tpe":0.35,"ee":0.2143},{"task_id":"task_137_enterprise_industry_analysis_easy_easy067","category":"enterprise_industry_analysis","score":1.0,"tokens":199164,"cost_usd":1.01804,"elapsed_time":89.96,"gpr":1.0,"tpe":0.6211,"ee":0.3},{"task_id":"task_138_enterprise_industry_analysis_easy_easy068","category":"enterprise_industry_analysis","score":1.0,"tokens":373265,"cost_usd":1.927425,"elapsed_time":181.08,"gpr":1.0,"tpe":0.1668,"ee":0.15},{"task_id":"task_139_enterprise_industry_analysis_easy_easy069","category":"enterprise_industry_analysis","score":0.0,"tokens":159224,"cost_usd":0.84042,"elapsed_time":109.92,"gpr":1.0,"tpe":0.5334,"ee":0.3},{"task_id":"task_140_enterprise_industry_analysis_easy_easy070","category":"enterprise_industry_analysis","score":0.0,"tokens":74115,"cost_usd":0.386615,"elapsed_time":42.71,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_141_enterprise_industry_analysis_easy_easy071","category":"enterprise_industry_analysis","score":1.0,"tokens":185974,"cost_usd":0.96909,"elapsed_time":92.56,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_142_enterprise_industry_analysis_easy_easy072","category":"enterprise_industry_analysis","score":1.0,"tokens":101798,"cost_usd":0.53187,"elapsed_time":67.37,"gpr":1.0,"tpe":0.7587,"ee":0.4286},{"task_id":"task_143_enterprise_industry_analysis_easy_easy073","category":"enterprise_industry_analysis","score":0.0,"tokens":107064,"cost_usd":0.5546,"elapsed_time":60.31,"gpr":0.6667,"tpe":1.0,"ee":0.4286},{"task_id":"task_144_enterprise_industry_analysis_easy_easy074","category":"enterprise_industry_analysis","score":1.0,"tokens":183543,"cost_usd":0.938415,"elapsed_time":59.42,"gpr":1.0,"tpe":0.756,"ee":0.5},{"task_id":"task_145_enterprise_industry_analysis_easy_easy075","category":"enterprise_industry_analysis","score":1.0,"tokens":64778,"cost_usd":0.33091,"elapsed_time":44.88,"gpr":1.0,"tpe":0.87,"ee":0.6},{"task_id":"task_146_enterprise_industry_analysis_easy_easy076","category":"enterprise_industry_analysis","score":1.0,"tokens":53868,"cost_usd":0.2756,"elapsed_time":32.51,"gpr":1.0,"tpe":0.9667,"ee":0.75},{"task_id":"task_147_enterprise_industry_analysis_easy_easy077","category":"enterprise_industry_analysis","score":1.0,"tokens":53976,"cost_usd":0.2785,"elapsed_time":28.89,"gpr":1.0,"tpe":0.9667,"ee":0.75},{"task_id":"task_148_enterprise_industry_analysis_easy_easy078","category":"enterprise_industry_analysis","score":1.0,"tokens":64769,"cost_usd":0.330205,"elapsed_time":47.14,"gpr":1.0,"tpe":0.87,"ee":0.6},{"task_id":"task_149_enterprise_industry_analysis_easy_easy079","category":"enterprise_industry_analysis","score":1.0,"tokens":65019,"cost_usd":0.333355,"elapsed_time":89.68,"gpr":1.0,"tpe":0.87,"ee":0.6},{"task_id":"task_150_enterprise_industry_analysis_easy_easy080","category":"enterprise_industry_analysis","score":1.0,"tokens":64795,"cost_usd":0.330575,"elapsed_time":32.94,"gpr":1.0,"tpe":0.87,"ee":0.6},{"task_id":"task_151_enterprise_industry_analysis_easy_easy081","category":"enterprise_industry_analysis","score":1.0,"tokens":56118,"cost_usd":0.29421,"elapsed_time":37.43,"gpr":1.0,"tpe":0.9667,"ee":0.75},{"task_id":"task_152_enterprise_industry_analysis_easy_easy082","category":"enterprise_industry_analysis","score":1.0,"tokens":67632,"cost_usd":0.35222,"elapsed_time":38.62,"gpr":1.0,"tpe":0.87,"ee":0.6},{"task_id":"task_153_enterprise_industry_analysis_easy_easy083","category":"enterprise_industry_analysis","score":1.0,"tokens":105015,"cost_usd":0.538935,"elapsed_time":52.02,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_154_enterprise_industry_analysis_easy_easy084","category":"enterprise_industry_analysis","score":1.0,"tokens":179588,"cost_usd":0.93534,"elapsed_time":87.97,"gpr":1.0,"tpe":0.5283,"ee":0.25},{"task_id":"task_155_enterprise_industry_analysis_easy_easy085","category":"enterprise_industry_analysis","score":1.0,"tokens":86595,"cost_usd":0.445335,"elapsed_time":74.88,"gpr":1.0,"tpe":0.813,"ee":0.5},{"task_id":"task_156_enterprise_industry_analysis_easy_easy086","category":"enterprise_industry_analysis","score":1.0,"tokens":106491,"cost_usd":0.547755,"elapsed_time":69.61,"gpr":1.0,"tpe":0.7047,"ee":0.4286},{"task_id":"task_157_enterprise_industry_analysis_easy_easy088","category":"enterprise_industry_analysis","score":1.0,"tokens":285745,"cost_usd":1.438385,"elapsed_time":41.58,"gpr":1.0,"tpe":0.81,"ee":0.5},{"task_id":"task_158_enterprise_industry_analysis_easy_easy090","category":"enterprise_industry_analysis","score":0.0,"tokens":189827,"cost_usd":0.973675,"elapsed_time":112.42,"gpr":1.0,"tpe":0.315,"ee":0.2},{"task_id":"task_159_enterprise_industry_analysis_easy_easy091","category":"enterprise_industry_analysis","score":0.0,"tokens":266096,"cost_usd":1.34672,"elapsed_time":91.49,"gpr":0.6667,"tpe":0.6561,"ee":0.375},{"task_id":"task_160_enterprise_industry_analysis_easy_easy092","category":"enterprise_industry_analysis","score":0.0,"tokens":102804,"cost_usd":0.53014,"elapsed_time":65.95,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_161_enterprise_industry_analysis_easy_easy093","category":"enterprise_industry_analysis","score":1.0,"tokens":84726,"cost_usd":0.43785,"elapsed_time":72.69,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_162_enterprise_industry_analysis_easy_easy094","category":"enterprise_industry_analysis","score":1.0,"tokens":186433,"cost_usd":0.953065,"elapsed_time":163.9,"gpr":1.0,"tpe":0.448,"ee":0.25},{"task_id":"task_163_enterprise_industry_analysis_easy_easy095","category":"enterprise_industry_analysis","score":1.0,"tokens":83479,"cost_usd":0.426255,"elapsed_time":95.66,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_164_enterprise_industry_analysis_easy_easy096","category":"enterprise_industry_analysis","score":1.0,"tokens":68769,"cost_usd":0.358145,"elapsed_time":81.05,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_165_enterprise_industry_analysis_easy_easy097","category":"enterprise_industry_analysis","score":0.0,"tokens":133706,"cost_usd":0.68127,"elapsed_time":71.84,"gpr":0.3333,"tpe":0.729,"ee":0.5},{"task_id":"task_166_enterprise_industry_analysis_easy_easy098","category":"enterprise_industry_analysis","score":1.0,"tokens":1067647,"cost_usd":5.360615,"elapsed_time":146.11,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_167_enterprise_industry_analysis_easy_easy099","category":"enterprise_industry_analysis","score":1.0,"tokens":51449,"cost_usd":0.265865,"elapsed_time":67.69,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_168_enterprise_industry_analysis_easy_easy100","category":"enterprise_industry_analysis","score":1.0,"tokens":118733,"cost_usd":0.606005,"elapsed_time":60.1,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_169_enterprise_industry_analysis_easy_easy101","category":"enterprise_industry_analysis","score":1.0,"tokens":38105,"cost_usd":0.194925,"elapsed_time":55.87,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_170_enterprise_industry_analysis_easy_easy102","category":"enterprise_industry_analysis","score":1.0,"tokens":207232,"cost_usd":1.06072,"elapsed_time":95.21,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_171_enterprise_industry_analysis_easy_easy103","category":"enterprise_industry_analysis","score":1.0,"tokens":105095,"cost_usd":0.541315,"elapsed_time":82.72,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_172_enterprise_industry_analysis_easy_easy104","category":"enterprise_industry_analysis","score":1.0,"tokens":109994,"cost_usd":0.56057,"elapsed_time":71.43,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_173_enterprise_industry_analysis_easy_easy105","category":"enterprise_industry_analysis","score":1.0,"tokens":107773,"cost_usd":0.556245,"elapsed_time":156.49,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_174_enterprise_industry_analysis_easy_easy106","category":"enterprise_industry_analysis","score":1.0,"tokens":77224,"cost_usd":0.40276,"elapsed_time":50.17,"gpr":1.0,"tpe":0.786,"ee":0.5},{"task_id":"task_175_enterprise_industry_analysis_easy_easy107","category":"enterprise_industry_analysis","score":1.0,"tokens":51908,"cost_usd":0.2671,"elapsed_time":33.28,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_176_enterprise_industry_analysis_easy_easy108","category":"enterprise_industry_analysis","score":1.0,"tokens":59019,"cost_usd":0.303075,"elapsed_time":57.8,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_177_enterprise_industry_analysis_easy_easy109","category":"enterprise_industry_analysis","score":0.0,"tokens":66099,"cost_usd":0.339215,"elapsed_time":63.88,"gpr":0.6667,"tpe":0.729,"ee":0.4286},{"task_id":"task_178_enterprise_industry_analysis_easy_easy110","category":"enterprise_industry_analysis","score":1.0,"tokens":62754,"cost_usd":0.32293,"elapsed_time":39.2,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_179_enterprise_industry_analysis_easy_easy111","category":"enterprise_industry_analysis","score":1.0,"tokens":43666,"cost_usd":0.22369,"elapsed_time":32.33,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_180_enterprise_industry_analysis_easy_easy112","category":"enterprise_industry_analysis","score":1.0,"tokens":124319,"cost_usd":0.638775,"elapsed_time":56.47,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_181_enterprise_industry_analysis_easy_easy113","category":"enterprise_industry_analysis","score":1.0,"tokens":107729,"cost_usd":0.556685,"elapsed_time":60.3,"gpr":0.6667,"tpe":0.6925,"ee":0.4286},{"task_id":"task_182_enterprise_industry_analysis_easy_easy114","category":"enterprise_industry_analysis","score":0.0,"tokens":129860,"cost_usd":0.66736,"elapsed_time":58.25,"gpr":0.6667,"tpe":0.729,"ee":0.375},{"task_id":"task_183_enterprise_industry_analysis_easy_easy115","category":"enterprise_industry_analysis","score":0.0,"tokens":139499,"cost_usd":0.719235,"elapsed_time":50.92,"gpr":0.6667,"tpe":1.0,"ee":0.6},{"task_id":"task_184_enterprise_industry_analysis_easy_easy116","category":"enterprise_industry_analysis","score":1.0,"tokens":128340,"cost_usd":0.66046,"elapsed_time":65.13,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_185_enterprise_industry_analysis_easy_easy117","category":"enterprise_industry_analysis","score":1.0,"tokens":72315,"cost_usd":0.371295,"elapsed_time":42.6,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_186_enterprise_industry_analysis_medium_medium001","category":"enterprise_industry_analysis","score":1.0,"tokens":257827,"cost_usd":1.325935,"elapsed_time":104.08,"gpr":1.0,"tpe":0.6677,"ee":0.2857},{"task_id":"task_187_enterprise_industry_analysis_medium_medium002","category":"enterprise_industry_analysis","score":1.0,"tokens":100626,"cost_usd":0.52111,"elapsed_time":96.31,"gpr":1.0,"tpe":0.9275,"ee":0.6667},{"task_id":"task_188_enterprise_industry_analysis_medium_medium003","category":"enterprise_industry_analysis","score":0.0,"tokens":286303,"cost_usd":1.472855,"elapsed_time":151.28,"gpr":0.75,"tpe":0.5201,"ee":0.2353},{"task_id":"task_189_enterprise_industry_analysis_medium_medium004","category":"enterprise_industry_analysis","score":1.0,"tokens":222680,"cost_usd":1.15338,"elapsed_time":126.9,"gpr":1.0,"tpe":0.6322,"ee":0.3077},{"task_id":"task_190_enterprise_industry_analysis_medium_medium005","category":"enterprise_industry_analysis","score":1.0,"tokens":287421,"cost_usd":1.473105,"elapsed_time":206.26,"gpr":1.0,"tpe":0.4693,"ee":0.25},{"task_id":"task_191_enterprise_industry_analysis_medium_medium006","category":"enterprise_industry_analysis","score":1.0,"tokens":233751,"cost_usd":1.207455,"elapsed_time":105.46,"gpr":1.0,"tpe":0.5818,"ee":0.3333},{"task_id":"task_192_enterprise_industry_analysis_medium_medium007","category":"enterprise_industry_analysis","score":0.0,"tokens":223923,"cost_usd":1.152675,"elapsed_time":143.19,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_193_enterprise_industry_analysis_medium_medium008","category":"enterprise_industry_analysis","score":1.0,"tokens":168625,"cost_usd":0.870665,"elapsed_time":74.25,"gpr":1.0,"tpe":0.7537,"ee":0.5},{"task_id":"task_194_enterprise_industry_analysis_medium_medium009","category":"enterprise_industry_analysis","score":1.0,"tokens":226051,"cost_usd":1.161235,"elapsed_time":104.75,"gpr":1.0,"tpe":0.4899,"ee":0.3077},{"task_id":"task_195_enterprise_industry_analysis_medium_medium010","category":"enterprise_industry_analysis","score":0.0,"tokens":73701,"cost_usd":0.388005,"elapsed_time":40.11,"gpr":0.4,"tpe":0.95,"ee":0.8},{"task_id":"task_196_enterprise_industry_analysis_medium_medium011","category":"enterprise_industry_analysis","score":1.0,"tokens":535905,"cost_usd":2.763045,"elapsed_time":267.6,"gpr":0.4,"tpe":0.1216,"ee":0.1481},{"task_id":"task_197_enterprise_industry_analysis_medium_medium012","category":"enterprise_industry_analysis","score":0.0,"tokens":615643,"cost_usd":3.178975,"elapsed_time":297.63,"gpr":0.0,"tpe":0.0,"ee":0.1333},{"task_id":"task_198_enterprise_industry_analysis_medium_medium013","category":"enterprise_industry_analysis","score":0.0,"tokens":1689710,"cost_usd":8.62283,"elapsed_time":586.4,"gpr":0.5,"tpe":0.3874,"ee":0.1111},{"task_id":"task_199_enterprise_industry_analysis_medium_medium014","category":"enterprise_industry_analysis","score":1.0,"tokens":205117,"cost_usd":1.058425,"elapsed_time":86.79,"gpr":1.0,"tpe":0.8097,"ee":0.5},{"task_id":"task_200_enterprise_industry_analysis_medium_medium015","category":"enterprise_industry_analysis","score":1.0,"tokens":161449,"cost_usd":0.855265,"elapsed_time":157.14,"gpr":1.0,"tpe":0.7143,"ee":0.5},{"task_id":"task_201_enterprise_industry_analysis_medium_medium016","category":"enterprise_industry_analysis","score":1.0,"tokens":269467,"cost_usd":1.394775,"elapsed_time":125.07,"gpr":1.0,"tpe":0.6069,"ee":0.3333},{"task_id":"task_202_enterprise_industry_analysis_medium_medium017","category":"enterprise_industry_analysis","score":1.0,"tokens":209867,"cost_usd":1.095715,"elapsed_time":88.62,"gpr":1.0,"tpe":0.7274,"ee":0.5},{"task_id":"task_203_enterprise_industry_analysis_medium_medium018","category":"enterprise_industry_analysis","score":1.0,"tokens":247120,"cost_usd":1.27882,"elapsed_time":155.45,"gpr":1.0,"tpe":0.5399,"ee":0.3846},{"task_id":"task_204_enterprise_industry_analysis_medium_medium019","category":"enterprise_industry_analysis","score":1.0,"tokens":196041,"cost_usd":1.025125,"elapsed_time":102.35,"gpr":1.0,"tpe":0.5232,"ee":0.4167},{"task_id":"task_205_enterprise_industry_analysis_medium_medium020","category":"enterprise_industry_analysis","score":1.0,"tokens":163131,"cost_usd":0.845655,"elapsed_time":76.32,"gpr":1.0,"tpe":0.7792,"ee":0.5556},{"task_id":"task_206_enterprise_industry_analysis_medium_medium021","category":"enterprise_industry_analysis","score":0.0,"tokens":321034,"cost_usd":1.67963,"elapsed_time":175.03,"gpr":0.8,"tpe":0.5438,"ee":0.3571},{"task_id":"task_207_enterprise_industry_analysis_medium_medium022","category":"enterprise_industry_analysis","score":1.0,"tokens":670857,"cost_usd":3.439465,"elapsed_time":265.93,"gpr":1.0,"tpe":0.118,"ee":0.1786},{"task_id":"task_208_enterprise_industry_analysis_medium_medium023","category":"enterprise_industry_analysis","score":1.0,"tokens":320463,"cost_usd":1.653775,"elapsed_time":157.12,"gpr":1.0,"tpe":0.414,"ee":0.2941},{"task_id":"task_209_enterprise_industry_analysis_medium_medium024","category":"enterprise_industry_analysis","score":1.0,"tokens":224311,"cost_usd":1.152875,"elapsed_time":102.89,"gpr":1.0,"tpe":0.839,"ee":0.5556},{"task_id":"task_210_enterprise_industry_analysis_medium_medium025","category":"enterprise_industry_analysis","score":0.0,"tokens":1584830,"cost_usd":8.03685,"elapsed_time":274.79,"gpr":0.6,"tpe":0.5425,"ee":0.1724},{"task_id":"task_211_enterprise_industry_analysis_medium_medium026","category":"enterprise_industry_analysis","score":1.0,"tokens":151314,"cost_usd":0.78931,"elapsed_time":87.5,"gpr":1.0,"tpe":0.8498,"ee":0.625},{"task_id":"task_212_enterprise_industry_analysis_medium_medium027","category":"enterprise_industry_analysis","score":1.0,"tokens":199288,"cost_usd":1.03156,"elapsed_time":95.68,"gpr":1.0,"tpe":0.7425,"ee":0.4545},{"task_id":"task_213_enterprise_industry_analysis_medium_medium028","category":"enterprise_industry_analysis","score":1.0,"tokens":88755,"cost_usd":0.463135,"elapsed_time":56.68,"gpr":1.0,"tpe":0.98,"ee":0.8333},{"task_id":"task_214_enterprise_industry_analysis_medium_medium029","category":"enterprise_industry_analysis","score":1.0,"tokens":137701,"cost_usd":0.715865,"elapsed_time":84.93,"gpr":1.0,"tpe":0.8698,"ee":0.625},{"task_id":"task_215_enterprise_industry_analysis_medium_medium030","category":"enterprise_industry_analysis","score":1.0,"tokens":185513,"cost_usd":0.968365,"elapsed_time":185.08,"gpr":1.0,"tpe":0.6927,"ee":0.4545},{"task_id":"task_216_enterprise_industry_analysis_medium_medium031","category":"enterprise_industry_analysis","score":1.0,"tokens":221598,"cost_usd":1.14923,"elapsed_time":112.63,"gpr":1.0,"tpe":0.6702,"ee":0.4167},{"task_id":"task_217_enterprise_industry_analysis_medium_medium032","category":"enterprise_industry_analysis","score":1.0,"tokens":250091,"cost_usd":1.291055,"elapsed_time":105.55,"gpr":1.0,"tpe":0.6603,"ee":0.4545},{"task_id":"task_218_enterprise_industry_analysis_medium_medium033","category":"enterprise_industry_analysis","score":0.0,"tokens":88416,"cost_usd":0.46176,"elapsed_time":69.53,"gpr":0.8,"tpe":1.0,"ee":0.8333},{"task_id":"task_219_enterprise_industry_analysis_medium_medium034","category":"enterprise_industry_analysis","score":1.0,"tokens":72714,"cost_usd":0.38245,"elapsed_time":57.84,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_220_enterprise_industry_analysis_medium_medium035","category":"enterprise_industry_analysis","score":1.0,"tokens":215117,"cost_usd":1.133985,"elapsed_time":121.44,"gpr":1.0,"tpe":0.5444,"ee":0.4167},{"task_id":"task_221_enterprise_industry_analysis_medium_medium036","category":"enterprise_industry_analysis","score":1.0,"tokens":188132,"cost_usd":0.99418,"elapsed_time":97.26,"gpr":1.0,"tpe":0.7405,"ee":0.5},{"task_id":"task_222_enterprise_industry_analysis_medium_medium037","category":"enterprise_industry_analysis","score":1.0,"tokens":108452,"cost_usd":0.56804,"elapsed_time":89.22,"gpr":1.0,"tpe":0.962,"ee":0.7143},{"task_id":"task_223_enterprise_industry_analysis_medium_medium038","category":"enterprise_industry_analysis","score":0.0,"tokens":164634,"cost_usd":0.86493,"elapsed_time":181.12,"gpr":0.8,"tpe":0.8645,"ee":0.5},{"task_id":"task_224_enterprise_industry_analysis_medium_medium039","category":"enterprise_industry_analysis","score":1.0,"tokens":280480,"cost_usd":1.46034,"elapsed_time":172.99,"gpr":1.0,"tpe":0.4226,"ee":0.3125},{"task_id":"task_225_enterprise_industry_analysis_medium_medium040","category":"enterprise_industry_analysis","score":1.0,"tokens":162475,"cost_usd":0.847735,"elapsed_time":73.46,"gpr":1.0,"tpe":0.6883,"ee":0.5},{"task_id":"task_226_enterprise_industry_analysis_medium_medium041","category":"enterprise_industry_analysis","score":1.0,"tokens":159995,"cost_usd":0.833895,"elapsed_time":116.38,"gpr":1.0,"tpe":0.7555,"ee":0.5},{"task_id":"task_227_enterprise_industry_analysis_medium_medium042","category":"enterprise_industry_analysis","score":0.0,"tokens":315353,"cost_usd":1.658445,"elapsed_time":211.99,"gpr":0.8,"tpe":0.6441,"ee":0.3333},{"task_id":"task_228_enterprise_industry_analysis_medium_medium043","category":"enterprise_industry_analysis","score":1.0,"tokens":132792,"cost_usd":0.68728,"elapsed_time":118.3,"gpr":1.0,"tpe":0.8512,"ee":0.5556},{"task_id":"task_229_enterprise_industry_analysis_medium_medium044","category":"enterprise_industry_analysis","score":1.0,"tokens":139816,"cost_usd":0.72708,"elapsed_time":117.09,"gpr":1.0,"tpe":0.942,"ee":0.7143},{"task_id":"task_230_enterprise_industry_analysis_medium_medium045","category":"enterprise_industry_analysis","score":0.0,"tokens":1077316,"cost_usd":5.48192,"elapsed_time":298.64,"gpr":0.6,"tpe":0.3575,"ee":0.1316},{"task_id":"task_231_enterprise_industry_analysis_medium_medium046","category":"enterprise_industry_analysis","score":1.0,"tokens":186064,"cost_usd":0.99442,"elapsed_time":130.94,"gpr":1.0,"tpe":0.7343,"ee":0.5},{"task_id":"task_232_enterprise_industry_analysis_medium_medium047","category":"enterprise_industry_analysis","score":1.0,"tokens":171296,"cost_usd":0.90094,"elapsed_time":159.78,"gpr":1.0,"tpe":0.8082,"ee":0.5556},{"task_id":"task_233_enterprise_industry_analysis_medium_medium048","category":"enterprise_industry_analysis","score":1.0,"tokens":150769,"cost_usd":0.787005,"elapsed_time":71.46,"gpr":1.0,"tpe":0.8118,"ee":0.625},{"task_id":"task_234_enterprise_industry_analysis_medium_medium049","category":"enterprise_industry_analysis","score":1.0,"tokens":159920,"cost_usd":0.84284,"elapsed_time":79.52,"gpr":1.0,"tpe":0.8244,"ee":0.5556},{"task_id":"task_235_enterprise_industry_analysis_medium_medium050","category":"enterprise_industry_analysis","score":1.0,"tokens":310604,"cost_usd":1.61662,"elapsed_time":132.85,"gpr":1.0,"tpe":0.5477,"ee":0.3125},{"task_id":"task_236_enterprise_industry_analysis_medium_medium051","category":"enterprise_industry_analysis","score":0.0,"tokens":191216,"cost_usd":1.00518,"elapsed_time":122.39,"gpr":0.6,"tpe":0.7047,"ee":0.4167},{"task_id":"task_237_enterprise_industry_analysis_medium_medium052","category":"enterprise_industry_analysis","score":1.0,"tokens":216765,"cost_usd":1.126645,"elapsed_time":101.75,"gpr":1.0,"tpe":0.5809,"ee":0.3571},{"task_id":"task_238_enterprise_industry_analysis_medium_medium053","category":"enterprise_industry_analysis","score":1.0,"tokens":153035,"cost_usd":0.806555,"elapsed_time":82.37,"gpr":1.0,"tpe":0.6721,"ee":0.5},{"task_id":"task_239_enterprise_industry_analysis_medium_medium054","category":"enterprise_industry_analysis","score":1.0,"tokens":141314,"cost_usd":0.73753,"elapsed_time":77.83,"gpr":1.0,"tpe":0.839,"ee":0.5556},{"task_id":"task_240_enterprise_industry_analysis_medium_medium055","category":"enterprise_industry_analysis","score":1.0,"tokens":182331,"cost_usd":0.954495,"elapsed_time":105.94,"gpr":1.0,"tpe":0.7025,"ee":0.4545},{"task_id":"task_241_enterprise_industry_analysis_medium_medium056","category":"enterprise_industry_analysis","score":1.0,"tokens":244095,"cost_usd":1.263215,"elapsed_time":737.08,"gpr":1.0,"tpe":0.4532,"ee":0.3333},{"task_id":"task_242_enterprise_industry_analysis_medium_medium057","category":"enterprise_industry_analysis","score":1.0,"tokens":674931,"cost_usd":3.437655,"elapsed_time":171.7,"gpr":1.0,"tpe":0.4503,"ee":0.2941},{"task_id":"task_243_enterprise_industry_analysis_medium_medium058","category":"enterprise_industry_analysis","score":0.0,"tokens":192559,"cost_usd":1.006695,"elapsed_time":103.68,"gpr":1.0,"tpe":0.7487,"ee":0.4545},{"task_id":"task_244_enterprise_industry_analysis_medium_medium059","category":"enterprise_industry_analysis","score":1.0,"tokens":369964,"cost_usd":1.89856,"elapsed_time":174.98,"gpr":1.0,"tpe":0.3089,"ee":0.2941},{"task_id":"task_245_enterprise_industry_analysis_medium_medium060","category":"enterprise_industry_analysis","score":1.0,"tokens":288163,"cost_usd":1.484475,"elapsed_time":134.22,"gpr":1.0,"tpe":0.5928,"ee":0.3571},{"task_id":"task_246_enterprise_industry_analysis_medium_medium061","category":"enterprise_industry_analysis","score":1.0,"tokens":117747,"cost_usd":0.617755,"elapsed_time":127.15,"gpr":1.0,"tpe":0.8698,"ee":0.625},{"task_id":"task_247_enterprise_industry_analysis_medium_medium062","category":"enterprise_industry_analysis","score":1.0,"tokens":151252,"cost_usd":0.78786,"elapsed_time":137.3,"gpr":1.0,"tpe":0.7828,"ee":0.5556},{"task_id":"task_248_enterprise_industry_analysis_medium_medium063","category":"enterprise_industry_analysis","score":1.0,"tokens":162239,"cost_usd":0.844215,"elapsed_time":157.88,"gpr":1.0,"tpe":0.6634,"ee":0.4},{"task_id":"task_249_enterprise_industry_analysis_medium_medium064","category":"enterprise_industry_analysis","score":1.0,"tokens":317704,"cost_usd":1.62498,"elapsed_time":148.21,"gpr":1.0,"tpe":0.3173,"ee":0.2353},{"task_id":"task_250_enterprise_industry_analysis_medium_medium065","category":"enterprise_industry_analysis","score":1.0,"tokens":208730,"cost_usd":1.07003,"elapsed_time":807.33,"gpr":1.0,"tpe":0.6618,"ee":0.4444},{"task_id":"task_251_enterprise_industry_analysis_medium_medium066","category":"enterprise_industry_analysis","score":1.0,"tokens":261189,"cost_usd":1.332825,"elapsed_time":111.31,"gpr":1.0,"tpe":0.4382,"ee":0.3077},{"task_id":"task_252_enterprise_industry_analysis_medium_medium067","category":"enterprise_industry_analysis","score":1.0,"tokens":197872,"cost_usd":1.02442,"elapsed_time":133.47,"gpr":1.0,"tpe":0.4944,"ee":0.3333},{"task_id":"task_253_enterprise_industry_analysis_medium_medium068","category":"enterprise_industry_analysis","score":0.0,"tokens":234438,"cost_usd":1.20869,"elapsed_time":133.94,"gpr":0.75,"tpe":0.795,"ee":0.4444},{"task_id":"task_254_enterprise_industry_analysis_medium_medium069","category":"enterprise_industry_analysis","score":1.0,"tokens":134149,"cost_usd":0.709965,"elapsed_time":79.18,"gpr":1.0,"tpe":0.7603,"ee":0.5},{"task_id":"task_255_enterprise_industry_analysis_medium_medium070","category":"enterprise_industry_analysis","score":0.0,"tokens":204041,"cost_usd":1.059425,"elapsed_time":113.8,"gpr":0.3333,"tpe":0.5905,"ee":0.4},{"task_id":"task_256_enterprise_industry_analysis_medium_medium071","category":"enterprise_industry_analysis","score":0.0,"tokens":59092,"cost_usd":0.30158,"elapsed_time":644.02,"gpr":0.0,"tpe":0.0,"ee":0.8},{"task_id":"task_257_enterprise_industry_analysis_medium_medium072","category":"enterprise_industry_analysis","score":1.0,"tokens":190187,"cost_usd":0.973015,"elapsed_time":144.15,"gpr":1.0,"tpe":0.5388,"ee":0.3333},{"task_id":"task_258_enterprise_industry_analysis_medium_medium073","category":"enterprise_industry_analysis","score":1.0,"tokens":128392,"cost_usd":0.6607,"elapsed_time":62.83,"gpr":1.0,"tpe":0.7513,"ee":0.5},{"task_id":"task_259_enterprise_industry_analysis_medium_medium074","category":"enterprise_industry_analysis","score":1.0,"tokens":191581,"cost_usd":0.988005,"elapsed_time":152.45,"gpr":1.0,"tpe":0.4424,"ee":0.3077},{"task_id":"task_260_enterprise_industry_analysis_medium_medium075","category":"enterprise_industry_analysis","score":1.0,"tokens":202945,"cost_usd":1.044445,"elapsed_time":88.0,"gpr":1.0,"tpe":0.5493,"ee":0.3636},{"task_id":"task_261_enterprise_industry_analysis_medium_medium076","category":"enterprise_industry_analysis","score":0.0,"tokens":288792,"cost_usd":1.4967,"elapsed_time":123.64,"gpr":0.75,"tpe":0.5583,"ee":0.2857},{"task_id":"task_262_enterprise_industry_analysis_medium_medium077","category":"enterprise_industry_analysis","score":1.0,"tokens":284051,"cost_usd":1.459195,"elapsed_time":151.91,"gpr":1.0,"tpe":0.4414,"ee":0.25},{"task_id":"task_263_enterprise_industry_analysis_medium_medium078","category":"enterprise_industry_analysis","score":0.0,"tokens":144630,"cost_usd":0.74519,"elapsed_time":61.88,"gpr":1.0,"tpe":0.7353,"ee":0.5},{"task_id":"task_264_enterprise_industry_analysis_medium_medium079","category":"enterprise_industry_analysis","score":1.0,"tokens":302456,"cost_usd":1.56954,"elapsed_time":796.18,"gpr":1.0,"tpe":0.574,"ee":0.3077},{"task_id":"task_265_enterprise_industry_analysis_medium_medium080","category":"enterprise_industry_analysis","score":0.0,"tokens":276698,"cost_usd":1.42205,"elapsed_time":108.94,"gpr":0.25,"tpe":0.3874,"ee":0.3077},{"task_id":"task_266_enterprise_industry_analysis_medium_medium081","category":"enterprise_industry_analysis","score":1.0,"tokens":407828,"cost_usd":2.09448,"elapsed_time":203.88,"gpr":1.0,"tpe":0.5452,"ee":0.2353},{"task_id":"task_267_enterprise_industry_analysis_medium_medium082","category":"enterprise_industry_analysis","score":0.0,"tokens":187975,"cost_usd":0.969275,"elapsed_time":89.79,"gpr":1.0,"tpe":0.5626,"ee":0.4},{"task_id":"task_268_enterprise_industry_analysis_medium_medium083","category":"enterprise_industry_analysis","score":1.0,"tokens":425629,"cost_usd":2.195265,"elapsed_time":203.02,"gpr":1.0,"tpe":0.3985,"ee":0.2222},{"task_id":"task_269_enterprise_industry_analysis_medium_medium084","category":"enterprise_industry_analysis","score":1.0,"tokens":283587,"cost_usd":1.460055,"elapsed_time":130.1,"gpr":1.0,"tpe":0.549,"ee":0.3077},{"task_id":"task_270_enterprise_industry_analysis_medium_medium085","category":"enterprise_industry_analysis","score":1.0,"tokens":158844,"cost_usd":0.82062,"elapsed_time":82.99,"gpr":1.0,"tpe":0.7785,"ee":0.5},{"task_id":"task_271_enterprise_industry_analysis_medium_medium086","category":"enterprise_industry_analysis","score":1.0,"tokens":213789,"cost_usd":1.100685,"elapsed_time":119.74,"gpr":1.0,"tpe":0.5758,"ee":0.3333},{"task_id":"task_272_enterprise_industry_analysis_medium_medium087","category":"enterprise_industry_analysis","score":1.0,"tokens":296105,"cost_usd":1.517525,"elapsed_time":103.59,"gpr":1.0,"tpe":0.6158,"ee":0.4},{"task_id":"task_273_enterprise_industry_analysis_medium_medium088","category":"enterprise_industry_analysis","score":1.0,"tokens":256417,"cost_usd":1.327725,"elapsed_time":120.78,"gpr":1.0,"tpe":0.6043,"ee":0.3333},{"task_id":"task_274_enterprise_industry_analysis_medium_medium089","category":"enterprise_industry_analysis","score":1.0,"tokens":562036,"cost_usd":2.88526,"elapsed_time":308.23,"gpr":1.0,"tpe":0.3356,"ee":0.1905},{"task_id":"task_275_enterprise_industry_analysis_medium_medium090","category":"enterprise_industry_analysis","score":1.0,"tokens":312809,"cost_usd":1.603045,"elapsed_time":192.15,"gpr":1.0,"tpe":0.6103,"ee":0.3333},{"task_id":"task_276_enterprise_industry_analysis_medium_medium091","category":"enterprise_industry_analysis","score":1.0,"tokens":186461,"cost_usd":0.961285,"elapsed_time":158.48,"gpr":1.0,"tpe":0.7109,"ee":0.4},{"task_id":"task_277_enterprise_industry_analysis_medium_medium092","category":"enterprise_industry_analysis","score":1.0,"tokens":160892,"cost_usd":0.83578,"elapsed_time":163.77,"gpr":1.0,"tpe":0.7785,"ee":0.5},{"task_id":"task_278_enterprise_industry_analysis_medium_medium093","category":"enterprise_industry_analysis","score":1.0,"tokens":127065,"cost_usd":0.659785,"elapsed_time":105.05,"gpr":1.0,"tpe":0.8373,"ee":0.5714},{"task_id":"task_279_enterprise_industry_analysis_medium_medium094","category":"enterprise_industry_analysis","score":1.0,"tokens":164289,"cost_usd":0.851545,"elapsed_time":110.28,"gpr":1.0,"tpe":0.7493,"ee":0.5},{"task_id":"task_280_enterprise_industry_analysis_medium_medium095","category":"enterprise_industry_analysis","score":1.0,"tokens":127079,"cost_usd":0.658135,"elapsed_time":96.75,"gpr":1.0,"tpe":0.8123,"ee":0.5714},{"task_id":"task_281_enterprise_industry_analysis_medium_medium096","category":"enterprise_industry_analysis","score":0.0,"tokens":207798,"cost_usd":1.06651,"elapsed_time":79.39,"gpr":1.0,"tpe":0.7898,"ee":0.5714},{"task_id":"task_282_enterprise_industry_analysis_medium_medium097","category":"enterprise_industry_analysis","score":1.0,"tokens":124468,"cost_usd":0.642,"elapsed_time":80.56,"gpr":1.0,"tpe":0.8123,"ee":0.5714},{"task_id":"task_283_enterprise_industry_analysis_medium_medium098","category":"enterprise_industry_analysis","score":1.0,"tokens":196035,"cost_usd":1.008715,"elapsed_time":82.73,"gpr":1.0,"tpe":0.5921,"ee":0.4},{"task_id":"task_284_enterprise_industry_analysis_medium_medium099","category":"enterprise_industry_analysis","score":0.0,"tokens":278669,"cost_usd":1.440145,"elapsed_time":177.36,"gpr":0.5,"tpe":0.368,"ee":0.2667},{"task_id":"task_285_enterprise_industry_analysis_medium_medium100","category":"enterprise_industry_analysis","score":1.0,"tokens":148546,"cost_usd":0.77001,"elapsed_time":119.82,"gpr":1.0,"tpe":0.6104,"ee":0.4},{"task_id":"task_286_enterprise_industry_analysis_medium_medium101","category":"enterprise_industry_analysis","score":1.0,"tokens":241299,"cost_usd":1.244035,"elapsed_time":104.06,"gpr":1.0,"tpe":0.5808,"ee":0.3636},{"task_id":"task_287_enterprise_industry_analysis_medium_medium102","category":"enterprise_industry_analysis","score":0.0,"tokens":192575,"cost_usd":0.995995,"elapsed_time":716.05,"gpr":0.25,"tpe":1.0,"ee":0.4444},{"task_id":"task_288_enterprise_industry_analysis_medium_medium103","category":"enterprise_industry_analysis","score":0.0,"tokens":150645,"cost_usd":0.778945,"elapsed_time":104.55,"gpr":0.25,"tpe":0.9,"ee":0.5714},{"task_id":"task_289_enterprise_industry_analysis_medium_medium104","category":"enterprise_industry_analysis","score":0.0,"tokens":147147,"cost_usd":0.761975,"elapsed_time":72.65,"gpr":0.5,"tpe":0.855,"ee":0.5},{"task_id":"task_290_enterprise_industry_analysis_medium_medium105","category":"enterprise_industry_analysis","score":0.0,"tokens":468306,"cost_usd":2.40875,"elapsed_time":243.06,"gpr":0.25,"tpe":1.0,"ee":0.2222},{"task_id":"task_291_enterprise_industry_analysis_medium_medium106","category":"enterprise_industry_analysis","score":0.0,"tokens":108152,"cost_usd":0.5597,"elapsed_time":129.12,"gpr":0.25,"tpe":0.9,"ee":0.6667},{"task_id":"task_292_enterprise_industry_analysis_medium_medium107","category":"enterprise_industry_analysis","score":0.0,"tokens":214970,"cost_usd":1.10481,"elapsed_time":65.94,"gpr":0.25,"tpe":0.6561,"ee":0.4444},{"task_id":"task_293_enterprise_industry_analysis_medium_medium108","category":"enterprise_industry_analysis","score":0.0,"tokens":83792,"cost_usd":0.43128,"elapsed_time":108.16,"gpr":0.25,"tpe":1.0,"ee":0.8},{"task_id":"task_294_enterprise_industry_analysis_medium_medium109","category":"enterprise_industry_analysis","score":1.0,"tokens":226221,"cost_usd":1.165705,"elapsed_time":219.41,"gpr":1.0,"tpe":0.5063,"ee":0.3636},{"task_id":"task_295_enterprise_industry_analysis_medium_medium110","category":"enterprise_industry_analysis","score":0.0,"tokens":179883,"cost_usd":0.936975,"elapsed_time":131.15,"gpr":0.25,"tpe":0.5905,"ee":0.3636},{"task_id":"task_296_enterprise_industry_analysis_medium_medium111","category":"enterprise_industry_analysis","score":0.0,"tokens":102771,"cost_usd":0.530575,"elapsed_time":60.27,"gpr":0.25,"tpe":0.81,"ee":0.6667},{"task_id":"task_297_enterprise_industry_policy_analysis_easy_easy067","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":116598,"cost_usd":0.60081,"elapsed_time":64.89,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_298_enterprise_industry_policy_analysis_easy_easy068","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":297718,"cost_usd":1.51923,"elapsed_time":135.88,"gpr":1.0,"tpe":0.6189,"ee":0.3},{"task_id":"task_299_enterprise_industry_policy_analysis_easy_easy069","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":112576,"cost_usd":0.5818,"elapsed_time":108.15,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_300_enterprise_industry_policy_analysis_easy_easy070","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":121042,"cost_usd":0.62873,"elapsed_time":202.43,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_301_enterprise_industry_policy_analysis_easy_easy071","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":124721,"cost_usd":0.640225,"elapsed_time":71.03,"gpr":1.0,"tpe":0.7047,"ee":0.4286},{"task_id":"task_302_enterprise_industry_policy_analysis_easy_easy072","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":56122,"cost_usd":0.28761,"elapsed_time":125.51,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_303_enterprise_industry_policy_analysis_easy_easy073","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":41861,"cost_usd":0.218785,"elapsed_time":64.85,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_304_enterprise_industry_policy_analysis_easy_easy074","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":151021,"cost_usd":0.775325,"elapsed_time":81.49,"gpr":0.2,"tpe":0.6561,"ee":0.4286},{"task_id":"task_305_enterprise_industry_policy_analysis_easy_easy075","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":194372,"cost_usd":0.99138,"elapsed_time":666.75,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_306_enterprise_industry_policy_analysis_easy_easy076","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":150776,"cost_usd":0.77808,"elapsed_time":94.91,"gpr":0.6667,"tpe":0.8145,"ee":0.5},{"task_id":"task_307_enterprise_industry_policy_analysis_medium_medium001","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":175986,"cost_usd":0.90467,"elapsed_time":200.06,"gpr":1.0,"tpe":0.8878,"ee":0.625},{"task_id":"task_308_enterprise_industry_policy_analysis_medium_medium002","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":176231,"cost_usd":0.916875,"elapsed_time":227.27,"gpr":1.0,"tpe":0.4996,"ee":0.4167},{"task_id":"task_309_enterprise_industry_policy_analysis_medium_medium003","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":204847,"cost_usd":1.062955,"elapsed_time":107.35,"gpr":0.8,"tpe":0.733,"ee":0.4545},{"task_id":"task_310_enterprise_industry_policy_analysis_medium_medium004","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":173840,"cost_usd":0.91652,"elapsed_time":278.29,"gpr":0.4,"tpe":0.81,"ee":0.4545},{"task_id":"task_311_enterprise_industry_policy_analysis_medium_medium005","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":530300,"cost_usd":2.73512,"elapsed_time":626.39,"gpr":1.0,"tpe":0.4097,"ee":0.2083},{"task_id":"task_312_enterprise_industry_policy_analysis_medium_medium006","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":262318,"cost_usd":1.36527,"elapsed_time":320.21,"gpr":0.6,"tpe":0.7171,"ee":0.3571},{"task_id":"task_313_enterprise_industry_policy_analysis_medium_medium007","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":256338,"cost_usd":1.32433,"elapsed_time":463.59,"gpr":1.0,"tpe":0.6463,"ee":0.3846},{"task_id":"task_314_enterprise_industry_policy_analysis_medium_medium008","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":243331,"cost_usd":1.250695,"elapsed_time":285.84,"gpr":1.0,"tpe":0.5575,"ee":0.3846},{"task_id":"task_315_enterprise_industry_policy_analysis_medium_medium009","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":262904,"cost_usd":1.35688,"elapsed_time":241.73,"gpr":1.0,"tpe":0.5217,"ee":0.3571},{"task_id":"task_316_enterprise_industry_policy_analysis_medium_medium010","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":440100,"cost_usd":2.29622,"elapsed_time":581.79,"gpr":0.6,"tpe":0.5968,"ee":0.25},{"task_id":"task_317_enterprise_industry_policy_analysis_medium_medium011","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":133300,"cost_usd":0.69258,"elapsed_time":67.38,"gpr":1.0,"tpe":0.922,"ee":0.7143},{"task_id":"task_318_enterprise_industry_policy_analysis_medium_medium012","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":218969,"cost_usd":1.138585,"elapsed_time":167.54,"gpr":0.6,"tpe":0.7732,"ee":0.4545},{"task_id":"task_319_enterprise_industry_policy_analysis_medium_medium013","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":237324,"cost_usd":1.2299,"elapsed_time":130.46,"gpr":1.0,"tpe":0.7176,"ee":0.4545},{"task_id":"task_320_enterprise_industry_policy_analysis_medium_medium014","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":153943,"cost_usd":0.801755,"elapsed_time":283.91,"gpr":1.0,"tpe":0.8716,"ee":0.625},{"task_id":"task_321_enterprise_industry_policy_analysis_medium_medium015","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":413209,"cost_usd":2.136525,"elapsed_time":833.49,"gpr":1.0,"tpe":0.6541,"ee":0.3125},{"task_id":"task_322_enterprise_industry_policy_analysis_medium_medium016","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":292833,"cost_usd":1.510285,"elapsed_time":215.75,"gpr":1.0,"tpe":0.6361,"ee":0.3333},{"task_id":"task_323_enterprise_industry_policy_analysis_medium_medium017","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":292123,"cost_usd":1.504035,"elapsed_time":304.92,"gpr":1.0,"tpe":0.7132,"ee":0.4167},{"task_id":"task_324_enterprise_industry_policy_analysis_medium_medium018","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":265905,"cost_usd":1.397965,"elapsed_time":161.5,"gpr":1.0,"tpe":0.7581,"ee":0.4167},{"task_id":"task_325_enterprise_industry_policy_analysis_medium_medium019","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":158254,"cost_usd":0.82743,"elapsed_time":207.98,"gpr":1.0,"tpe":0.8878,"ee":0.625},{"task_id":"task_326_enterprise_industry_policy_analysis_medium_medium020","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":167751,"cost_usd":0.873635,"elapsed_time":179.25,"gpr":0.6,"tpe":0.9367,"ee":0.625},{"task_id":"task_327_enterprise_industry_policy_analysis_medium_medium021","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":303367,"cost_usd":1.573675,"elapsed_time":391.83,"gpr":0.6,"tpe":0.9367,"ee":0.3571},{"task_id":"task_328_enterprise_industry_policy_analysis_medium_medium022","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":416258,"cost_usd":2.16253,"elapsed_time":216.65,"gpr":0.4,"tpe":0.729,"ee":0.3125},{"task_id":"task_329_enterprise_industry_policy_analysis_medium_medium023","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":380281,"cost_usd":1.962505,"elapsed_time":384.48,"gpr":1.0,"tpe":0.5467,"ee":0.3125},{"task_id":"task_330_enterprise_industry_policy_analysis_medium_medium024","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":404729,"cost_usd":2.091905,"elapsed_time":293.74,"gpr":0.8,"tpe":0.6597,"ee":0.2941},{"task_id":"task_331_enterprise_industry_policy_analysis_medium_medium025","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":374252,"cost_usd":1.93284,"elapsed_time":324.41,"gpr":1.0,"tpe":0.5901,"ee":0.3333},{"task_id":"task_332_enterprise_industry_policy_analysis_medium_medium026","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":413245,"cost_usd":2.144745,"elapsed_time":273.24,"gpr":1.0,"tpe":0.551,"ee":0.2941},{"task_id":"task_333_enterprise_industry_policy_analysis_medium_medium027","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":272169,"cost_usd":1.419545,"elapsed_time":236.08,"gpr":1.0,"tpe":0.6999,"ee":0.3571},{"task_id":"task_334_enterprise_industry_policy_analysis_medium_medium028","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":249544,"cost_usd":1.29816,"elapsed_time":224.2,"gpr":1.0,"tpe":0.6796,"ee":0.4545},{"task_id":"task_335_enterprise_industry_policy_analysis_medium_medium029","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":509938,"cost_usd":2.58873,"elapsed_time":176.75,"gpr":1.0,"tpe":0.6883,"ee":0.4545},{"task_id":"task_336_enterprise_industry_policy_analysis_medium_medium030","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":264575,"cost_usd":1.364515,"elapsed_time":194.99,"gpr":0.6,"tpe":0.6145,"ee":0.4167},{"task_id":"task_337_enterprise_industry_policy_analysis_medium_medium031","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":350181,"cost_usd":1.801245,"elapsed_time":536.76,"gpr":1.0,"tpe":0.5808,"ee":0.2667},{"task_id":"task_338_enterprise_industry_policy_analysis_medium_medium032","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":428684,"cost_usd":2.19436,"elapsed_time":336.59,"gpr":0.25,"tpe":0.9,"ee":0.2222},{"task_id":"task_339_enterprise_industry_policy_analysis_medium_medium033","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":256157,"cost_usd":1.347065,"elapsed_time":168.36,"gpr":1.0,"tpe":0.4605,"ee":0.2857},{"task_id":"task_340_enterprise_industry_policy_analysis_medium_medium034","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":260178,"cost_usd":1.32955,"elapsed_time":243.02,"gpr":0.5,"tpe":0.6302,"ee":0.3333},{"task_id":"task_341_enterprise_industry_policy_analysis_medium_medium035","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":321022,"cost_usd":1.64443,"elapsed_time":395.48,"gpr":1.0,"tpe":0.5338,"ee":0.25},{"task_id":"task_342_enterprise_industry_policy_analysis_medium_medium036","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":294505,"cost_usd":1.522565,"elapsed_time":210.69,"gpr":1.0,"tpe":0.5937,"ee":0.3333},{"task_id":"task_343_enterprise_industry_policy_analysis_medium_medium037","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":490217,"cost_usd":2.552365,"elapsed_time":334.57,"gpr":0.75,"tpe":0.49,"ee":0.1739},{"task_id":"task_344_enterprise_industry_policy_analysis_medium_medium038","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":372028,"cost_usd":1.8997,"elapsed_time":579.19,"gpr":1.0,"tpe":0.464,"ee":0.2222},{"task_id":"task_345_enterprise_industry_policy_analysis_medium_medium039","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":371491,"cost_usd":1.914395,"elapsed_time":386.44,"gpr":0.75,"tpe":0.7374,"ee":0.2222},{"task_id":"task_346_enterprise_industry_policy_analysis_medium_medium040","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":128992,"cost_usd":0.68428,"elapsed_time":200.37,"gpr":1.0,"tpe":0.925,"ee":0.5714},{"task_id":"task_347_enterprise_industry_policy_analysis_medium_medium041","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":375217,"cost_usd":1.925825,"elapsed_time":527.56,"gpr":0.75,"tpe":0.3772,"ee":0.25},{"task_id":"task_348_enterprise_industry_policy_analysis_medium_medium042","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":422342,"cost_usd":2.16617,"elapsed_time":294.76,"gpr":0.5,"tpe":0.6202,"ee":0.25},{"task_id":"task_349_enterprise_industry_policy_analysis_medium_medium043","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":396325,"cost_usd":2.051765,"elapsed_time":331.99,"gpr":0.75,"tpe":0.5636,"ee":0.2667},{"task_id":"task_350_enterprise_industry_policy_analysis_medium_medium044","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":321466,"cost_usd":1.68089,"elapsed_time":342.22,"gpr":0.75,"tpe":0.3487,"ee":0.2667},{"task_id":"task_351_enterprise_industry_policy_analysis_medium_medium045","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":398757,"cost_usd":2.041305,"elapsed_time":273.95,"gpr":0.5,"tpe":0.4696,"ee":0.2667},{"task_id":"task_352_enterprise_industry_policy_analysis_medium_medium046","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":325049,"cost_usd":1.694925,"elapsed_time":391.62,"gpr":0.5,"tpe":0.828,"ee":0.3333},{"task_id":"task_353_enterprise_industry_policy_analysis_medium_medium047","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":376570,"cost_usd":1.94065,"elapsed_time":367.01,"gpr":1.0,"tpe":0.4385,"ee":0.2667},{"task_id":"task_354_enterprise_industry_policy_analysis_medium_medium048","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":451639,"cost_usd":2.341755,"elapsed_time":337.49,"gpr":1.0,"tpe":0.4185,"ee":0.25},{"task_id":"task_355_enterprise_industry_policy_analysis_medium_medium049","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":334615,"cost_usd":1.729915,"elapsed_time":307.54,"gpr":1.0,"tpe":0.549,"ee":0.3077},{"task_id":"task_356_enterprise_industry_policy_analysis_medium_medium050","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":234783,"cost_usd":1.200675,"elapsed_time":196.87,"gpr":0.5,"tpe":0.6597,"ee":0.3636},{"task_id":"task_357_enterprise_industry_policy_analysis_medium_medium051","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":161660,"cost_usd":0.83778,"elapsed_time":199.21,"gpr":0.5,"tpe":0.8145,"ee":0.5714},{"task_id":"task_358_enterprise_industry_policy_analysis_medium_medium052","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":498792,"cost_usd":2.58172,"elapsed_time":318.44,"gpr":0.75,"tpe":0.4687,"ee":0.2105},{"task_id":"task_359_enterprise_industry_policy_analysis_medium_medium053","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":110323,"cost_usd":0.573255,"elapsed_time":707.28,"gpr":0.75,"tpe":0.9667,"ee":0.6667},{"task_id":"task_360_enterprise_industry_policy_analysis_medium_medium054","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":17081,"cost_usd":0.085605,"elapsed_time":17.68,"gpr":0.0,"tpe":0.0,"ee":4.0},{"task_id":"task_361_enterprise_industry_policy_analysis_medium_medium055","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":205376,"cost_usd":1.0537,"elapsed_time":110.89,"gpr":1.0,"tpe":0.6306,"ee":0.4},{"task_id":"task_362_enterprise_industry_policy_analysis_medium_medium056","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":449386,"cost_usd":2.33975,"elapsed_time":334.21,"gpr":0.5,"tpe":0.733,"ee":0.2222},{"task_id":"task_363_enterprise_industry_policy_analysis_medium_medium057","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":370751,"cost_usd":1.916655,"elapsed_time":206.6,"gpr":0.25,"tpe":0.5905,"ee":0.2353},{"task_id":"task_364_enterprise_industry_policy_analysis_medium_medium058","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":558353,"cost_usd":2.866485,"elapsed_time":403.22,"gpr":0.25,"tpe":0.4783,"ee":0.1667},{"task_id":"task_365_enterprise_industry_policy_analysis_medium_medium059","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":320568,"cost_usd":1.67336,"elapsed_time":254.43,"gpr":0.5,"tpe":0.6441,"ee":0.3333},{"task_id":"task_366_enterprise_industry_policy_analysis_medium_medium060","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":647979,"cost_usd":3.331215,"elapsed_time":279.4,"gpr":0.25,"tpe":0.9,"ee":0.2222},{"task_id":"task_367_enterprise_industry_policy_analysis_medium_medium061","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2959726,"cost_usd":14.87537,"elapsed_time":870.24,"gpr":0.25,"tpe":0.81,"ee":0.1429},{"task_id":"task_368_enterprise_industry_policy_analysis_medium_medium062","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":867733,"cost_usd":4.434005,"elapsed_time":420.02,"gpr":0.5,"tpe":0.8645,"ee":0.2105},{"task_id":"task_369_enterprise_industry_policy_analysis_medium_medium063","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":293221,"cost_usd":1.525665,"elapsed_time":248.42,"gpr":0.25,"tpe":0.9,"ee":0.25},{"task_id":"task_370_enterprise_industry_policy_analysis_medium_medium064","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":407403,"cost_usd":2.130255,"elapsed_time":631.77,"gpr":0.5,"tpe":0.5433,"ee":0.2105},{"task_id":"task_371_enterprise_industry_policy_analysis_medium_medium065","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":904111,"cost_usd":4.617635,"elapsed_time":1207.4,"gpr":0.25,"tpe":0.5905,"ee":0.125},{"task_id":"task_372_enterprise_industry_policy_analysis_medium_medium066","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":694342,"cost_usd":3.51627,"elapsed_time":201.75,"gpr":1.0,"tpe":0.7994,"ee":0.3077},{"task_id":"task_373_hypothesis_verification_hard_hard001","category":"hypothesis_verification","score":0.6667,"tokens":545016,"cost_usd":2.8411,"elapsed_time":210.25,"gpr":0.5556,"tpe":0.81,"ee":0.6364},{"task_id":"task_374_hypothesis_verification_hard_hard002","category":"hypothesis_verification","score":0.0,"tokens":330066,"cost_usd":1.73159,"elapsed_time":125.99,"gpr":0.375,"tpe":1.0,"ee":0.5},{"task_id":"task_375_hypothesis_verification_hard_hard003","category":"hypothesis_verification","score":0.0,"tokens":2377292,"cost_usd":12.18238,"elapsed_time":650.7,"gpr":0.3333,"tpe":0.4284,"ee":0.2424},{"task_id":"task_376_hypothesis_verification_hard_hard004","category":"hypothesis_verification","score":0.0,"tokens":1872224,"cost_usd":9.48908,"elapsed_time":512.59,"gpr":0.1,"tpe":0.4783,"ee":0.35},{"task_id":"task_377_hypothesis_verification_hard_hard005","category":"hypothesis_verification","score":0.0,"tokens":1075378,"cost_usd":5.68289,"elapsed_time":511.59,"gpr":0.0,"tpe":0.0,"ee":0.45},{"task_id":"task_378_hypothesis_verification_hard_hard006","category":"hypothesis_verification","score":0.0,"tokens":3118676,"cost_usd":15.8541,"elapsed_time":483.77,"gpr":0.1818,"tpe":0.81,"ee":0.175},{"task_id":"task_379_hypothesis_verification_hard_hard007","category":"hypothesis_verification","score":0.0,"tokens":4562260,"cost_usd":23.54856,"elapsed_time":1207.21,"gpr":0.0625,"tpe":0.81,"ee":0.1552},{"task_id":"task_380_hypothesis_verification_hard_hard008","category":"hypothesis_verification","score":1.0,"tokens":1070775,"cost_usd":5.566195,"elapsed_time":322.73,"gpr":1.0,"tpe":0.4882,"ee":0.3043},{"task_id":"task_381_hypothesis_verification_hard_hard009","category":"hypothesis_verification","score":0.0,"tokens":2598172,"cost_usd":13.16608,"elapsed_time":515.2,"gpr":0.2222,"tpe":0.4114,"ee":0.2609},{"task_id":"task_382_hypothesis_verification_hard_hard010","category":"hypothesis_verification","score":0.0,"tokens":2877003,"cost_usd":14.791895,"elapsed_time":649.96,"gpr":0.0667,"tpe":0.3874,"ee":0.2059},{"task_id":"task_383_hypothesis_verification_hard_hard011","category":"hypothesis_verification","score":1.0,"tokens":979559,"cost_usd":5.114015,"elapsed_time":415.58,"gpr":1.0,"tpe":0.3741,"ee":0.3333},{"task_id":"task_384_hypothesis_verification_hard_hard012","category":"hypothesis_verification","score":0.0,"tokens":1450651,"cost_usd":7.519935,"elapsed_time":501.01,"gpr":0.3636,"tpe":0.4858,"ee":0.3333},{"task_id":"task_385_hypothesis_verification_hard_hard013","category":"hypothesis_verification","score":0.25,"tokens":1548550,"cost_usd":8.06019,"elapsed_time":458.69,"gpr":0.0,"tpe":0.0,"ee":0.35},{"task_id":"task_386_hypothesis_verification_hard_hard014","category":"hypothesis_verification","score":1.0,"tokens":1721985,"cost_usd":9.017345,"elapsed_time":505.45,"gpr":1.0,"tpe":0.5937,"ee":0.2857},{"task_id":"task_387_hypothesis_verification_hard_hard015","category":"hypothesis_verification","score":0.0,"tokens":1684165,"cost_usd":8.686105,"elapsed_time":628.54,"gpr":0.1818,"tpe":0.7952,"ee":0.3889},{"task_id":"task_388_hypothesis_verification_medium_medium001","category":"hypothesis_verification","score":0.0,"tokens":555703,"cost_usd":2.880815,"elapsed_time":215.72,"gpr":0.4,"tpe":1.0,"ee":0.5385},{"task_id":"task_389_hypothesis_verification_medium_medium002","category":"hypothesis_verification","score":0.0,"tokens":341698,"cost_usd":1.80881,"elapsed_time":243.52,"gpr":0.2,"tpe":1.0,"ee":0.4667},{"task_id":"task_390_hypothesis_verification_medium_medium003","category":"hypothesis_verification","score":1.0,"tokens":2262000,"cost_usd":11.72614,"elapsed_time":668.27,"gpr":1.0,"tpe":0.1719,"ee":0.1579},{"task_id":"task_391_hypothesis_verification_medium_medium004","category":"hypothesis_verification","score":1.0,"tokens":255306,"cost_usd":1.33915,"elapsed_time":143.67,"gpr":0.6,"tpe":0.7468,"ee":0.875},{"task_id":"task_392_hypothesis_verification_medium_medium005","category":"hypothesis_verification","score":0.0,"tokens":462794,"cost_usd":2.42737,"elapsed_time":285.8,"gpr":0.4,"tpe":0.7781,"ee":0.3889},{"task_id":"task_393_hypothesis_verification_medium_medium006","category":"hypothesis_verification","score":0.0,"tokens":572332,"cost_usd":3.01622,"elapsed_time":280.48,"gpr":0.0,"tpe":0.0,"ee":0.3529},{"task_id":"task_394_hypothesis_verification_medium_medium007","category":"hypothesis_verification","score":0.0,"tokens":792470,"cost_usd":4.19349,"elapsed_time":369.77,"gpr":0.2222,"tpe":0.9,"ee":0.3684},{"task_id":"task_395_hypothesis_verification_medium_medium008","category":"hypothesis_verification","score":0.0,"tokens":392608,"cost_usd":2.07462,"elapsed_time":211.59,"gpr":1.0,"tpe":0.7602,"ee":0.5},{"task_id":"task_396_hypothesis_verification_medium_medium009","category":"hypothesis_verification","score":0.0,"tokens":1162088,"cost_usd":5.9593,"elapsed_time":249.93,"gpr":0.25,"tpe":0.2824,"ee":0.2727},{"task_id":"task_397_hypothesis_verification_medium_medium010","category":"hypothesis_verification","score":1.0,"tokens":525774,"cost_usd":2.78865,"elapsed_time":243.21,"gpr":1.0,"tpe":0.4907,"ee":0.4375},{"task_id":"task_398_hypothesis_verification_medium_medium011","category":"hypothesis_verification","score":1.0,"tokens":642949,"cost_usd":3.539385,"elapsed_time":345.66,"gpr":1.0,"tpe":0.2059,"ee":0.2857},{"task_id":"task_399_hypothesis_verification_medium_medium012","category":"hypothesis_verification","score":1.0,"tokens":228720,"cost_usd":1.2264,"elapsed_time":235.4,"gpr":1.0,"tpe":0.7484,"ee":0.6},{"task_id":"task_400_hypothesis_verification_medium_medium013","category":"hypothesis_verification","score":1.0,"tokens":490354,"cost_usd":2.53619,"elapsed_time":298.43,"gpr":1.0,"tpe":0.3138,"ee":0.35},{"task_id":"task_401_hypothesis_verification_medium_medium014","category":"hypothesis_verification","score":1.0,"tokens":581491,"cost_usd":2.990915,"elapsed_time":131.01,"gpr":0.0,"tpe":0.0,"ee":0.7273},{"task_id":"task_402_industry_planning_hard_hard001","category":"industry_planning","score":0.5,"tokens":531316,"cost_usd":2.75488,"elapsed_time":164.18,"gpr":0.5556,"tpe":0.6561,"ee":0.5385},{"task_id":"task_403_industry_planning_hard_hard002","category":"industry_planning","score":0.0,"tokens":1623372,"cost_usd":8.46956,"elapsed_time":446.69,"gpr":0.25,"tpe":0.7317,"ee":0.28},{"task_id":"task_404_industry_planning_hard_hard003","category":"industry_planning","score":1.0,"tokens":913857,"cost_usd":4.826245,"elapsed_time":370.62,"gpr":1.0,"tpe":0.287,"ee":0.3478},{"task_id":"task_405_industry_planning_hard_hard004","category":"industry_planning","score":0.5,"tokens":788080,"cost_usd":4.18632,"elapsed_time":361.86,"gpr":1.0,"tpe":0.4851,"ee":0.2727},{"task_id":"task_406_industry_planning_hard_hard005","category":"industry_planning","score":1.0,"tokens":1054259,"cost_usd":5.367975,"elapsed_time":171.1,"gpr":1.0,"tpe":0.7892,"ee":0.5},{"task_id":"task_407_industry_planning_hard_hard006","category":"industry_planning","score":1.0,"tokens":1847237,"cost_usd":9.614445,"elapsed_time":696.7,"gpr":1.0,"tpe":0.439,"ee":0.2692},{"task_id":"task_408_industry_planning_hard_hard007","category":"industry_planning","score":0.0,"tokens":1590737,"cost_usd":8.293545,"elapsed_time":546.6,"gpr":0.2,"tpe":0.5582,"ee":0.2308},{"task_id":"task_409_industry_planning_hard_hard008","category":"industry_planning","score":0.0,"tokens":3176801,"cost_usd":16.167685,"elapsed_time":440.64,"gpr":0.875,"tpe":0.3998,"ee":0.2222},{"task_id":"task_410_industry_planning_hard_hard009","category":"industry_planning","score":0.67,"tokens":2749686,"cost_usd":14.22395,"elapsed_time":1207.13,"gpr":0.5,"tpe":0.0703,"ee":0.1053},{"task_id":"task_411_industry_planning_hard_hard010","category":"industry_planning","score":0.0,"tokens":5756327,"cost_usd":29.078115,"elapsed_time":683.69,"gpr":0.6667,"tpe":0.2089,"ee":0.1765},{"task_id":"task_412_industry_planning_hard_hard011","category":"industry_planning","score":1.0,"tokens":1285286,"cost_usd":6.62613,"elapsed_time":425.47,"gpr":1.0,"tpe":0.2697,"ee":0.2857},{"task_id":"task_413_industry_planning_hard_hard012","category":"industry_planning","score":1.0,"tokens":3300782,"cost_usd":16.79963,"elapsed_time":540.5,"gpr":1.0,"tpe":0.11,"ee":0.1875},{"task_id":"task_414_industry_planning_hard_hard013","category":"industry_planning","score":0.0,"tokens":3699064,"cost_usd":19.19042,"elapsed_time":1087.07,"gpr":0.0,"tpe":0.0,"ee":0.1321},{"task_id":"task_415_industry_planning_hard_hard014","category":"industry_planning","score":1.0,"tokens":3461260,"cost_usd":17.74504,"elapsed_time":824.92,"gpr":1.0,"tpe":0.1212,"ee":0.1111},{"task_id":"task_416_industry_planning_medium_medium001","category":"industry_planning","score":1.0,"tokens":489839,"cost_usd":2.596255,"elapsed_time":225.15,"gpr":1.0,"tpe":0.8281,"ee":0.6667},{"task_id":"task_417_industry_planning_medium_medium002","category":"industry_planning","score":1.0,"tokens":899102,"cost_usd":4.68187,"elapsed_time":343.38,"gpr":1.0,"tpe":0.3625,"ee":0.3077},{"task_id":"task_418_industry_planning_medium_medium003","category":"industry_planning","score":0.0,"tokens":1826381,"cost_usd":9.344225,"elapsed_time":498.11,"gpr":0.0,"tpe":0.0,"ee":0.2286},{"task_id":"task_419_industry_planning_medium_medium004","category":"industry_planning","score":1.0,"tokens":293157,"cost_usd":1.576625,"elapsed_time":200.7,"gpr":1.0,"tpe":0.7543,"ee":0.5833},{"task_id":"task_420_industry_planning_medium_medium005","category":"industry_planning","score":0.0,"tokens":3179073,"cost_usd":16.598305,"elapsed_time":1206.61,"gpr":0.0,"tpe":0.0,"ee":0.1296},{"task_id":"task_421_industry_planning_medium_medium006","category":"industry_planning","score":0.0,"tokens":2817582,"cost_usd":14.51295,"elapsed_time":681.23,"gpr":0.0,"tpe":0.0,"ee":0.2222},{"task_id":"task_422_industry_planning_medium_medium007","category":"industry_planning","score":1.0,"tokens":552096,"cost_usd":2.9998,"elapsed_time":522.25,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_423_industry_planning_medium_medium008","category":"industry_planning","score":1.0,"tokens":640991,"cost_usd":3.448035,"elapsed_time":356.67,"gpr":1.0,"tpe":0.3879,"ee":0.4},{"task_id":"task_424_industry_planning_medium_medium009","category":"industry_planning","score":1.0,"tokens":941313,"cost_usd":4.988945,"elapsed_time":362.87,"gpr":1.0,"tpe":0.311,"ee":0.3636},{"task_id":"task_425_industry_planning_medium_medium010","category":"industry_planning","score":0.0,"tokens":474947,"cost_usd":2.493655,"elapsed_time":159.93,"gpr":0.4,"tpe":1.0,"ee":0.7273},{"task_id":"task_426_industry_planning_medium_medium011","category":"industry_planning","score":1.0,"tokens":477723,"cost_usd":2.590875,"elapsed_time":537.84,"gpr":1.0,"tpe":0.6929,"ee":0.6154},{"task_id":"task_427_industry_planning_medium_medium012","category":"industry_planning","score":1.0,"tokens":888374,"cost_usd":4.62255,"elapsed_time":246.89,"gpr":1.0,"tpe":0.8311,"ee":0.6667},{"task_id":"task_428_industry_planning_medium_medium013","category":"industry_planning","score":1.0,"tokens":612025,"cost_usd":3.219225,"elapsed_time":398.21,"gpr":1.0,"tpe":0.4908,"ee":0.4211},{"task_id":"task_429_industry_planning_medium_medium014","category":"industry_planning","score":1.0,"tokens":1487598,"cost_usd":7.70319,"elapsed_time":773.61,"gpr":1.0,"tpe":0.3616,"ee":0.2308},{"task_id":"task_430_international_comparison_hard_hard001","category":"international_comparison","score":0.5,"tokens":1102340,"cost_usd":5.66086,"elapsed_time":373.68,"gpr":0.5,"tpe":0.4245,"ee":0.2857},{"task_id":"task_431_international_comparison_hard_hard002","category":"international_comparison","score":0.0,"tokens":7353058,"cost_usd":37.11711,"elapsed_time":1207.14,"gpr":1.0,"tpe":0.0021,"ee":0.087},{"task_id":"task_432_international_comparison_hard_hard003","category":"international_comparison","score":0.0,"tokens":1164764,"cost_usd":5.97896,"elapsed_time":391.98,"gpr":0.5556,"tpe":0.4102,"ee":0.24},{"task_id":"task_433_international_comparison_hard_hard004","category":"international_comparison","score":0.0,"tokens":723849,"cost_usd":3.745585,"elapsed_time":305.33,"gpr":0.0,"tpe":0.0,"ee":0.2609},{"task_id":"task_434_international_comparison_hard_hard005","category":"international_comparison","score":1.0,"tokens":1072826,"cost_usd":5.52301,"elapsed_time":394.32,"gpr":1.0,"tpe":0.3217,"ee":0.25},{"task_id":"task_435_international_comparison_hard_hard006","category":"international_comparison","score":0.25,"tokens":1069521,"cost_usd":5.505545,"elapsed_time":311.94,"gpr":0.1667,"tpe":0.2288,"ee":0.2917},{"task_id":"task_436_international_comparison_hard_hard007","category":"international_comparison","score":0.5,"tokens":1621281,"cost_usd":8.259705,"elapsed_time":621.37,"gpr":0.4545,"tpe":0.6586,"ee":0.3043},{"task_id":"task_437_international_comparison_hard_hard008","category":"international_comparison","score":0.5,"tokens":745511,"cost_usd":3.847095,"elapsed_time":311.39,"gpr":0.1667,"tpe":0.3874,"ee":0.375},{"task_id":"task_438_international_comparison_hard_hard009","category":"international_comparison","score":1.0,"tokens":1491242,"cost_usd":7.73825,"elapsed_time":787.92,"gpr":1.0,"tpe":0.2288,"ee":0.3333},{"task_id":"task_439_international_comparison_hard_hard010","category":"international_comparison","score":0.5,"tokens":998367,"cost_usd":5.097395,"elapsed_time":297.6,"gpr":0.3333,"tpe":0.368,"ee":0.3158},{"task_id":"task_440_international_comparison_hard_hard011","category":"international_comparison","score":0.5,"tokens":668869,"cost_usd":3.438245,"elapsed_time":200.01,"gpr":0.8,"tpe":0.3874,"ee":0.375},{"task_id":"task_441_international_comparison_hard_hard012","category":"international_comparison","score":1.0,"tokens":745745,"cost_usd":3.880845,"elapsed_time":467.23,"gpr":1.0,"tpe":0.2836,"ee":0.2727},{"task_id":"task_442_international_comparison_hard_hard013","category":"international_comparison","score":0.25,"tokens":1503563,"cost_usd":7.656595,"elapsed_time":463.83,"gpr":0.6,"tpe":0.5916,"ee":0.3158},{"task_id":"task_443_international_comparison_hard_hard014","category":"international_comparison","score":1.0,"tokens":1184433,"cost_usd":6.094425,"elapsed_time":610.27,"gpr":1.0,"tpe":0.0921,"ee":0.2},{"task_id":"task_444_international_comparison_medium_medium001","category":"international_comparison","score":0.0,"tokens":901887,"cost_usd":4.639655,"elapsed_time":307.15,"gpr":1.0,"tpe":0.2344,"ee":0.2},{"task_id":"task_445_international_comparison_medium_medium002","category":"international_comparison","score":1.0,"tokens":970120,"cost_usd":5.00582,"elapsed_time":378.07,"gpr":1.0,"tpe":0.3202,"ee":0.1739},{"task_id":"task_446_international_comparison_medium_medium003","category":"international_comparison","score":0.0,"tokens":415078,"cost_usd":2.13337,"elapsed_time":253.63,"gpr":0.3333,"tpe":0.2824,"ee":0.2353},{"task_id":"task_447_international_comparison_medium_medium004","category":"international_comparison","score":0.0,"tokens":1260305,"cost_usd":6.412025,"elapsed_time":463.64,"gpr":1.0,"tpe":0.1714,"ee":0.1667},{"task_id":"task_448_international_comparison_medium_medium005","category":"international_comparison","score":0.0,"tokens":118934,"cost_usd":0.61241,"elapsed_time":69.56,"gpr":0.0,"tpe":0.0,"ee":0.6667},{"task_id":"task_449_international_comparison_medium_medium006","category":"international_comparison","score":1.0,"tokens":371640,"cost_usd":1.95748,"elapsed_time":208.03,"gpr":1.0,"tpe":0.5967,"ee":0.3571},{"task_id":"task_450_international_comparison_medium_medium007","category":"international_comparison","score":0.33,"tokens":755879,"cost_usd":3.883275,"elapsed_time":278.33,"gpr":0.5,"tpe":0.3745,"ee":0.2353},{"task_id":"task_451_international_comparison_medium_medium008","category":"international_comparison","score":1.0,"tokens":665530,"cost_usd":3.41549,"elapsed_time":211.09,"gpr":1.0,"tpe":0.5413,"ee":0.3636},{"task_id":"task_452_international_comparison_medium_medium009","category":"international_comparison","score":0.0,"tokens":1645807,"cost_usd":8.379995,"elapsed_time":442.07,"gpr":0.5714,"tpe":0.1371,"ee":0.1562},{"task_id":"task_453_international_comparison_medium_medium010","category":"international_comparison","score":0.5,"tokens":677810,"cost_usd":3.52679,"elapsed_time":349.6,"gpr":0.6,"tpe":0.4225,"ee":0.2778},{"task_id":"task_454_international_comparison_medium_medium011","category":"international_comparison","score":1.0,"tokens":86168,"cost_usd":0.4566,"elapsed_time":50.82,"gpr":1.0,"tpe":0.95,"ee":0.8},{"task_id":"task_455_international_comparison_medium_medium012","category":"international_comparison","score":1.0,"tokens":589164,"cost_usd":3.02814,"elapsed_time":247.52,"gpr":1.0,"tpe":0.3301,"ee":0.2778},{"task_id":"task_456_international_comparison_medium_medium013","category":"international_comparison","score":1.0,"tokens":854701,"cost_usd":4.341645,"elapsed_time":312.77,"gpr":1.0,"tpe":0.2931,"ee":0.2105},{"task_id":"task_457_international_comparison_medium_medium014","category":"international_comparison","score":1.0,"tokens":970281,"cost_usd":5.012865,"elapsed_time":339.29,"gpr":0.875,"tpe":0.3534,"ee":0.2381},{"task_id":"task_458_international_comparison_medium_medium015","category":"international_comparison","score":0.0,"tokens":1555691,"cost_usd":8.018775,"elapsed_time":656.49,"gpr":0.2857,"tpe":0.3639,"ee":0.1163},{"task_id":"task_459_international_comparison_medium_medium016","category":"international_comparison","score":1.0,"tokens":356440,"cost_usd":1.84832,"elapsed_time":115.74,"gpr":0.8333,"tpe":0.6108,"ee":0.4545},{"task_id":"task_460_international_comparison_medium_medium017","category":"international_comparison","score":0.0,"tokens":402361,"cost_usd":2.152905,"elapsed_time":291.03,"gpr":0.3,"tpe":0.4305,"ee":0.2941},{"task_id":"task_461_international_comparison_medium_medium018","category":"international_comparison","score":1.0,"tokens":275815,"cost_usd":1.448775,"elapsed_time":189.9,"gpr":1.0,"tpe":0.6026,"ee":0.3636},{"task_id":"task_462_international_comparison_medium_medium019","category":"international_comparison","score":0.5,"tokens":1463054,"cost_usd":7.45987,"elapsed_time":530.28,"gpr":0.6667,"tpe":0.2119,"ee":0.1333},{"task_id":"task_463_international_comparison_medium_medium020","category":"international_comparison","score":0.0,"tokens":3905144,"cost_usd":19.92982,"elapsed_time":1207.09,"gpr":0.4444,"tpe":0.3244,"ee":0.0847},{"task_id":"task_464_international_comparison_medium_medium021","category":"international_comparison","score":1.0,"tokens":1683752,"cost_usd":8.67562,"elapsed_time":512.29,"gpr":1.0,"tpe":0.2639,"ee":0.2174},{"task_id":"task_465_international_comparison_medium_medium022","category":"international_comparison","score":0.0,"tokens":601738,"cost_usd":3.11219,"elapsed_time":325.04,"gpr":1.0,"tpe":0.5452,"ee":0.2778},{"task_id":"task_466_international_comparison_medium_medium023","category":"international_comparison","score":0.75,"tokens":401069,"cost_usd":2.121945,"elapsed_time":284.41,"gpr":0.5,"tpe":0.9033,"ee":0.3333},{"task_id":"task_467_international_comparison_medium_medium024","category":"international_comparison","score":0.0,"tokens":1507506,"cost_usd":7.71647,"elapsed_time":415.26,"gpr":1.0,"tpe":0.143,"ee":0.1852},{"task_id":"task_468_international_comparison_medium_medium025","category":"international_comparison","score":1.0,"tokens":2123440,"cost_usd":10.78556,"elapsed_time":605.3,"gpr":1.0,"tpe":0.2185,"ee":0.1379},{"task_id":"task_469_risk_assessment_hard_hard001","category":"risk_assessment","score":1.0,"tokens":192982,"cost_usd":1.04007,"elapsed_time":167.56,"gpr":1.0,"tpe":0.81,"ee":0.7},{"task_id":"task_470_risk_assessment_hard_hard002","category":"risk_assessment","score":1.0,"tokens":218709,"cost_usd":1.200925,"elapsed_time":301.36,"gpr":1.0,"tpe":0.828,"ee":0.6364},{"task_id":"task_471_risk_assessment_hard_hard003","category":"risk_assessment","score":1.0,"tokens":1286963,"cost_usd":6.644915,"elapsed_time":557.87,"gpr":1.0,"tpe":0.5885,"ee":0.3333},{"task_id":"task_472_risk_assessment_hard_hard004","category":"risk_assessment","score":0.0,"tokens":2144677,"cost_usd":11.215645,"elapsed_time":1207.14,"gpr":0.0,"tpe":0.0,"ee":0.1591},{"task_id":"task_473_risk_assessment_hard_hard005","category":"risk_assessment","score":1.0,"tokens":198102,"cost_usd":1.04301,"elapsed_time":146.79,"gpr":1.0,"tpe":0.84,"ee":0.7778},{"task_id":"task_474_risk_assessment_hard_hard006","category":"risk_assessment","score":1.0,"tokens":702378,"cost_usd":3.67017,"elapsed_time":533.81,"gpr":1.0,"tpe":0.5683,"ee":0.3889},{"task_id":"task_475_risk_assessment_hard_hard007","category":"risk_assessment","score":1.0,"tokens":214315,"cost_usd":1.134575,"elapsed_time":159.25,"gpr":1.0,"tpe":0.7587,"ee":0.6364},{"task_id":"task_476_risk_assessment_hard_hard008","category":"risk_assessment","score":0.0,"tokens":171270,"cost_usd":0.91403,"elapsed_time":156.26,"gpr":0.5,"tpe":1.0,"ee":0.7},{"task_id":"task_477_risk_assessment_hard_hard009","category":"risk_assessment","score":0.0,"tokens":1426056,"cost_usd":7.39484,"elapsed_time":707.17,"gpr":1.0,"tpe":0.4557,"ee":0.2414},{"task_id":"task_478_risk_assessment_hard_hard010","category":"risk_assessment","score":0.0,"tokens":684999,"cost_usd":3.573075,"elapsed_time":409.64,"gpr":0.25,"tpe":1.0,"ee":0.5},{"task_id":"task_479_risk_assessment_hard_hard011","category":"risk_assessment","score":1.0,"tokens":2617584,"cost_usd":13.32368,"elapsed_time":503.66,"gpr":1.0,"tpe":0.6578,"ee":0.3889},{"task_id":"task_480_risk_assessment_hard_hard012","category":"risk_assessment","score":0.0,"tokens":372801,"cost_usd":2.040085,"elapsed_time":356.82,"gpr":0.75,"tpe":0.4412,"ee":0.3889},{"task_id":"task_481_risk_assessment_hard_hard013","category":"risk_assessment","score":1.0,"tokens":477144,"cost_usd":2.60416,"elapsed_time":260.94,"gpr":0.6667,"tpe":0.7152,"ee":0.45},{"task_id":"task_482_risk_assessment_medium_medium001","category":"risk_assessment","score":1.0,"tokens":2012158,"cost_usd":10.14589,"elapsed_time":341.35,"gpr":1.0,"tpe":0.6306,"ee":0.4286},{"task_id":"task_483_risk_assessment_medium_medium002","category":"risk_assessment","score":0.0,"tokens":210891,"cost_usd":1.142615,"elapsed_time":278.2,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_484_risk_assessment_medium_medium003","category":"risk_assessment","score":0.0,"tokens":1343543,"cost_usd":7.043415,"elapsed_time":716.53,"gpr":0.2,"tpe":1.0,"ee":0.1579},{"task_id":"task_485_risk_assessment_medium_medium004","category":"risk_assessment","score":1.0,"tokens":1866379,"cost_usd":9.613755,"elapsed_time":524.87,"gpr":1.0,"tpe":0.5044,"ee":0.2083},{"task_id":"task_486_risk_assessment_medium_medium005","category":"risk_assessment","score":1.0,"tokens":1479778,"cost_usd":7.66247,"elapsed_time":387.56,"gpr":1.0,"tpe":0.4522,"ee":0.2778},{"task_id":"task_487_risk_assessment_medium_medium006","category":"risk_assessment","score":1.0,"tokens":1422739,"cost_usd":7.289075,"elapsed_time":449.06,"gpr":1.0,"tpe":0.695,"ee":0.3125},{"task_id":"task_488_risk_assessment_medium_medium007","category":"risk_assessment","score":0.3333,"tokens":546285,"cost_usd":2.837745,"elapsed_time":227.18,"gpr":0.25,"tpe":0.8854,"ee":0.4286},{"task_id":"task_489_risk_assessment_medium_medium008","category":"risk_assessment","score":1.0,"tokens":1657035,"cost_usd":8.491095,"elapsed_time":664.65,"gpr":0.0625,"tpe":0.81,"ee":0.2609},{"task_id":"task_490_risk_assessment_medium_medium009","category":"risk_assessment","score":1.0,"tokens":1116698,"cost_usd":5.78819,"elapsed_time":448.29,"gpr":1.0,"tpe":0.5054,"ee":0.2727},{"task_id":"task_491_risk_assessment_medium_medium010","category":"risk_assessment","score":1.0,"tokens":3928466,"cost_usd":19.89729,"elapsed_time":653.71,"gpr":1.0,"tpe":0.4549,"ee":0.3158},{"task_id":"task_492_risk_assessment_medium_medium011","category":"risk_assessment","score":0.0,"tokens":1334311,"cost_usd":6.852215,"elapsed_time":300.57,"gpr":0.0909,"tpe":1.0,"ee":0.3}]},{"model":"gemini-3.1-pro-preview","acc":0.4579,"ee":0.3234,"gpr":0.3363,"tpe":0.2591,"efficiency":{"total_tokens":750885045,"total_cost_usd":1027.96,"total_requests":15730},"category_scores":{"comprehensive_decision":{"acc":0.3949,"ee":0.3005,"gpr":0.3484,"tpe":0.2438,"count":70},"enterprise_industry_analysis":{"acc":0.615,"ee":0.3657,"gpr":0.3347,"tpe":0.289,"count":226},"enterprise_industry_policy_analysis":{"acc":0.4737,"ee":0.2651,"gpr":0.6112,"tpe":0.5407,"count":76},"hypothesis_verification":{"acc":0.1034,"ee":0.2814,"gpr":0.1476,"tpe":0.1371,"count":29},"industry_planning":{"acc":0.2679,"ee":0.1607,"gpr":0.2893,"tpe":0.2207,"count":28},"international_comparison":{"acc":0.1026,"ee":0.0948,"gpr":0.2181,"tpe":0.0641,"count":39},"risk_assessment":{"acc":0.3401,"ee":0.1124,"gpr":0.2671,"tpe":0.1382,"count":24}},"tasks":[{"task_id":"task_001_comprehensive_decision_easy_easy001","category":"comprehensive_decision","score":0.0,"tokens":118267,"cost_usd":0.181121,"elapsed_time":160.74,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_002_comprehensive_decision_easy_easy002","category":"comprehensive_decision","score":1.0,"tokens":99542,"cost_usd":0.225298,"elapsed_time":151.41,"gpr":1.0,"tpe":0.5806,"ee":0.25},{"task_id":"task_003_comprehensive_decision_easy_easy003","category":"comprehensive_decision","score":1.0,"tokens":150820,"cost_usd":0.237764,"elapsed_time":192.37,"gpr":1.0,"tpe":0.6679,"ee":0.25},{"task_id":"task_004_comprehensive_decision_easy_easy004","category":"comprehensive_decision","score":1.0,"tokens":236677,"cost_usd":0.291461,"elapsed_time":215.31,"gpr":1.0,"tpe":0.2864,"ee":0.1111},{"task_id":"task_005_comprehensive_decision_easy_easy005","category":"comprehensive_decision","score":0.6,"tokens":2428907,"cost_usd":4.210254,"elapsed_time":456.8,"gpr":0.6667,"tpe":0.306,"ee":0.1154},{"task_id":"task_006_comprehensive_decision_easy_easy006","category":"comprehensive_decision","score":0.0,"tokens":255898,"cost_usd":0.457557,"elapsed_time":1208.11,"gpr":0.0,"tpe":0.0,"ee":0.1765},{"task_id":"task_007_comprehensive_decision_hard_hard001","category":"comprehensive_decision","score":0.0,"tokens":3392867,"cost_usd":3.433314,"elapsed_time":1208.22,"gpr":0.25,"tpe":0.6937,"ee":0.1944},{"task_id":"task_008_comprehensive_decision_hard_hard002","category":"comprehensive_decision","score":0.0,"tokens":932746,"cost_usd":1.409312,"elapsed_time":606.99,"gpr":0.2,"tpe":0.1501,"ee":0.1707},{"task_id":"task_009_comprehensive_decision_hard_hard003","category":"comprehensive_decision","score":0.0,"tokens":941222,"cost_usd":1.324092,"elapsed_time":1208.24,"gpr":0.7143,"tpe":0.1429,"ee":0.2},{"task_id":"task_010_comprehensive_decision_hard_hard004","category":"comprehensive_decision","score":0.0,"tokens":455884,"cost_usd":0.855165,"elapsed_time":394.14,"gpr":0.0,"tpe":0.0,"ee":0.28},{"task_id":"task_011_comprehensive_decision_hard_hard005","category":"comprehensive_decision","score":0.0,"tokens":785988,"cost_usd":1.440589,"elapsed_time":1207.42,"gpr":0.7143,"tpe":0.1441,"ee":0.1892},{"task_id":"task_012_comprehensive_decision_hard_hard006","category":"comprehensive_decision","score":0.5,"tokens":2299273,"cost_usd":4.373061,"elapsed_time":1056.08,"gpr":0.5,"tpe":0.067,"ee":0.1944},{"task_id":"task_013_comprehensive_decision_hard_hard007","category":"comprehensive_decision","score":0.0,"tokens":1115021,"cost_usd":1.619921,"elapsed_time":1207.35,"gpr":0.8333,"tpe":0.5099,"ee":0.2121},{"task_id":"task_014_comprehensive_decision_hard_hard008","category":"comprehensive_decision","score":0.0,"tokens":795102,"cost_usd":1.473727,"elapsed_time":1207.37,"gpr":0.5,"tpe":0.1306,"ee":0.2121},{"task_id":"task_015_comprehensive_decision_hard_hard009","category":"comprehensive_decision","score":0.0,"tokens":265500,"cost_usd":0.469459,"elapsed_time":1207.44,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_016_comprehensive_decision_hard_hard010","category":"comprehensive_decision","score":0.0,"tokens":537807,"cost_usd":0.929292,"elapsed_time":1182.13,"gpr":0.0,"tpe":0.0,"ee":0.2917},{"task_id":"task_017_comprehensive_decision_hard_hard011","category":"comprehensive_decision","score":0.0,"tokens":966196,"cost_usd":1.550987,"elapsed_time":1207.43,"gpr":1.0,"tpe":0.1668,"ee":0.1892},{"task_id":"task_018_comprehensive_decision_hard_hard012","category":"comprehensive_decision","score":0.0,"tokens":293411,"cost_usd":0.548537,"elapsed_time":1207.58,"gpr":0.0,"tpe":0.0,"ee":0.35},{"task_id":"task_019_comprehensive_decision_hard_hard013","category":"comprehensive_decision","score":0.0,"tokens":348832,"cost_usd":0.562722,"elapsed_time":1207.29,"gpr":0.0,"tpe":0.0,"ee":0.3684},{"task_id":"task_020_comprehensive_decision_hard_hard014","category":"comprehensive_decision","score":0.0,"tokens":1187818,"cost_usd":1.978441,"elapsed_time":979.97,"gpr":0.4286,"tpe":0.4351,"ee":0.2121},{"task_id":"task_021_comprehensive_decision_hard_hard015","category":"comprehensive_decision","score":0.0,"tokens":673371,"cost_usd":1.527872,"elapsed_time":484.17,"gpr":0.0,"tpe":0.0,"ee":0.3636},{"task_id":"task_022_comprehensive_decision_hard_hard016","category":"comprehensive_decision","score":0.75,"tokens":476571,"cost_usd":0.949536,"elapsed_time":664.61,"gpr":1.0,"tpe":0.254,"ee":0.2143},{"task_id":"task_023_comprehensive_decision_hard_hard017","category":"comprehensive_decision","score":0.0,"tokens":1364665,"cost_usd":2.783643,"elapsed_time":674.6,"gpr":0.5,"tpe":0.4305,"ee":0.4286},{"task_id":"task_024_comprehensive_decision_hard_hard018","category":"comprehensive_decision","score":0.0,"tokens":207941,"cost_usd":0.52617,"elapsed_time":306.13,"gpr":0.2857,"tpe":0.81,"ee":0.5833},{"task_id":"task_025_comprehensive_decision_hard_hard019","category":"comprehensive_decision","score":1.0,"tokens":76687,"cost_usd":0.258014,"elapsed_time":103.29,"gpr":1.0,"tpe":1.0,"ee":1.5},{"task_id":"task_026_comprehensive_decision_medium_medium001","category":"comprehensive_decision","score":0.0,"tokens":828440,"cost_usd":1.226869,"elapsed_time":1207.59,"gpr":0.625,"tpe":0.2766,"ee":0.1389},{"task_id":"task_027_comprehensive_decision_medium_medium002","category":"comprehensive_decision","score":1.0,"tokens":846311,"cost_usd":1.729192,"elapsed_time":628.19,"gpr":1.0,"tpe":0.3988,"ee":0.2273},{"task_id":"task_028_comprehensive_decision_medium_medium003","category":"comprehensive_decision","score":0.0,"tokens":906537,"cost_usd":1.646886,"elapsed_time":1207.36,"gpr":0.1429,"tpe":0.3138,"ee":0.1667},{"task_id":"task_029_comprehensive_decision_medium_medium004","category":"comprehensive_decision","score":0.0,"tokens":590722,"cost_usd":1.1048,"elapsed_time":1207.45,"gpr":0.625,"tpe":0.351,"ee":0.1471},{"task_id":"task_030_comprehensive_decision_medium_medium005","category":"comprehensive_decision","score":1.0,"tokens":444255,"cost_usd":0.825234,"elapsed_time":543.4,"gpr":1.0,"tpe":0.3487,"ee":0.2083},{"task_id":"task_031_comprehensive_decision_medium_medium006","category":"comprehensive_decision","score":1.0,"tokens":308623,"cost_usd":0.486461,"elapsed_time":638.99,"gpr":1.0,"tpe":0.1956,"ee":0.1739},{"task_id":"task_032_comprehensive_decision_medium_medium007","category":"comprehensive_decision","score":0.0,"tokens":613929,"cost_usd":0.880929,"elapsed_time":910.54,"gpr":0.125,"tpe":0.4783,"ee":0.1667},{"task_id":"task_033_comprehensive_decision_medium_medium008","category":"comprehensive_decision","score":1.0,"tokens":633331,"cost_usd":1.095275,"elapsed_time":1133.9,"gpr":1.0,"tpe":0.3048,"ee":0.1515},{"task_id":"task_034_comprehensive_decision_medium_medium009","category":"comprehensive_decision","score":1.0,"tokens":216246,"cost_usd":0.471869,"elapsed_time":307.87,"gpr":1.0,"tpe":0.4641,"ee":0.3846},{"task_id":"task_035_comprehensive_decision_medium_medium010","category":"comprehensive_decision","score":0.5,"tokens":2151183,"cost_usd":3.003018,"elapsed_time":1121.52,"gpr":0.5,"tpe":0.4479,"ee":0.0877},{"task_id":"task_036_comprehensive_decision_medium_medium011","category":"comprehensive_decision","score":1.0,"tokens":410364,"cost_usd":0.696655,"elapsed_time":1028.08,"gpr":1.0,"tpe":0.2857,"ee":0.2174},{"task_id":"task_037_comprehensive_decision_medium_medium012","category":"comprehensive_decision","score":0.0,"tokens":1295727,"cost_usd":1.827217,"elapsed_time":1207.36,"gpr":0.0,"tpe":0.0,"ee":0.1064},{"task_id":"task_038_comprehensive_decision_medium_medium013","category":"comprehensive_decision","score":0.0,"tokens":469081,"cost_usd":0.767506,"elapsed_time":1207.46,"gpr":0.6,"tpe":0.4229,"ee":0.1852},{"task_id":"task_039_comprehensive_decision_medium_medium014","category":"comprehensive_decision","score":0.0,"tokens":442236,"cost_usd":0.539189,"elapsed_time":1207.48,"gpr":0.3333,"tpe":0.855,"ee":0.7143},{"task_id":"task_040_comprehensive_decision_medium_medium015","category":"comprehensive_decision","score":0.0,"tokens":313356,"cost_usd":0.595843,"elapsed_time":1207.48,"gpr":0.5,"tpe":0.95,"ee":0.2941},{"task_id":"task_041_comprehensive_decision_medium_medium016","category":"comprehensive_decision","score":0.0,"tokens":463903,"cost_usd":0.872255,"elapsed_time":1207.4,"gpr":0.0,"tpe":0.0,"ee":0.1562},{"task_id":"task_042_comprehensive_decision_medium_medium017","category":"comprehensive_decision","score":1.0,"tokens":474109,"cost_usd":1.002802,"elapsed_time":961.87,"gpr":1.0,"tpe":0.0581,"ee":0.1562},{"task_id":"task_043_comprehensive_decision_medium_medium018","category":"comprehensive_decision","score":0.0,"tokens":661415,"cost_usd":1.090954,"elapsed_time":1207.52,"gpr":0.2857,"tpe":0.4783,"ee":0.3571},{"task_id":"task_044_comprehensive_decision_medium_medium019","category":"comprehensive_decision","score":1.0,"tokens":344924,"cost_usd":0.63651,"elapsed_time":594.08,"gpr":1.0,"tpe":0.1668,"ee":0.2174},{"task_id":"task_045_comprehensive_decision_medium_medium020","category":"comprehensive_decision","score":0.0,"tokens":966880,"cost_usd":1.725171,"elapsed_time":1207.39,"gpr":0.0,"tpe":0.0,"ee":0.1},{"task_id":"task_046_comprehensive_decision_medium_medium021","category":"comprehensive_decision","score":0.0,"tokens":539384,"cost_usd":1.07668,"elapsed_time":756.29,"gpr":0.0,"tpe":0.0,"ee":0.1852},{"task_id":"task_047_comprehensive_decision_medium_medium022","category":"comprehensive_decision","score":1.0,"tokens":208839,"cost_usd":0.30984,"elapsed_time":163.31,"gpr":1.0,"tpe":0.3487,"ee":0.3333},{"task_id":"task_048_comprehensive_decision_medium_medium023","category":"comprehensive_decision","score":1.0,"tokens":418545,"cost_usd":0.719592,"elapsed_time":325.41,"gpr":1.0,"tpe":0.0804,"ee":0.1333},{"task_id":"task_049_comprehensive_decision_medium_medium024","category":"comprehensive_decision","score":1.0,"tokens":166080,"cost_usd":0.258692,"elapsed_time":210.06,"gpr":1.0,"tpe":0.5314,"ee":0.4545},{"task_id":"task_050_comprehensive_decision_medium_medium025","category":"comprehensive_decision","score":1.0,"tokens":893845,"cost_usd":1.384382,"elapsed_time":938.33,"gpr":1.0,"tpe":0.2265,"ee":0.1389},{"task_id":"task_051_comprehensive_decision_medium_medium026","category":"comprehensive_decision","score":1.0,"tokens":175017,"cost_usd":0.2793,"elapsed_time":142.54,"gpr":1.0,"tpe":0.6561,"ee":0.5556},{"task_id":"task_052_comprehensive_decision_medium_medium027","category":"comprehensive_decision","score":1.0,"tokens":3955264,"cost_usd":5.861597,"elapsed_time":935.72,"gpr":1.0,"tpe":0.0582,"ee":0.1562},{"task_id":"task_053_comprehensive_decision_medium_medium028","category":"comprehensive_decision","score":0.29,"tokens":959122,"cost_usd":1.492858,"elapsed_time":1158.57,"gpr":0.5,"tpe":0.0097,"ee":0.102},{"task_id":"task_054_comprehensive_decision_medium_medium029","category":"comprehensive_decision","score":0.0,"tokens":512190,"cost_usd":0.804399,"elapsed_time":1207.46,"gpr":1.0,"tpe":0.267,"ee":0.1515},{"task_id":"task_055_comprehensive_decision_medium_medium030","category":"comprehensive_decision","score":1.0,"tokens":224024,"cost_usd":0.374698,"elapsed_time":215.35,"gpr":1.0,"tpe":0.3941,"ee":0.3125},{"task_id":"task_056_comprehensive_decision_medium_medium031","category":"comprehensive_decision","score":1.0,"tokens":186941,"cost_usd":0.349817,"elapsed_time":290.83,"gpr":1.0,"tpe":0.4589,"ee":0.2667},{"task_id":"task_057_comprehensive_decision_medium_medium032","category":"comprehensive_decision","score":0.0,"tokens":820712,"cost_usd":1.530718,"elapsed_time":1021.84,"gpr":0.5714,"tpe":0.6441,"ee":0.1724},{"task_id":"task_058_comprehensive_decision_medium_medium033","category":"comprehensive_decision","score":1.0,"tokens":267735,"cost_usd":0.426584,"elapsed_time":724.31,"gpr":1.0,"tpe":0.472,"ee":0.3125},{"task_id":"task_059_comprehensive_decision_medium_medium034","category":"comprehensive_decision","score":0.0,"tokens":2107359,"cost_usd":3.185993,"elapsed_time":1207.38,"gpr":0.0,"tpe":0.0,"ee":0.2778},{"task_id":"task_060_comprehensive_decision_medium_medium035","category":"comprehensive_decision","score":0.0,"tokens":732495,"cost_usd":1.115203,"elapsed_time":1207.33,"gpr":0.4444,"tpe":0.6714,"ee":0.1562},{"task_id":"task_061_comprehensive_decision_medium_medium036","category":"comprehensive_decision","score":1.0,"tokens":722762,"cost_usd":1.292358,"elapsed_time":1195.33,"gpr":1.0,"tpe":0.0886,"ee":0.1667},{"task_id":"task_062_comprehensive_decision_medium_medium037","category":"comprehensive_decision","score":0.0,"tokens":762678,"cost_usd":1.052801,"elapsed_time":1207.58,"gpr":1.0,"tpe":0.1501,"ee":0.1667},{"task_id":"task_063_comprehensive_decision_medium_medium038","category":"comprehensive_decision","score":0.0,"tokens":182406,"cost_usd":0.308543,"elapsed_time":1207.47,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_064_comprehensive_decision_medium_medium039","category":"comprehensive_decision","score":1.0,"tokens":1374145,"cost_usd":1.718359,"elapsed_time":975.75,"gpr":1.0,"tpe":0.0798,"ee":0.1429},{"task_id":"task_065_comprehensive_decision_medium_medium040","category":"comprehensive_decision","score":0.0,"tokens":763215,"cost_usd":1.248389,"elapsed_time":491.28,"gpr":0.3333,"tpe":0.2288,"ee":0.2},{"task_id":"task_066_comprehensive_decision_medium_medium041","category":"comprehensive_decision","score":1.0,"tokens":194772,"cost_usd":0.379928,"elapsed_time":360.74,"gpr":1.0,"tpe":0.4305,"ee":0.3846},{"task_id":"task_067_comprehensive_decision_medium_medium042","category":"comprehensive_decision","score":0.0,"tokens":320248,"cost_usd":0.597985,"elapsed_time":522.02,"gpr":0.0,"tpe":0.0,"ee":0.2083},{"task_id":"task_068_comprehensive_decision_medium_medium043","category":"comprehensive_decision","score":1.0,"tokens":559505,"cost_usd":0.873057,"elapsed_time":650.78,"gpr":1.0,"tpe":0.3874,"ee":0.3077},{"task_id":"task_069_comprehensive_decision_medium_medium044","category":"comprehensive_decision","score":0.0,"tokens":1044812,"cost_usd":1.415377,"elapsed_time":1207.33,"gpr":0.25,"tpe":0.0985,"ee":0.1282},{"task_id":"task_070_comprehensive_decision_medium_medium045","category":"comprehensive_decision","score":0.0,"tokens":415425,"cost_usd":0.546484,"elapsed_time":1207.33,"gpr":0.25,"tpe":0.0886,"ee":0.1613},{"task_id":"task_071_enterprise_industry_analysis_easy_easy001","category":"enterprise_industry_analysis","score":1.0,"tokens":331368,"cost_usd":0.548814,"elapsed_time":677.66,"gpr":1.0,"tpe":0.2131,"ee":0.1304},{"task_id":"task_072_enterprise_industry_analysis_easy_easy002","category":"enterprise_industry_analysis","score":0.0,"tokens":480384,"cost_usd":0.71329,"elapsed_time":1207.56,"gpr":0.75,"tpe":0.1908,"ee":0.0938},{"task_id":"task_073_enterprise_industry_analysis_easy_easy003","category":"enterprise_industry_analysis","score":1.0,"tokens":509193,"cost_usd":0.707591,"elapsed_time":1001.23,"gpr":1.0,"tpe":0.1216,"ee":0.1304},{"task_id":"task_074_enterprise_industry_analysis_easy_easy004","category":"enterprise_industry_analysis","score":1.0,"tokens":101137,"cost_usd":0.180652,"elapsed_time":282.36,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_075_enterprise_industry_analysis_easy_easy005","category":"enterprise_industry_analysis","score":1.0,"tokens":905569,"cost_usd":1.614338,"elapsed_time":454.82,"gpr":1.0,"tpe":0.5031,"ee":0.25},{"task_id":"task_076_enterprise_industry_analysis_easy_easy006","category":"enterprise_industry_analysis","score":1.0,"tokens":165162,"cost_usd":0.286836,"elapsed_time":989.29,"gpr":1.0,"tpe":0.5073,"ee":0.25},{"task_id":"task_077_enterprise_industry_analysis_easy_easy007","category":"enterprise_industry_analysis","score":0.0,"tokens":185544,"cost_usd":0.326351,"elapsed_time":879.16,"gpr":0.0,"tpe":0.0,"ee":0.1875},{"task_id":"task_078_enterprise_industry_analysis_easy_easy008","category":"enterprise_industry_analysis","score":1.0,"tokens":89711,"cost_usd":0.169811,"elapsed_time":592.04,"gpr":1.0,"tpe":0.6585,"ee":0.375},{"task_id":"task_079_enterprise_industry_analysis_easy_easy009","category":"enterprise_industry_analysis","score":1.0,"tokens":119271,"cost_usd":0.202343,"elapsed_time":331.06,"gpr":1.0,"tpe":0.7155,"ee":0.375},{"task_id":"task_080_enterprise_industry_analysis_easy_easy010","category":"enterprise_industry_analysis","score":1.0,"tokens":122141,"cost_usd":0.200096,"elapsed_time":398.44,"gpr":1.0,"tpe":0.7887,"ee":0.4286},{"task_id":"task_081_enterprise_industry_analysis_easy_easy011","category":"enterprise_industry_analysis","score":0.0,"tokens":3017543,"cost_usd":5.28656,"elapsed_time":1207.33,"gpr":0.0,"tpe":0.0,"ee":0.1034},{"task_id":"task_082_enterprise_industry_analysis_easy_easy012","category":"enterprise_industry_analysis","score":1.0,"tokens":88862,"cost_usd":0.15527,"elapsed_time":292.19,"gpr":1.0,"tpe":0.81,"ee":0.5},{"task_id":"task_083_enterprise_industry_analysis_easy_easy013","category":"enterprise_industry_analysis","score":1.0,"tokens":698456,"cost_usd":1.298297,"elapsed_time":1052.7,"gpr":1.0,"tpe":0.0718,"ee":0.1071},{"task_id":"task_084_enterprise_industry_analysis_easy_easy014","category":"enterprise_industry_analysis","score":1.0,"tokens":105752,"cost_usd":0.230394,"elapsed_time":412.0,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_085_enterprise_industry_analysis_easy_easy015","category":"enterprise_industry_analysis","score":0.0,"tokens":115692,"cost_usd":0.188708,"elapsed_time":1207.41,"gpr":0.3333,"tpe":0.4783,"ee":0.3},{"task_id":"task_086_enterprise_industry_analysis_easy_easy016","category":"enterprise_industry_analysis","score":1.0,"tokens":162622,"cost_usd":0.293923,"elapsed_time":578.1,"gpr":1.0,"tpe":0.4321,"ee":0.25},{"task_id":"task_087_enterprise_industry_analysis_easy_easy017","category":"enterprise_industry_analysis","score":0.0,"tokens":141466,"cost_usd":0.219218,"elapsed_time":1207.4,"gpr":0.6667,"tpe":0.828,"ee":0.2727},{"task_id":"task_088_enterprise_industry_analysis_easy_easy018","category":"enterprise_industry_analysis","score":0.0,"tokens":567435,"cost_usd":1.071313,"elapsed_time":990.55,"gpr":0.0,"tpe":0.0,"ee":0.15},{"task_id":"task_089_enterprise_industry_analysis_easy_easy019","category":"enterprise_industry_analysis","score":1.0,"tokens":121830,"cost_usd":0.220435,"elapsed_time":228.46,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_090_enterprise_industry_analysis_easy_easy020","category":"enterprise_industry_analysis","score":0.0,"tokens":187837,"cost_usd":0.353636,"elapsed_time":1207.45,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_091_enterprise_industry_analysis_easy_easy021","category":"enterprise_industry_analysis","score":0.0,"tokens":133732,"cost_usd":0.213266,"elapsed_time":1207.42,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_092_enterprise_industry_analysis_easy_easy022","category":"enterprise_industry_analysis","score":1.0,"tokens":159539,"cost_usd":0.299183,"elapsed_time":683.9,"gpr":1.0,"tpe":0.5318,"ee":0.2308},{"task_id":"task_093_enterprise_industry_analysis_easy_easy023","category":"enterprise_industry_analysis","score":0.0,"tokens":280646,"cost_usd":0.428236,"elapsed_time":1207.36,"gpr":0.0,"tpe":0.0,"ee":0.1364},{"task_id":"task_094_enterprise_industry_analysis_easy_easy024","category":"enterprise_industry_analysis","score":1.0,"tokens":86865,"cost_usd":0.153391,"elapsed_time":422.4,"gpr":1.0,"tpe":0.8763,"ee":0.5},{"task_id":"task_095_enterprise_industry_analysis_easy_easy025","category":"enterprise_industry_analysis","score":1.0,"tokens":201723,"cost_usd":0.370938,"elapsed_time":564.42,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_096_enterprise_industry_analysis_easy_easy026","category":"enterprise_industry_analysis","score":1.0,"tokens":176868,"cost_usd":0.263661,"elapsed_time":673.15,"gpr":1.0,"tpe":0.315,"ee":0.2},{"task_id":"task_097_enterprise_industry_analysis_easy_easy027","category":"enterprise_industry_analysis","score":1.0,"tokens":493039,"cost_usd":0.758571,"elapsed_time":952.52,"gpr":1.0,"tpe":0.1674,"ee":0.1429},{"task_id":"task_098_enterprise_industry_analysis_easy_easy028","category":"enterprise_industry_analysis","score":1.0,"tokens":185487,"cost_usd":0.318264,"elapsed_time":734.48,"gpr":1.0,"tpe":0.2824,"ee":0.2},{"task_id":"task_099_enterprise_industry_analysis_easy_easy029","category":"enterprise_industry_analysis","score":1.0,"tokens":103228,"cost_usd":0.168017,"elapsed_time":297.02,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_100_enterprise_industry_analysis_easy_easy030","category":"enterprise_industry_analysis","score":1.0,"tokens":211093,"cost_usd":0.382754,"elapsed_time":937.13,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_101_enterprise_industry_analysis_easy_easy031","category":"enterprise_industry_analysis","score":1.0,"tokens":230604,"cost_usd":0.417309,"elapsed_time":863.27,"gpr":1.0,"tpe":0.4422,"ee":0.2308},{"task_id":"task_102_enterprise_industry_analysis_easy_easy032","category":"enterprise_industry_analysis","score":1.0,"tokens":256242,"cost_usd":0.433921,"elapsed_time":895.79,"gpr":1.0,"tpe":0.5516,"ee":0.1875},{"task_id":"task_103_enterprise_industry_analysis_easy_easy033","category":"enterprise_industry_analysis","score":0.0,"tokens":57857,"cost_usd":0.145226,"elapsed_time":211.69,"gpr":0.3333,"tpe":1.0,"ee":0.75},{"task_id":"task_104_enterprise_industry_analysis_easy_easy034","category":"enterprise_industry_analysis","score":0.0,"tokens":20606,"cost_usd":0.033056,"elapsed_time":564.75,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_105_enterprise_industry_analysis_easy_easy035","category":"enterprise_industry_analysis","score":1.0,"tokens":79721,"cost_usd":0.128665,"elapsed_time":631.3,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_106_enterprise_industry_analysis_easy_easy036","category":"enterprise_industry_analysis","score":1.0,"tokens":144823,"cost_usd":0.267381,"elapsed_time":580.3,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_107_enterprise_industry_analysis_easy_easy037","category":"enterprise_industry_analysis","score":1.0,"tokens":90256,"cost_usd":0.159603,"elapsed_time":569.36,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_108_enterprise_industry_analysis_easy_easy038","category":"enterprise_industry_analysis","score":0.0,"tokens":125881,"cost_usd":0.269199,"elapsed_time":480.32,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_109_enterprise_industry_analysis_easy_easy039","category":"enterprise_industry_analysis","score":1.0,"tokens":49413,"cost_usd":0.123823,"elapsed_time":298.2,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_110_enterprise_industry_analysis_easy_easy040","category":"enterprise_industry_analysis","score":1.0,"tokens":70541,"cost_usd":0.133496,"elapsed_time":242.17,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_111_enterprise_industry_analysis_easy_easy041","category":"enterprise_industry_analysis","score":1.0,"tokens":57729,"cost_usd":0.118591,"elapsed_time":130.45,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_112_enterprise_industry_analysis_easy_easy042","category":"enterprise_industry_analysis","score":1.0,"tokens":98137,"cost_usd":0.171136,"elapsed_time":484.68,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_113_enterprise_industry_analysis_easy_easy043","category":"enterprise_industry_analysis","score":1.0,"tokens":69807,"cost_usd":0.151964,"elapsed_time":64.09,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_114_enterprise_industry_analysis_easy_easy044","category":"enterprise_industry_analysis","score":1.0,"tokens":80653,"cost_usd":0.143725,"elapsed_time":386.76,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_115_enterprise_industry_analysis_easy_easy045","category":"enterprise_industry_analysis","score":1.0,"tokens":61433,"cost_usd":0.119153,"elapsed_time":313.57,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_116_enterprise_industry_analysis_easy_easy046","category":"enterprise_industry_analysis","score":1.0,"tokens":188780,"cost_usd":0.325285,"elapsed_time":585.79,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_117_enterprise_industry_analysis_easy_easy047","category":"enterprise_industry_analysis","score":1.0,"tokens":139778,"cost_usd":0.279446,"elapsed_time":404.24,"gpr":1.0,"tpe":0.3874,"ee":0.25},{"task_id":"task_118_enterprise_industry_analysis_easy_easy048","category":"enterprise_industry_analysis","score":0.0,"tokens":33889,"cost_usd":0.073848,"elapsed_time":333.34,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_119_enterprise_industry_analysis_easy_easy049","category":"enterprise_industry_analysis","score":1.0,"tokens":228831,"cost_usd":0.468362,"elapsed_time":487.58,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_120_enterprise_industry_analysis_easy_easy050","category":"enterprise_industry_analysis","score":1.0,"tokens":46995,"cost_usd":0.086933,"elapsed_time":320.88,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_121_enterprise_industry_analysis_easy_easy051","category":"enterprise_industry_analysis","score":1.0,"tokens":98938,"cost_usd":0.196355,"elapsed_time":394.0,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_122_enterprise_industry_analysis_easy_easy052","category":"enterprise_industry_analysis","score":1.0,"tokens":101502,"cost_usd":0.171343,"elapsed_time":626.38,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_123_enterprise_industry_analysis_easy_easy053","category":"enterprise_industry_analysis","score":1.0,"tokens":71325,"cost_usd":0.141533,"elapsed_time":249.72,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_124_enterprise_industry_analysis_easy_easy054","category":"enterprise_industry_analysis","score":1.0,"tokens":66872,"cost_usd":0.13425,"elapsed_time":173.55,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_125_enterprise_industry_analysis_easy_easy055","category":"enterprise_industry_analysis","score":1.0,"tokens":32356,"cost_usd":0.071079,"elapsed_time":123.15,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_126_enterprise_industry_analysis_easy_easy056","category":"enterprise_industry_analysis","score":0.0,"tokens":1033719,"cost_usd":1.966141,"elapsed_time":1207.51,"gpr":0.0,"tpe":0.0,"ee":0.1364},{"task_id":"task_127_enterprise_industry_analysis_easy_easy057","category":"enterprise_industry_analysis","score":1.0,"tokens":68587,"cost_usd":0.142596,"elapsed_time":525.16,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_128_enterprise_industry_analysis_easy_easy058","category":"enterprise_industry_analysis","score":1.0,"tokens":37083,"cost_usd":0.063643,"elapsed_time":543.13,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_129_enterprise_industry_analysis_easy_easy059","category":"enterprise_industry_analysis","score":1.0,"tokens":348413,"cost_usd":0.75218,"elapsed_time":548.4,"gpr":1.0,"tpe":0.2824,"ee":0.2},{"task_id":"task_130_enterprise_industry_analysis_easy_easy060","category":"enterprise_industry_analysis","score":1.0,"tokens":105661,"cost_usd":0.231312,"elapsed_time":588.22,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_131_enterprise_industry_analysis_easy_easy061","category":"enterprise_industry_analysis","score":0.0,"tokens":4906872,"cost_usd":6.7078,"elapsed_time":1207.55,"gpr":0.0,"tpe":0.0,"ee":0.1579},{"task_id":"task_132_enterprise_industry_analysis_easy_easy062","category":"enterprise_industry_analysis","score":0.0,"tokens":60622,"cost_usd":0.114321,"elapsed_time":720.98,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_133_enterprise_industry_analysis_easy_easy063","category":"enterprise_industry_analysis","score":1.0,"tokens":741713,"cost_usd":1.458557,"elapsed_time":1001.85,"gpr":1.0,"tpe":0.3138,"ee":0.2143},{"task_id":"task_134_enterprise_industry_analysis_easy_easy064","category":"enterprise_industry_analysis","score":0.0,"tokens":282053,"cost_usd":0.469113,"elapsed_time":1207.52,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_135_enterprise_industry_analysis_easy_easy065","category":"enterprise_industry_analysis","score":0.0,"tokens":129336,"cost_usd":0.225428,"elapsed_time":1207.38,"gpr":0.6667,"tpe":0.81,"ee":0.3},{"task_id":"task_136_enterprise_industry_analysis_easy_easy066","category":"enterprise_industry_analysis","score":0.0,"tokens":212333,"cost_usd":0.402964,"elapsed_time":1207.38,"gpr":0.6667,"tpe":0.3312,"ee":0.1875},{"task_id":"task_137_enterprise_industry_analysis_easy_easy067","category":"enterprise_industry_analysis","score":0.0,"tokens":186695,"cost_usd":0.369104,"elapsed_time":1207.42,"gpr":1.0,"tpe":0.5104,"ee":0.2},{"task_id":"task_138_enterprise_industry_analysis_easy_easy068","category":"enterprise_industry_analysis","score":1.0,"tokens":1914676,"cost_usd":2.983852,"elapsed_time":1143.52,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_139_enterprise_industry_analysis_easy_easy069","category":"enterprise_industry_analysis","score":0.0,"tokens":2496265,"cost_usd":4.374688,"elapsed_time":1033.7,"gpr":0.0,"tpe":0.0,"ee":0.15},{"task_id":"task_140_enterprise_industry_analysis_easy_easy070","category":"enterprise_industry_analysis","score":1.0,"tokens":148975,"cost_usd":0.31733,"elapsed_time":137.13,"gpr":1.0,"tpe":0.6804,"ee":0.4286},{"task_id":"task_141_enterprise_industry_analysis_easy_easy071","category":"enterprise_industry_analysis","score":1.0,"tokens":182238,"cost_usd":0.345624,"elapsed_time":581.58,"gpr":1.0,"tpe":0.2288,"ee":0.1765},{"task_id":"task_142_enterprise_industry_analysis_easy_easy072","category":"enterprise_industry_analysis","score":0.0,"tokens":144848,"cost_usd":0.313976,"elapsed_time":431.76,"gpr":0.6667,"tpe":0.5314,"ee":0.25},{"task_id":"task_143_enterprise_industry_analysis_easy_easy073","category":"enterprise_industry_analysis","score":0.0,"tokens":110235,"cost_usd":0.211407,"elapsed_time":1207.38,"gpr":0.3333,"tpe":0.6561,"ee":0.4286},{"task_id":"task_144_enterprise_industry_analysis_easy_easy074","category":"enterprise_industry_analysis","score":1.0,"tokens":131114,"cost_usd":0.282278,"elapsed_time":147.94,"gpr":1.0,"tpe":0.496,"ee":0.3},{"task_id":"task_145_enterprise_industry_analysis_easy_easy075","category":"enterprise_industry_analysis","score":1.0,"tokens":79573,"cost_usd":0.164103,"elapsed_time":274.75,"gpr":1.0,"tpe":0.813,"ee":0.5},{"task_id":"task_146_enterprise_industry_analysis_easy_easy076","category":"enterprise_industry_analysis","score":1.0,"tokens":76788,"cost_usd":0.184336,"elapsed_time":132.01,"gpr":1.0,"tpe":0.7047,"ee":0.4286},{"task_id":"task_147_enterprise_industry_analysis_easy_easy077","category":"enterprise_industry_analysis","score":0.0,"tokens":85620,"cost_usd":0.166099,"elapsed_time":1207.4,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_148_enterprise_industry_analysis_easy_easy078","category":"enterprise_industry_analysis","score":1.0,"tokens":45190,"cost_usd":0.10819,"elapsed_time":52.61,"gpr":1.0,"tpe":1.0,"ee":1.5},{"task_id":"task_149_enterprise_industry_analysis_easy_easy079","category":"enterprise_industry_analysis","score":1.0,"tokens":657379,"cost_usd":1.314919,"elapsed_time":554.25,"gpr":1.0,"tpe":0.5175,"ee":0.2727},{"task_id":"task_150_enterprise_industry_analysis_easy_easy080","category":"enterprise_industry_analysis","score":1.0,"tokens":46578,"cost_usd":0.084815,"elapsed_time":463.03,"gpr":1.0,"tpe":0.9667,"ee":0.75},{"task_id":"task_151_enterprise_industry_analysis_easy_easy081","category":"enterprise_industry_analysis","score":0.0,"tokens":34408,"cost_usd":0.078906,"elapsed_time":1207.4,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_152_enterprise_industry_analysis_easy_easy082","category":"enterprise_industry_analysis","score":1.0,"tokens":139661,"cost_usd":0.250009,"elapsed_time":876.23,"gpr":1.0,"tpe":0.7771,"ee":0.3333},{"task_id":"task_153_enterprise_industry_analysis_easy_easy083","category":"enterprise_industry_analysis","score":1.0,"tokens":156196,"cost_usd":0.315399,"elapsed_time":499.9,"gpr":1.0,"tpe":0.6145,"ee":0.3333},{"task_id":"task_154_enterprise_industry_analysis_easy_easy084","category":"enterprise_industry_analysis","score":0.0,"tokens":122213,"cost_usd":0.201675,"elapsed_time":1207.46,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_155_enterprise_industry_analysis_easy_easy085","category":"enterprise_industry_analysis","score":0.0,"tokens":68930,"cost_usd":0.133387,"elapsed_time":1207.32,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_156_enterprise_industry_analysis_easy_easy086","category":"enterprise_industry_analysis","score":0.0,"tokens":99320,"cost_usd":0.18192,"elapsed_time":868.42,"gpr":0.6667,"tpe":0.6561,"ee":0.375},{"task_id":"task_157_enterprise_industry_analysis_easy_easy088","category":"enterprise_industry_analysis","score":0.0,"tokens":386394,"cost_usd":0.627168,"elapsed_time":1207.39,"gpr":0.0,"tpe":0.0,"ee":0.069},{"task_id":"task_158_enterprise_industry_analysis_easy_easy090","category":"enterprise_industry_analysis","score":0.0,"tokens":233281,"cost_usd":0.466892,"elapsed_time":1207.38,"gpr":0.6667,"tpe":0.4783,"ee":0.2727},{"task_id":"task_159_enterprise_industry_analysis_easy_easy091","category":"enterprise_industry_analysis","score":0.0,"tokens":151214,"cost_usd":0.288467,"elapsed_time":835.92,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_160_enterprise_industry_analysis_easy_easy092","category":"enterprise_industry_analysis","score":0.0,"tokens":322125,"cost_usd":0.642125,"elapsed_time":879.28,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_161_enterprise_industry_analysis_easy_easy093","category":"enterprise_industry_analysis","score":1.0,"tokens":171044,"cost_usd":0.302793,"elapsed_time":471.1,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_162_enterprise_industry_analysis_easy_easy094","category":"enterprise_industry_analysis","score":0.0,"tokens":329018,"cost_usd":0.579772,"elapsed_time":299.68,"gpr":0.0,"tpe":0.0,"ee":0.2143},{"task_id":"task_163_enterprise_industry_analysis_easy_easy095","category":"enterprise_industry_analysis","score":1.0,"tokens":58119,"cost_usd":0.114891,"elapsed_time":243.65,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_164_enterprise_industry_analysis_easy_easy096","category":"enterprise_industry_analysis","score":1.0,"tokens":75775,"cost_usd":0.121774,"elapsed_time":126.88,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_165_enterprise_industry_analysis_easy_easy097","category":"enterprise_industry_analysis","score":1.0,"tokens":71222,"cost_usd":0.09997,"elapsed_time":355.35,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_166_enterprise_industry_analysis_easy_easy098","category":"enterprise_industry_analysis","score":1.0,"tokens":242703,"cost_usd":0.424818,"elapsed_time":413.76,"gpr":1.0,"tpe":0.1501,"ee":0.1429},{"task_id":"task_167_enterprise_industry_analysis_easy_easy099","category":"enterprise_industry_analysis","score":1.0,"tokens":99743,"cost_usd":0.196166,"elapsed_time":120.47,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_168_enterprise_industry_analysis_easy_easy100","category":"enterprise_industry_analysis","score":1.0,"tokens":1725682,"cost_usd":1.742501,"elapsed_time":506.66,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_169_enterprise_industry_analysis_easy_easy101","category":"enterprise_industry_analysis","score":1.0,"tokens":53395,"cost_usd":0.102009,"elapsed_time":78.74,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_170_enterprise_industry_analysis_easy_easy102","category":"enterprise_industry_analysis","score":1.0,"tokens":32557,"cost_usd":0.060892,"elapsed_time":222.38,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_171_enterprise_industry_analysis_easy_easy103","category":"enterprise_industry_analysis","score":1.0,"tokens":2044365,"cost_usd":3.070331,"elapsed_time":499.81,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_172_enterprise_industry_analysis_easy_easy104","category":"enterprise_industry_analysis","score":1.0,"tokens":29318,"cost_usd":0.05101,"elapsed_time":121.85,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_173_enterprise_industry_analysis_easy_easy105","category":"enterprise_industry_analysis","score":0.0,"tokens":4961856,"cost_usd":6.60792,"elapsed_time":1207.34,"gpr":0.0,"tpe":0.0,"ee":0.125},{"task_id":"task_174_enterprise_industry_analysis_easy_easy106","category":"enterprise_industry_analysis","score":0.0,"tokens":5105949,"cost_usd":6.800637,"elapsed_time":1207.36,"gpr":0.0,"tpe":0.0,"ee":0.1154},{"task_id":"task_175_enterprise_industry_analysis_easy_easy107","category":"enterprise_industry_analysis","score":1.0,"tokens":34655,"cost_usd":0.066927,"elapsed_time":60.35,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_176_enterprise_industry_analysis_easy_easy108","category":"enterprise_industry_analysis","score":1.0,"tokens":61246,"cost_usd":0.132452,"elapsed_time":84.41,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_177_enterprise_industry_analysis_easy_easy109","category":"enterprise_industry_analysis","score":0.0,"tokens":334907,"cost_usd":0.567707,"elapsed_time":166.18,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_178_enterprise_industry_analysis_easy_easy110","category":"enterprise_industry_analysis","score":1.0,"tokens":34143,"cost_usd":0.077386,"elapsed_time":52.52,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_179_enterprise_industry_analysis_easy_easy111","category":"enterprise_industry_analysis","score":1.0,"tokens":57317,"cost_usd":0.10968,"elapsed_time":80.7,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_180_enterprise_industry_analysis_easy_easy112","category":"enterprise_industry_analysis","score":1.0,"tokens":588393,"cost_usd":1.032297,"elapsed_time":1130.15,"gpr":1.0,"tpe":0.0562,"ee":0.0968},{"task_id":"task_181_enterprise_industry_analysis_easy_easy113","category":"enterprise_industry_analysis","score":0.0,"tokens":6888021,"cost_usd":8.820096,"elapsed_time":1207.77,"gpr":0.0,"tpe":0.0,"ee":0.0833},{"task_id":"task_182_enterprise_industry_analysis_easy_easy114","category":"enterprise_industry_analysis","score":1.0,"tokens":317841,"cost_usd":0.583767,"elapsed_time":693.59,"gpr":1.0,"tpe":0.3138,"ee":0.2143},{"task_id":"task_183_enterprise_industry_analysis_easy_easy115","category":"enterprise_industry_analysis","score":0.0,"tokens":222651,"cost_usd":0.440543,"elapsed_time":399.55,"gpr":0.0,"tpe":0.0,"ee":0.2143},{"task_id":"task_184_enterprise_industry_analysis_easy_easy116","category":"enterprise_industry_analysis","score":1.0,"tokens":55367,"cost_usd":0.110778,"elapsed_time":204.93,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_185_enterprise_industry_analysis_easy_easy117","category":"enterprise_industry_analysis","score":1.0,"tokens":84476,"cost_usd":0.141135,"elapsed_time":212.21,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_186_enterprise_industry_analysis_medium_medium001","category":"enterprise_industry_analysis","score":1.0,"tokens":296436,"cost_usd":0.523536,"elapsed_time":959.84,"gpr":1.0,"tpe":0.2059,"ee":0.2105},{"task_id":"task_187_enterprise_industry_analysis_medium_medium002","category":"enterprise_industry_analysis","score":1.0,"tokens":326698,"cost_usd":0.65412,"elapsed_time":394.03,"gpr":1.0,"tpe":0.3684,"ee":0.2105},{"task_id":"task_188_enterprise_industry_analysis_medium_medium003","category":"enterprise_industry_analysis","score":0.0,"tokens":909361,"cost_usd":1.548764,"elapsed_time":1207.42,"gpr":0.25,"tpe":0.5314,"ee":0.0851},{"task_id":"task_189_enterprise_industry_analysis_medium_medium004","category":"enterprise_industry_analysis","score":1.0,"tokens":172649,"cost_usd":0.317343,"elapsed_time":156.4,"gpr":1.0,"tpe":0.4305,"ee":0.3333},{"task_id":"task_190_enterprise_industry_analysis_medium_medium005","category":"enterprise_industry_analysis","score":1.0,"tokens":2869639,"cost_usd":5.183215,"elapsed_time":716.18,"gpr":1.0,"tpe":0.1501,"ee":0.1818},{"task_id":"task_191_enterprise_industry_analysis_medium_medium006","category":"enterprise_industry_analysis","score":1.0,"tokens":841248,"cost_usd":1.682055,"elapsed_time":443.92,"gpr":1.0,"tpe":0.4783,"ee":0.3636},{"task_id":"task_192_enterprise_industry_analysis_medium_medium007","category":"enterprise_industry_analysis","score":0.0,"tokens":236412,"cost_usd":0.395397,"elapsed_time":409.6,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_193_enterprise_industry_analysis_medium_medium008","category":"enterprise_industry_analysis","score":1.0,"tokens":542939,"cost_usd":0.762318,"elapsed_time":472.55,"gpr":1.0,"tpe":0.4102,"ee":0.2857},{"task_id":"task_194_enterprise_industry_analysis_medium_medium009","category":"enterprise_industry_analysis","score":0.0,"tokens":330487,"cost_usd":0.61184,"elapsed_time":1207.45,"gpr":0.0,"tpe":0.0,"ee":0.1818},{"task_id":"task_195_enterprise_industry_analysis_medium_medium010","category":"enterprise_industry_analysis","score":0.0,"tokens":165199,"cost_usd":0.329504,"elapsed_time":202.95,"gpr":0.0,"tpe":0.0,"ee":0.4444},{"task_id":"task_196_enterprise_industry_analysis_medium_medium011","category":"enterprise_industry_analysis","score":0.0,"tokens":303836,"cost_usd":0.512142,"elapsed_time":621.17,"gpr":0.4,"tpe":1.0,"ee":0.2667},{"task_id":"task_197_enterprise_industry_analysis_medium_medium012","category":"enterprise_industry_analysis","score":1.0,"tokens":98867,"cost_usd":0.166333,"elapsed_time":131.2,"gpr":1.0,"tpe":0.81,"ee":0.6667},{"task_id":"task_198_enterprise_industry_analysis_medium_medium013","category":"enterprise_industry_analysis","score":1.0,"tokens":165899,"cost_usd":0.314895,"elapsed_time":188.56,"gpr":1.0,"tpe":0.5314,"ee":0.4},{"task_id":"task_199_enterprise_industry_analysis_medium_medium014","category":"enterprise_industry_analysis","score":0.0,"tokens":11854463,"cost_usd":15.366431,"elapsed_time":1207.49,"gpr":0.4,"tpe":1.0,"ee":0.1724},{"task_id":"task_200_enterprise_industry_analysis_medium_medium015","category":"enterprise_industry_analysis","score":1.0,"tokens":2506849,"cost_usd":4.09158,"elapsed_time":882.87,"gpr":1.0,"tpe":0.0718,"ee":0.1667},{"task_id":"task_201_enterprise_industry_analysis_medium_medium016","category":"enterprise_industry_analysis","score":0.0,"tokens":218142,"cost_usd":0.391438,"elapsed_time":1207.4,"gpr":0.2,"tpe":0.729,"ee":0.25},{"task_id":"task_202_enterprise_industry_analysis_medium_medium017","category":"enterprise_industry_analysis","score":0.0,"tokens":1724723,"cost_usd":3.188075,"elapsed_time":1207.41,"gpr":0.0,"tpe":0.0,"ee":0.1923},{"task_id":"task_203_enterprise_industry_analysis_medium_medium018","category":"enterprise_industry_analysis","score":0.0,"tokens":368077,"cost_usd":0.551202,"elapsed_time":1207.37,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_204_enterprise_industry_analysis_medium_medium019","category":"enterprise_industry_analysis","score":1.0,"tokens":197787,"cost_usd":0.370739,"elapsed_time":906.9,"gpr":1.0,"tpe":0.5553,"ee":0.4167},{"task_id":"task_205_enterprise_industry_analysis_medium_medium020","category":"enterprise_industry_analysis","score":1.0,"tokens":94802,"cost_usd":0.208316,"elapsed_time":248.62,"gpr":1.0,"tpe":0.8698,"ee":0.625},{"task_id":"task_206_enterprise_industry_analysis_medium_medium021","category":"enterprise_industry_analysis","score":0.0,"tokens":303969,"cost_usd":0.556329,"elapsed_time":962.74,"gpr":0.8,"tpe":0.1853,"ee":0.2381},{"task_id":"task_207_enterprise_industry_analysis_medium_medium022","category":"enterprise_industry_analysis","score":1.0,"tokens":103317,"cost_usd":0.237851,"elapsed_time":436.0,"gpr":1.0,"tpe":0.9312,"ee":0.5556},{"task_id":"task_208_enterprise_industry_analysis_medium_medium023","category":"enterprise_industry_analysis","score":0.0,"tokens":621394,"cost_usd":1.173727,"elapsed_time":1207.38,"gpr":0.8,"tpe":0.468,"ee":0.125},{"task_id":"task_209_enterprise_industry_analysis_medium_medium024","category":"enterprise_industry_analysis","score":1.0,"tokens":5949634,"cost_usd":5.997931,"elapsed_time":1155.47,"gpr":1.0,"tpe":0.5336,"ee":0.1786},{"task_id":"task_210_enterprise_industry_analysis_medium_medium025","category":"enterprise_industry_analysis","score":1.0,"tokens":4590480,"cost_usd":7.772356,"elapsed_time":1013.59,"gpr":1.0,"tpe":0.4198,"ee":0.1163},{"task_id":"task_211_enterprise_industry_analysis_medium_medium026","category":"enterprise_industry_analysis","score":1.0,"tokens":250772,"cost_usd":0.394954,"elapsed_time":886.66,"gpr":1.0,"tpe":0.2558,"ee":0.25},{"task_id":"task_212_enterprise_industry_analysis_medium_medium027","category":"enterprise_industry_analysis","score":1.0,"tokens":139388,"cost_usd":0.27557,"elapsed_time":278.89,"gpr":1.0,"tpe":0.8957,"ee":0.4167},{"task_id":"task_213_enterprise_industry_analysis_medium_medium028","category":"enterprise_industry_analysis","score":1.0,"tokens":297891,"cost_usd":0.484281,"elapsed_time":389.77,"gpr":1.0,"tpe":0.4826,"ee":0.2},{"task_id":"task_214_enterprise_industry_analysis_medium_medium029","category":"enterprise_industry_analysis","score":1.0,"tokens":538856,"cost_usd":1.007582,"elapsed_time":1129.25,"gpr":1.0,"tpe":0.2697,"ee":0.122},{"task_id":"task_215_enterprise_industry_analysis_medium_medium030","category":"enterprise_industry_analysis","score":1.0,"tokens":949038,"cost_usd":1.893292,"elapsed_time":241.81,"gpr":1.0,"tpe":0.4783,"ee":0.4167},{"task_id":"task_216_enterprise_industry_analysis_medium_medium031","category":"enterprise_industry_analysis","score":1.0,"tokens":94469,"cost_usd":0.249321,"elapsed_time":474.18,"gpr":1.0,"tpe":0.8698,"ee":0.625},{"task_id":"task_217_enterprise_industry_analysis_medium_medium032","category":"enterprise_industry_analysis","score":1.0,"tokens":119862,"cost_usd":0.248859,"elapsed_time":435.67,"gpr":1.0,"tpe":0.7937,"ee":0.5556},{"task_id":"task_218_enterprise_industry_analysis_medium_medium033","category":"enterprise_industry_analysis","score":1.0,"tokens":187552,"cost_usd":0.371455,"elapsed_time":445.61,"gpr":1.0,"tpe":0.4134,"ee":0.3333},{"task_id":"task_219_enterprise_industry_analysis_medium_medium034","category":"enterprise_industry_analysis","score":1.0,"tokens":103063,"cost_usd":0.192886,"elapsed_time":399.35,"gpr":1.0,"tpe":0.6561,"ee":0.5556},{"task_id":"task_220_enterprise_industry_analysis_medium_medium035","category":"enterprise_industry_analysis","score":0.0,"tokens":326375,"cost_usd":0.629887,"elapsed_time":1207.45,"gpr":0.8,"tpe":0.6041,"ee":0.1852},{"task_id":"task_221_enterprise_industry_analysis_medium_medium036","category":"enterprise_industry_analysis","score":0.0,"tokens":1038269,"cost_usd":2.000955,"elapsed_time":1207.4,"gpr":0.0,"tpe":0.0,"ee":0.2174},{"task_id":"task_222_enterprise_industry_analysis_medium_medium037","category":"enterprise_industry_analysis","score":1.0,"tokens":298657,"cost_usd":0.47181,"elapsed_time":251.98,"gpr":1.0,"tpe":0.2059,"ee":0.25},{"task_id":"task_223_enterprise_industry_analysis_medium_medium038","category":"enterprise_industry_analysis","score":0.0,"tokens":294798,"cost_usd":0.593544,"elapsed_time":1207.44,"gpr":0.2,"tpe":0.1501,"ee":0.2},{"task_id":"task_224_enterprise_industry_analysis_medium_medium039","category":"enterprise_industry_analysis","score":0.0,"tokens":207765,"cost_usd":0.410145,"elapsed_time":138.04,"gpr":0.0,"tpe":0.0,"ee":0.625},{"task_id":"task_225_enterprise_industry_analysis_medium_medium040","category":"enterprise_industry_analysis","score":1.0,"tokens":154148,"cost_usd":0.369686,"elapsed_time":167.93,"gpr":1.0,"tpe":0.7602,"ee":0.5},{"task_id":"task_226_enterprise_industry_analysis_medium_medium041","category":"enterprise_industry_analysis","score":1.0,"tokens":299104,"cost_usd":0.509594,"elapsed_time":400.42,"gpr":1.0,"tpe":0.2824,"ee":0.2941},{"task_id":"task_227_enterprise_industry_analysis_medium_medium042","category":"enterprise_industry_analysis","score":0.0,"tokens":4604430,"cost_usd":6.421947,"elapsed_time":1129.28,"gpr":0.8,"tpe":0.5834,"ee":0.1087},{"task_id":"task_228_enterprise_industry_analysis_medium_medium043","category":"enterprise_industry_analysis","score":1.0,"tokens":317730,"cost_usd":0.698228,"elapsed_time":1080.15,"gpr":1.0,"tpe":0.7878,"ee":0.2},{"task_id":"task_229_enterprise_industry_analysis_medium_medium044","category":"enterprise_industry_analysis","score":1.0,"tokens":180134,"cost_usd":0.407597,"elapsed_time":742.88,"gpr":1.0,"tpe":0.8775,"ee":0.3571},{"task_id":"task_230_enterprise_industry_analysis_medium_medium045","category":"enterprise_industry_analysis","score":0.0,"tokens":132486,"cost_usd":0.438572,"elapsed_time":609.6,"gpr":0.2,"tpe":1.0,"ee":0.5},{"task_id":"task_231_enterprise_industry_analysis_medium_medium046","category":"enterprise_industry_analysis","score":1.0,"tokens":454154,"cost_usd":0.833413,"elapsed_time":981.98,"gpr":1.0,"tpe":0.1094,"ee":0.1923},{"task_id":"task_232_enterprise_industry_analysis_medium_medium047","category":"enterprise_industry_analysis","score":1.0,"tokens":442687,"cost_usd":0.727289,"elapsed_time":799.71,"gpr":1.0,"tpe":0.1676,"ee":0.1852},{"task_id":"task_233_enterprise_industry_analysis_medium_medium048","category":"enterprise_industry_analysis","score":1.0,"tokens":158598,"cost_usd":0.383479,"elapsed_time":638.75,"gpr":1.0,"tpe":0.4601,"ee":0.3571},{"task_id":"task_234_enterprise_industry_analysis_medium_medium049","category":"enterprise_industry_analysis","score":1.0,"tokens":214508,"cost_usd":0.415902,"elapsed_time":360.06,"gpr":1.0,"tpe":0.7206,"ee":0.2778},{"task_id":"task_235_enterprise_industry_analysis_medium_medium050","category":"enterprise_industry_analysis","score":0.0,"tokens":314942,"cost_usd":0.531947,"elapsed_time":1068.07,"gpr":0.8,"tpe":0.3961,"ee":0.2},{"task_id":"task_236_enterprise_industry_analysis_medium_medium051","category":"enterprise_industry_analysis","score":0.0,"tokens":7833009,"cost_usd":11.022509,"elapsed_time":1207.42,"gpr":0.4,"tpe":0.481,"ee":0.1282},{"task_id":"task_237_enterprise_industry_analysis_medium_medium052","category":"enterprise_industry_analysis","score":1.0,"tokens":642995,"cost_usd":1.32653,"elapsed_time":149.53,"gpr":1.0,"tpe":0.8552,"ee":0.5556},{"task_id":"task_238_enterprise_industry_analysis_medium_medium053","category":"enterprise_industry_analysis","score":1.0,"tokens":306653,"cost_usd":0.548985,"elapsed_time":312.13,"gpr":1.0,"tpe":0.2032,"ee":0.2381},{"task_id":"task_239_enterprise_industry_analysis_medium_medium054","category":"enterprise_industry_analysis","score":1.0,"tokens":2176200,"cost_usd":3.255539,"elapsed_time":867.26,"gpr":1.0,"tpe":0.4512,"ee":0.1786},{"task_id":"task_240_enterprise_industry_analysis_medium_medium055","category":"enterprise_industry_analysis","score":1.0,"tokens":256820,"cost_usd":0.499493,"elapsed_time":290.49,"gpr":1.0,"tpe":0.4623,"ee":0.2273},{"task_id":"task_241_enterprise_industry_analysis_medium_medium056","category":"enterprise_industry_analysis","score":1.0,"tokens":206147,"cost_usd":0.40368,"elapsed_time":292.24,"gpr":1.0,"tpe":0.2598,"ee":0.2778},{"task_id":"task_242_enterprise_industry_analysis_medium_medium057","category":"enterprise_industry_analysis","score":1.0,"tokens":336155,"cost_usd":0.608247,"elapsed_time":700.11,"gpr":1.0,"tpe":0.4691,"ee":0.1786},{"task_id":"task_243_enterprise_industry_analysis_medium_medium058","category":"enterprise_industry_analysis","score":0.0,"tokens":1600887,"cost_usd":2.453384,"elapsed_time":249.71,"gpr":1.0,"tpe":0.5955,"ee":0.25},{"task_id":"task_244_enterprise_industry_analysis_medium_medium059","category":"enterprise_industry_analysis","score":1.0,"tokens":210026,"cost_usd":0.401117,"elapsed_time":259.34,"gpr":1.0,"tpe":0.3608,"ee":0.2941},{"task_id":"task_245_enterprise_industry_analysis_medium_medium060","category":"enterprise_industry_analysis","score":0.0,"tokens":10021342,"cost_usd":13.571133,"elapsed_time":1207.38,"gpr":0.4,"tpe":0.6925,"ee":0.0633},{"task_id":"task_246_enterprise_industry_analysis_medium_medium061","category":"enterprise_industry_analysis","score":1.0,"tokens":3441424,"cost_usd":4.797371,"elapsed_time":953.89,"gpr":1.0,"tpe":0.4851,"ee":0.1351},{"task_id":"task_247_enterprise_industry_analysis_medium_medium062","category":"enterprise_industry_analysis","score":1.0,"tokens":255343,"cost_usd":0.507961,"elapsed_time":214.56,"gpr":1.0,"tpe":0.3674,"ee":0.3125},{"task_id":"task_248_enterprise_industry_analysis_medium_medium063","category":"enterprise_industry_analysis","score":1.0,"tokens":165653,"cost_usd":0.358638,"elapsed_time":277.1,"gpr":1.0,"tpe":0.481,"ee":0.2667},{"task_id":"task_249_enterprise_industry_analysis_medium_medium064","category":"enterprise_industry_analysis","score":0.0,"tokens":7018463,"cost_usd":9.699852,"elapsed_time":1207.46,"gpr":0.5,"tpe":0.0966,"ee":0.0571},{"task_id":"task_250_enterprise_industry_analysis_medium_medium065","category":"enterprise_industry_analysis","score":0.0,"tokens":151181,"cost_usd":0.314732,"elapsed_time":33.33,"gpr":0.0,"tpe":0.0,"ee":2.0},{"task_id":"task_251_enterprise_industry_analysis_medium_medium066","category":"enterprise_industry_analysis","score":1.0,"tokens":187117,"cost_usd":0.335304,"elapsed_time":387.63,"gpr":1.0,"tpe":0.2824,"ee":0.25},{"task_id":"task_252_enterprise_industry_analysis_medium_medium067","category":"enterprise_industry_analysis","score":1.0,"tokens":163983,"cost_usd":0.264382,"elapsed_time":321.77,"gpr":1.0,"tpe":0.5219,"ee":0.2857},{"task_id":"task_253_enterprise_industry_analysis_medium_medium068","category":"enterprise_industry_analysis","score":0.0,"tokens":320639,"cost_usd":0.590997,"elapsed_time":358.35,"gpr":0.75,"tpe":0.4284,"ee":0.1481},{"task_id":"task_254_enterprise_industry_analysis_medium_medium069","category":"enterprise_industry_analysis","score":1.0,"tokens":373882,"cost_usd":0.718163,"elapsed_time":565.27,"gpr":1.0,"tpe":0.0775,"ee":0.129},{"task_id":"task_255_enterprise_industry_analysis_medium_medium070","category":"enterprise_industry_analysis","score":0.0,"tokens":182675,"cost_usd":0.381981,"elapsed_time":247.06,"gpr":0.6667,"tpe":0.81,"ee":0.3333},{"task_id":"task_256_enterprise_industry_analysis_medium_medium071","category":"enterprise_industry_analysis","score":0.0,"tokens":85508,"cost_usd":0.180323,"elapsed_time":112.24,"gpr":0.5,"tpe":0.8645,"ee":0.5},{"task_id":"task_257_enterprise_industry_analysis_medium_medium072","category":"enterprise_industry_analysis","score":1.0,"tokens":166263,"cost_usd":0.296262,"elapsed_time":230.63,"gpr":1.0,"tpe":0.5536,"ee":0.2857},{"task_id":"task_258_enterprise_industry_analysis_medium_medium073","category":"enterprise_industry_analysis","score":0.0,"tokens":470811,"cost_usd":0.885403,"elapsed_time":813.04,"gpr":0.75,"tpe":0.3578,"ee":0.1379},{"task_id":"task_259_enterprise_industry_analysis_medium_medium074","category":"enterprise_industry_analysis","score":0.0,"tokens":1165585,"cost_usd":2.011122,"elapsed_time":1207.45,"gpr":0.25,"tpe":0.5905,"ee":0.1143},{"task_id":"task_260_enterprise_industry_analysis_medium_medium075","category":"enterprise_industry_analysis","score":0.0,"tokens":436026,"cost_usd":0.781091,"elapsed_time":1207.45,"gpr":1.0,"tpe":0.4029,"ee":0.129},{"task_id":"task_261_enterprise_industry_analysis_medium_medium076","category":"enterprise_industry_analysis","score":0.0,"tokens":10397234,"cost_usd":15.913518,"elapsed_time":1207.45,"gpr":0.75,"tpe":0.795,"ee":0.0833},{"task_id":"task_262_enterprise_industry_analysis_medium_medium077","category":"enterprise_industry_analysis","score":1.0,"tokens":176935,"cost_usd":0.355921,"elapsed_time":643.31,"gpr":1.0,"tpe":0.5423,"ee":0.3333},{"task_id":"task_263_enterprise_industry_analysis_medium_medium078","category":"enterprise_industry_analysis","score":1.0,"tokens":896626,"cost_usd":1.736246,"elapsed_time":708.32,"gpr":1.0,"tpe":0.2569,"ee":0.129},{"task_id":"task_264_enterprise_industry_analysis_medium_medium079","category":"enterprise_industry_analysis","score":0.0,"tokens":1077577,"cost_usd":1.942196,"elapsed_time":1207.44,"gpr":1.0,"tpe":0.407,"ee":0.0667},{"task_id":"task_265_enterprise_industry_analysis_medium_medium080","category":"enterprise_industry_analysis","score":0.0,"tokens":283137,"cost_usd":0.459839,"elapsed_time":289.29,"gpr":0.0,"tpe":0.0,"ee":0.2353},{"task_id":"task_266_enterprise_industry_analysis_medium_medium081","category":"enterprise_industry_analysis","score":0.0,"tokens":646704,"cost_usd":0.909364,"elapsed_time":717.65,"gpr":0.75,"tpe":0.3299,"ee":0.1026},{"task_id":"task_267_enterprise_industry_analysis_medium_medium082","category":"enterprise_industry_analysis","score":1.0,"tokens":1126784,"cost_usd":1.783551,"elapsed_time":231.79,"gpr":1.0,"tpe":0.2542,"ee":0.2353},{"task_id":"task_268_enterprise_industry_analysis_medium_medium083","category":"enterprise_industry_analysis","score":0.0,"tokens":389764,"cost_usd":0.617082,"elapsed_time":1207.39,"gpr":1.0,"tpe":0.2149,"ee":0.16},{"task_id":"task_269_enterprise_industry_analysis_medium_medium084","category":"enterprise_industry_analysis","score":1.0,"tokens":275024,"cost_usd":0.492367,"elapsed_time":268.83,"gpr":1.0,"tpe":0.3376,"ee":0.2105},{"task_id":"task_270_enterprise_industry_analysis_medium_medium085","category":"enterprise_industry_analysis","score":1.0,"tokens":375596,"cost_usd":0.710765,"elapsed_time":385.62,"gpr":1.0,"tpe":0.4166,"ee":0.1379},{"task_id":"task_271_enterprise_industry_analysis_medium_medium086","category":"enterprise_industry_analysis","score":1.0,"tokens":171378,"cost_usd":0.29467,"elapsed_time":231.15,"gpr":1.0,"tpe":0.6499,"ee":0.3077},{"task_id":"task_272_enterprise_industry_analysis_medium_medium087","category":"enterprise_industry_analysis","score":1.0,"tokens":1695422,"cost_usd":2.701941,"elapsed_time":653.89,"gpr":1.0,"tpe":0.5998,"ee":0.2},{"task_id":"task_273_enterprise_industry_analysis_medium_medium088","category":"enterprise_industry_analysis","score":1.0,"tokens":4100353,"cost_usd":7.51323,"elapsed_time":578.46,"gpr":1.0,"tpe":0.0861,"ee":0.1143},{"task_id":"task_274_enterprise_industry_analysis_medium_medium089","category":"enterprise_industry_analysis","score":0.0,"tokens":7489652,"cost_usd":10.892968,"elapsed_time":1207.5,"gpr":1.0,"tpe":0.4037,"ee":0.0784},{"task_id":"task_275_enterprise_industry_analysis_medium_medium090","category":"enterprise_industry_analysis","score":0.0,"tokens":8755452,"cost_usd":14.368265,"elapsed_time":817.52,"gpr":0.75,"tpe":0.3486,"ee":0.0727},{"task_id":"task_276_enterprise_industry_analysis_medium_medium091","category":"enterprise_industry_analysis","score":0.0,"tokens":18045818,"cost_usd":24.07719,"elapsed_time":1207.37,"gpr":1.0,"tpe":0.4554,"ee":0.0548},{"task_id":"task_277_enterprise_industry_analysis_medium_medium092","category":"enterprise_industry_analysis","score":1.0,"tokens":1792217,"cost_usd":3.248554,"elapsed_time":553.59,"gpr":1.0,"tpe":0.1177,"ee":0.1429},{"task_id":"task_278_enterprise_industry_analysis_medium_medium093","category":"enterprise_industry_analysis","score":1.0,"tokens":699364,"cost_usd":1.28249,"elapsed_time":728.65,"gpr":1.0,"tpe":0.3472,"ee":0.1111},{"task_id":"task_279_enterprise_industry_analysis_medium_medium094","category":"enterprise_industry_analysis","score":1.0,"tokens":159353,"cost_usd":0.392986,"elapsed_time":222.47,"gpr":1.0,"tpe":0.7017,"ee":0.2857},{"task_id":"task_280_enterprise_industry_analysis_medium_medium095","category":"enterprise_industry_analysis","score":0.0,"tokens":393716,"cost_usd":0.66295,"elapsed_time":965.46,"gpr":0.75,"tpe":0.1351,"ee":0.1667},{"task_id":"task_281_enterprise_industry_analysis_medium_medium096","category":"enterprise_industry_analysis","score":1.0,"tokens":147279,"cost_usd":0.265353,"elapsed_time":248.83,"gpr":1.0,"tpe":0.6585,"ee":0.3077},{"task_id":"task_282_enterprise_industry_analysis_medium_medium097","category":"enterprise_industry_analysis","score":1.0,"tokens":284349,"cost_usd":0.58557,"elapsed_time":635.16,"gpr":1.0,"tpe":0.1647,"ee":0.1667},{"task_id":"task_283_enterprise_industry_analysis_medium_medium098","category":"enterprise_industry_analysis","score":0.0,"tokens":483483,"cost_usd":0.95253,"elapsed_time":1207.37,"gpr":0.25,"tpe":0.0985,"ee":0.1053},{"task_id":"task_284_enterprise_industry_analysis_medium_medium099","category":"enterprise_industry_analysis","score":1.0,"tokens":694205,"cost_usd":1.319499,"elapsed_time":995.75,"gpr":1.0,"tpe":0.0464,"ee":0.0851},{"task_id":"task_285_enterprise_industry_analysis_medium_medium100","category":"enterprise_industry_analysis","score":0.0,"tokens":885421,"cost_usd":1.165137,"elapsed_time":1207.4,"gpr":0.5,"tpe":0.0985,"ee":0.1538},{"task_id":"task_286_enterprise_industry_analysis_medium_medium101","category":"enterprise_industry_analysis","score":1.0,"tokens":284744,"cost_usd":0.472504,"elapsed_time":828.9,"gpr":1.0,"tpe":0.925,"ee":0.1667},{"task_id":"task_287_enterprise_industry_analysis_medium_medium102","category":"enterprise_industry_analysis","score":0.0,"tokens":215079,"cost_usd":0.375412,"elapsed_time":622.14,"gpr":0.0,"tpe":0.0,"ee":0.2222},{"task_id":"task_288_enterprise_industry_analysis_medium_medium103","category":"enterprise_industry_analysis","score":0.0,"tokens":215783,"cost_usd":0.346296,"elapsed_time":346.15,"gpr":0.0,"tpe":0.0,"ee":0.2105},{"task_id":"task_289_enterprise_industry_analysis_medium_medium104","category":"enterprise_industry_analysis","score":0.0,"tokens":153130,"cost_usd":0.259458,"elapsed_time":361.82,"gpr":0.75,"tpe":0.5905,"ee":0.2857},{"task_id":"task_290_enterprise_industry_analysis_medium_medium105","category":"enterprise_industry_analysis","score":0.0,"tokens":178061,"cost_usd":0.293097,"elapsed_time":561.52,"gpr":0.75,"tpe":0.5787,"ee":0.2667},{"task_id":"task_291_enterprise_industry_analysis_medium_medium106","category":"enterprise_industry_analysis","score":0.0,"tokens":491087,"cost_usd":0.883173,"elapsed_time":1207.66,"gpr":0.25,"tpe":0.4305,"ee":0.1},{"task_id":"task_292_enterprise_industry_analysis_medium_medium107","category":"enterprise_industry_analysis","score":0.0,"tokens":438924,"cost_usd":0.702257,"elapsed_time":941.11,"gpr":0.25,"tpe":0.3874,"ee":0.2353},{"task_id":"task_293_enterprise_industry_analysis_medium_medium108","category":"enterprise_industry_analysis","score":0.0,"tokens":441826,"cost_usd":0.803805,"elapsed_time":1207.35,"gpr":0.25,"tpe":0.2824,"ee":0.1379},{"task_id":"task_294_enterprise_industry_analysis_medium_medium109","category":"enterprise_industry_analysis","score":1.0,"tokens":240499,"cost_usd":0.464522,"elapsed_time":202.63,"gpr":1.0,"tpe":0.3487,"ee":0.2857},{"task_id":"task_295_enterprise_industry_analysis_medium_medium110","category":"enterprise_industry_analysis","score":0.0,"tokens":4594666,"cost_usd":8.008358,"elapsed_time":1207.37,"gpr":0.25,"tpe":0.1501,"ee":0.0851},{"task_id":"task_296_enterprise_industry_analysis_medium_medium111","category":"enterprise_industry_analysis","score":0.0,"tokens":183756,"cost_usd":0.337163,"elapsed_time":557.29,"gpr":0.25,"tpe":1.0,"ee":0.25},{"task_id":"task_297_enterprise_industry_policy_analysis_easy_easy067","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":176869,"cost_usd":0.331275,"elapsed_time":758.94,"gpr":1.0,"tpe":0.3138,"ee":0.2143},{"task_id":"task_298_enterprise_industry_policy_analysis_easy_easy068","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":540023,"cost_usd":0.882369,"elapsed_time":1207.53,"gpr":0.0,"tpe":0.0,"ee":0.0968},{"task_id":"task_299_enterprise_industry_policy_analysis_easy_easy069","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1281815,"cost_usd":2.373063,"elapsed_time":1041.01,"gpr":1.0,"tpe":0.0133,"ee":0.0682},{"task_id":"task_300_enterprise_industry_policy_analysis_easy_easy070","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":343949,"cost_usd":0.650226,"elapsed_time":759.79,"gpr":1.0,"tpe":0.1351,"ee":0.1364},{"task_id":"task_301_enterprise_industry_policy_analysis_easy_easy071","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":90796,"cost_usd":0.142232,"elapsed_time":398.84,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_302_enterprise_industry_policy_analysis_easy_easy072","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":53268,"cost_usd":0.117996,"elapsed_time":87.52,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_303_enterprise_industry_policy_analysis_easy_easy073","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":58165,"cost_usd":0.14341,"elapsed_time":337.74,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_304_enterprise_industry_policy_analysis_easy_easy074","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":44321,"cost_usd":0.096632,"elapsed_time":62.77,"gpr":0.0,"tpe":0.0,"ee":1.5},{"task_id":"task_305_enterprise_industry_policy_analysis_easy_easy075","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":223174,"cost_usd":0.450607,"elapsed_time":200.12,"gpr":1.0,"tpe":0.2824,"ee":0.2},{"task_id":"task_306_enterprise_industry_policy_analysis_easy_easy076","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":118145,"cost_usd":0.33887,"elapsed_time":763.48,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_307_enterprise_industry_policy_analysis_medium_medium001","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":260467,"cost_usd":0.485334,"elapsed_time":436.62,"gpr":1.0,"tpe":0.3254,"ee":0.2273},{"task_id":"task_308_enterprise_industry_policy_analysis_medium_medium002","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":3090598,"cost_usd":6.017779,"elapsed_time":738.73,"gpr":1.0,"tpe":0.495,"ee":0.122},{"task_id":"task_309_enterprise_industry_policy_analysis_medium_medium003","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":212847,"cost_usd":0.336495,"elapsed_time":1207.38,"gpr":0.8,"tpe":0.4607,"ee":0.2632},{"task_id":"task_310_enterprise_industry_policy_analysis_medium_medium004","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":84619,"cost_usd":0.161611,"elapsed_time":114.9,"gpr":0.4,"tpe":1.0,"ee":0.7143},{"task_id":"task_311_enterprise_industry_policy_analysis_medium_medium005","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":287292,"cost_usd":0.515008,"elapsed_time":417.75,"gpr":1.0,"tpe":0.449,"ee":0.2778},{"task_id":"task_312_enterprise_industry_policy_analysis_medium_medium006","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":293171,"cost_usd":0.556085,"elapsed_time":577.57,"gpr":1.0,"tpe":0.5552,"ee":0.2273},{"task_id":"task_313_enterprise_industry_policy_analysis_medium_medium007","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":176080,"cost_usd":0.381593,"elapsed_time":257.38,"gpr":1.0,"tpe":0.6501,"ee":0.4545},{"task_id":"task_314_enterprise_industry_policy_analysis_medium_medium008","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":138677,"cost_usd":0.307574,"elapsed_time":328.84,"gpr":1.0,"tpe":0.8097,"ee":0.5},{"task_id":"task_315_enterprise_industry_policy_analysis_medium_medium009","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":320894,"cost_usd":0.644068,"elapsed_time":459.59,"gpr":1.0,"tpe":0.4319,"ee":0.2941},{"task_id":"task_316_enterprise_industry_policy_analysis_medium_medium010","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2375990,"cost_usd":4.53056,"elapsed_time":1207.4,"gpr":0.8,"tpe":0.3729,"ee":0.1562},{"task_id":"task_317_enterprise_industry_policy_analysis_medium_medium011","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":725526,"cost_usd":1.310204,"elapsed_time":509.34,"gpr":1.0,"tpe":0.6665,"ee":0.119},{"task_id":"task_318_enterprise_industry_policy_analysis_medium_medium012","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":116988,"cost_usd":0.223283,"elapsed_time":222.59,"gpr":0.4,"tpe":0.9,"ee":0.5556},{"task_id":"task_319_enterprise_industry_policy_analysis_medium_medium013","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":278454,"cost_usd":0.543386,"elapsed_time":638.93,"gpr":1.0,"tpe":0.4088,"ee":0.2083},{"task_id":"task_320_enterprise_industry_policy_analysis_medium_medium014","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":145448,"cost_usd":0.308002,"elapsed_time":205.62,"gpr":1.0,"tpe":0.8035,"ee":0.4167},{"task_id":"task_321_enterprise_industry_policy_analysis_medium_medium015","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":202586,"cost_usd":0.417718,"elapsed_time":193.96,"gpr":1.0,"tpe":0.5821,"ee":0.2941},{"task_id":"task_322_enterprise_industry_policy_analysis_medium_medium016","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":260420,"cost_usd":0.599132,"elapsed_time":286.68,"gpr":1.0,"tpe":0.5286,"ee":0.2381},{"task_id":"task_323_enterprise_industry_policy_analysis_medium_medium017","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":195560,"cost_usd":0.349781,"elapsed_time":667.48,"gpr":1.0,"tpe":0.7828,"ee":0.3125},{"task_id":"task_324_enterprise_industry_policy_analysis_medium_medium018","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":183035,"cost_usd":0.335338,"elapsed_time":523.86,"gpr":1.0,"tpe":0.8175,"ee":0.3571},{"task_id":"task_325_enterprise_industry_policy_analysis_medium_medium019","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":397105,"cost_usd":0.705117,"elapsed_time":368.07,"gpr":1.0,"tpe":0.345,"ee":0.2273},{"task_id":"task_326_enterprise_industry_policy_analysis_medium_medium020","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":4621728,"cost_usd":6.406125,"elapsed_time":1207.39,"gpr":0.4,"tpe":0.4725,"ee":0.1389},{"task_id":"task_327_enterprise_industry_policy_analysis_medium_medium021","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":390183,"cost_usd":0.709246,"elapsed_time":303.79,"gpr":0.4,"tpe":0.8645,"ee":0.2083},{"task_id":"task_328_enterprise_industry_policy_analysis_medium_medium022","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1970938,"cost_usd":3.408038,"elapsed_time":1207.39,"gpr":0.4,"tpe":0.6925,"ee":0.1282},{"task_id":"task_329_enterprise_industry_policy_analysis_medium_medium023","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":472313,"cost_usd":0.760525,"elapsed_time":374.9,"gpr":1.0,"tpe":0.1304,"ee":0.1852},{"task_id":"task_330_enterprise_industry_policy_analysis_medium_medium024","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2218734,"cost_usd":3.226977,"elapsed_time":1207.4,"gpr":0.8,"tpe":0.5892,"ee":0.0658},{"task_id":"task_331_enterprise_industry_policy_analysis_medium_medium025","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":918307,"cost_usd":1.396803,"elapsed_time":1076.09,"gpr":1.0,"tpe":0.7014,"ee":0.1087},{"task_id":"task_332_enterprise_industry_policy_analysis_medium_medium026","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":356784,"cost_usd":0.528918,"elapsed_time":297.26,"gpr":1.0,"tpe":0.199,"ee":0.1667},{"task_id":"task_333_enterprise_industry_policy_analysis_medium_medium027","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":129260,"cost_usd":0.220876,"elapsed_time":178.34,"gpr":1.0,"tpe":0.8303,"ee":0.4545},{"task_id":"task_334_enterprise_industry_policy_analysis_medium_medium028","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":4699039,"cost_usd":7.605478,"elapsed_time":1189.84,"gpr":1.0,"tpe":0.4051,"ee":0.1087},{"task_id":"task_335_enterprise_industry_policy_analysis_medium_medium029","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":189237,"cost_usd":0.321058,"elapsed_time":228.38,"gpr":1.0,"tpe":0.8683,"ee":0.4545},{"task_id":"task_336_enterprise_industry_policy_analysis_medium_medium030","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":323991,"cost_usd":0.572144,"elapsed_time":314.29,"gpr":0.8,"tpe":0.6535,"ee":0.2273},{"task_id":"task_337_enterprise_industry_policy_analysis_medium_medium031","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1486466,"cost_usd":2.496357,"elapsed_time":1207.48,"gpr":1.0,"tpe":0.2281,"ee":0.0714},{"task_id":"task_338_enterprise_industry_policy_analysis_medium_medium032","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2907687,"cost_usd":4.973214,"elapsed_time":907.66,"gpr":0.75,"tpe":0.5728,"ee":0.1111},{"task_id":"task_339_enterprise_industry_policy_analysis_medium_medium033","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":342015,"cost_usd":0.612489,"elapsed_time":311.61,"gpr":1.0,"tpe":0.2073,"ee":0.1538},{"task_id":"task_340_enterprise_industry_policy_analysis_medium_medium034","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":422874,"cost_usd":0.827788,"elapsed_time":408.1,"gpr":0.5,"tpe":0.5771,"ee":0.1212},{"task_id":"task_341_enterprise_industry_policy_analysis_medium_medium035","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1304282,"cost_usd":2.367921,"elapsed_time":824.26,"gpr":1.0,"tpe":0.2744,"ee":0.0635},{"task_id":"task_342_enterprise_industry_policy_analysis_medium_medium036","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":3605613,"cost_usd":5.453466,"elapsed_time":1207.78,"gpr":0.75,"tpe":1.0,"ee":0.1481},{"task_id":"task_343_enterprise_industry_policy_analysis_medium_medium037","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1903850,"cost_usd":2.84634,"elapsed_time":1207.41,"gpr":0.75,"tpe":0.3185,"ee":0.0404},{"task_id":"task_344_enterprise_industry_policy_analysis_medium_medium038","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":428459,"cost_usd":0.837996,"elapsed_time":611.4,"gpr":1.0,"tpe":0.2036,"ee":0.16},{"task_id":"task_345_enterprise_industry_policy_analysis_medium_medium039","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2838779,"cost_usd":4.490997,"elapsed_time":587.32,"gpr":0.75,"tpe":0.4004,"ee":0.2353},{"task_id":"task_346_enterprise_industry_policy_analysis_medium_medium040","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1154935,"cost_usd":2.140316,"elapsed_time":635.12,"gpr":1.0,"tpe":0.1913,"ee":0.0816},{"task_id":"task_347_enterprise_industry_policy_analysis_medium_medium041","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":487822,"cost_usd":0.811991,"elapsed_time":406.42,"gpr":0.75,"tpe":0.4943,"ee":0.1212},{"task_id":"task_348_enterprise_industry_policy_analysis_medium_medium042","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2942930,"cost_usd":4.15503,"elapsed_time":1207.45,"gpr":1.0,"tpe":0.228,"ee":0.0615},{"task_id":"task_349_enterprise_industry_policy_analysis_medium_medium043","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1807436,"cost_usd":2.516242,"elapsed_time":1207.43,"gpr":0.75,"tpe":0.3305,"ee":0.0488},{"task_id":"task_350_enterprise_industry_policy_analysis_medium_medium044","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2712950,"cost_usd":3.354953,"elapsed_time":1207.37,"gpr":0.75,"tpe":0.3139,"ee":0.0364},{"task_id":"task_351_enterprise_industry_policy_analysis_medium_medium045","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":447639,"cost_usd":0.589298,"elapsed_time":801.37,"gpr":0.75,"tpe":0.5333,"ee":0.1143},{"task_id":"task_352_enterprise_industry_policy_analysis_medium_medium046","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1129826,"cost_usd":1.526992,"elapsed_time":463.84,"gpr":0.0,"tpe":0.0,"ee":0.1176},{"task_id":"task_353_enterprise_industry_policy_analysis_medium_medium047","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":339156,"cost_usd":0.56422,"elapsed_time":258.13,"gpr":1.0,"tpe":0.4216,"ee":0.1667},{"task_id":"task_354_enterprise_industry_policy_analysis_medium_medium048","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":190986,"cost_usd":0.320444,"elapsed_time":199.32,"gpr":1.0,"tpe":0.6089,"ee":0.2667},{"task_id":"task_355_enterprise_industry_policy_analysis_medium_medium049","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":12616725,"cost_usd":18.80849,"elapsed_time":1207.4,"gpr":0.75,"tpe":0.6083,"ee":0.0702},{"task_id":"task_356_enterprise_industry_policy_analysis_medium_medium050","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":8826251,"cost_usd":10.533949,"elapsed_time":1207.5,"gpr":0.5,"tpe":0.368,"ee":0.0597},{"task_id":"task_357_enterprise_industry_policy_analysis_medium_medium051","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":981833,"cost_usd":1.310344,"elapsed_time":815.81,"gpr":0.5,"tpe":0.6437,"ee":0.0727},{"task_id":"task_358_enterprise_industry_policy_analysis_medium_medium052","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":4092770,"cost_usd":5.861986,"elapsed_time":1207.59,"gpr":0.75,"tpe":0.4237,"ee":0.0482},{"task_id":"task_359_enterprise_industry_policy_analysis_medium_medium053","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":270763,"cost_usd":0.467283,"elapsed_time":259.74,"gpr":1.0,"tpe":0.5742,"ee":0.1818},{"task_id":"task_360_enterprise_industry_policy_analysis_medium_medium054","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":4786144,"cost_usd":6.587044,"elapsed_time":1007.43,"gpr":0.5,"tpe":0.3828,"ee":0.1176},{"task_id":"task_361_enterprise_industry_policy_analysis_medium_medium055","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1172828,"cost_usd":1.959597,"elapsed_time":1188.75,"gpr":1.0,"tpe":0.2033,"ee":0.0526},{"task_id":"task_362_enterprise_industry_policy_analysis_medium_medium056","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2178019,"cost_usd":3.180818,"elapsed_time":1207.48,"gpr":1.0,"tpe":0.3745,"ee":0.0533},{"task_id":"task_363_enterprise_industry_policy_analysis_medium_medium057","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1357255,"cost_usd":2.198097,"elapsed_time":629.43,"gpr":0.25,"tpe":0.9,"ee":0.1},{"task_id":"task_364_enterprise_industry_policy_analysis_medium_medium058","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":448966,"cost_usd":0.657861,"elapsed_time":505.03,"gpr":0.5,"tpe":1.0,"ee":0.125},{"task_id":"task_365_enterprise_industry_policy_analysis_medium_medium059","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1052650,"cost_usd":1.442881,"elapsed_time":1207.41,"gpr":0.75,"tpe":0.5782,"ee":0.0727},{"task_id":"task_366_enterprise_industry_policy_analysis_medium_medium060","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":763210,"cost_usd":1.295801,"elapsed_time":654.19,"gpr":0.25,"tpe":0.9,"ee":0.0909},{"task_id":"task_367_enterprise_industry_policy_analysis_medium_medium061","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":7961305,"cost_usd":9.718317,"elapsed_time":1207.46,"gpr":0.25,"tpe":0.5314,"ee":0.069},{"task_id":"task_368_enterprise_industry_policy_analysis_medium_medium062","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2991590,"cost_usd":3.473107,"elapsed_time":1207.48,"gpr":1.0,"tpe":0.5201,"ee":0.0548},{"task_id":"task_369_enterprise_industry_policy_analysis_medium_medium063","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":7529865,"cost_usd":8.657543,"elapsed_time":1207.46,"gpr":0.25,"tpe":1.0,"ee":0.0727},{"task_id":"task_370_enterprise_industry_policy_analysis_medium_medium064","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1520437,"cost_usd":2.498808,"elapsed_time":654.44,"gpr":0.75,"tpe":0.3422,"ee":0.1081},{"task_id":"task_371_enterprise_industry_policy_analysis_medium_medium065","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":5075571,"cost_usd":6.882365,"elapsed_time":1207.42,"gpr":1.0,"tpe":0.5886,"ee":0.093},{"task_id":"task_372_enterprise_industry_policy_analysis_medium_medium066","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2973399,"cost_usd":3.839956,"elapsed_time":1207.42,"gpr":0.75,"tpe":0.7608,"ee":0.069},{"task_id":"task_373_hypothesis_verification_hard_hard001","category":"hypothesis_verification","score":0.0,"tokens":926756,"cost_usd":1.393428,"elapsed_time":894.81,"gpr":0.0,"tpe":0.0,"ee":0.1795},{"task_id":"task_374_hypothesis_verification_hard_hard002","category":"hypothesis_verification","score":0.0,"tokens":2169644,"cost_usd":3.244955,"elapsed_time":1207.44,"gpr":0.375,"tpe":0.1778,"ee":0.0725},{"task_id":"task_375_hypothesis_verification_hard_hard003","category":"hypothesis_verification","score":0.0,"tokens":1566507,"cost_usd":2.298383,"elapsed_time":1207.43,"gpr":0.1111,"tpe":0.0278,"ee":0.1404},{"task_id":"task_376_hypothesis_verification_hard_hard004","category":"hypothesis_verification","score":0.0,"tokens":1751805,"cost_usd":2.662791,"elapsed_time":1207.37,"gpr":0.1,"tpe":0.2288,"ee":0.1111},{"task_id":"task_377_hypothesis_verification_hard_hard005","category":"hypothesis_verification","score":0.0,"tokens":1782059,"cost_usd":3.045821,"elapsed_time":1207.4,"gpr":0.0,"tpe":0.0,"ee":0.1667},{"task_id":"task_378_hypothesis_verification_hard_hard006","category":"hypothesis_verification","score":0.0,"tokens":710843,"cost_usd":1.083378,"elapsed_time":1207.37,"gpr":0.0,"tpe":0.0,"ee":0.2188},{"task_id":"task_379_hypothesis_verification_hard_hard007","category":"hypothesis_verification","score":0.0,"tokens":2937871,"cost_usd":4.92741,"elapsed_time":1207.33,"gpr":0.3125,"tpe":0.4591,"ee":0.1579},{"task_id":"task_380_hypothesis_verification_hard_hard008","category":"hypothesis_verification","score":0.0,"tokens":1443592,"cost_usd":2.406,"elapsed_time":1207.38,"gpr":0.1111,"tpe":0.1216,"ee":0.1458},{"task_id":"task_381_hypothesis_verification_hard_hard009","category":"hypothesis_verification","score":0.0,"tokens":2749021,"cost_usd":4.108328,"elapsed_time":1207.44,"gpr":0.0,"tpe":0.0,"ee":0.0845},{"task_id":"task_382_hypothesis_verification_hard_hard010","category":"hypothesis_verification","score":0.0,"tokens":3057097,"cost_usd":4.46827,"elapsed_time":1207.41,"gpr":0.0667,"tpe":0.0064,"ee":0.0986},{"task_id":"task_383_hypothesis_verification_hard_hard011","category":"hypothesis_verification","score":0.0,"tokens":2750531,"cost_usd":3.894051,"elapsed_time":1207.5,"gpr":0.0,"tpe":0.0,"ee":0.0959},{"task_id":"task_384_hypothesis_verification_hard_hard012","category":"hypothesis_verification","score":0.0,"tokens":1729450,"cost_usd":2.844306,"elapsed_time":1207.43,"gpr":0.0,"tpe":0.0,"ee":0.0952},{"task_id":"task_385_hypothesis_verification_hard_hard013","category":"hypothesis_verification","score":0.0,"tokens":3116954,"cost_usd":4.3178,"elapsed_time":1207.4,"gpr":0.7333,"tpe":0.1141,"ee":0.1148},{"task_id":"task_386_hypothesis_verification_hard_hard014","category":"hypothesis_verification","score":0.0,"tokens":2066790,"cost_usd":3.173991,"elapsed_time":1207.6,"gpr":0.125,"tpe":1.0,"ee":0.0822},{"task_id":"task_387_hypothesis_verification_hard_hard015","category":"hypothesis_verification","score":0.0,"tokens":1367300,"cost_usd":2.205034,"elapsed_time":774.92,"gpr":0.0909,"tpe":0.1501,"ee":0.1458},{"task_id":"task_388_hypothesis_verification_medium_medium001","category":"hypothesis_verification","score":0.0,"tokens":1342404,"cost_usd":2.297611,"elapsed_time":1207.59,"gpr":0.0,"tpe":0.0,"ee":0.14},{"task_id":"task_389_hypothesis_verification_medium_medium002","category":"hypothesis_verification","score":0.0,"tokens":2228004,"cost_usd":4.077688,"elapsed_time":1207.52,"gpr":0.0,"tpe":0.0,"ee":0.1228},{"task_id":"task_390_hypothesis_verification_medium_medium003","category":"hypothesis_verification","score":0.0,"tokens":2087501,"cost_usd":2.991567,"elapsed_time":1076.55,"gpr":0.0,"tpe":0.0,"ee":0.12},{"task_id":"task_391_hypothesis_verification_medium_medium004","category":"hypothesis_verification","score":0.0,"tokens":3132770,"cost_usd":4.66158,"elapsed_time":1207.41,"gpr":0.2,"tpe":0.3138,"ee":0.0986},{"task_id":"task_392_hypothesis_verification_medium_medium005","category":"hypothesis_verification","score":0.0,"tokens":1589591,"cost_usd":2.110038,"elapsed_time":1207.49,"gpr":0.6,"tpe":0.0459,"ee":0.1014},{"task_id":"task_393_hypothesis_verification_medium_medium006","category":"hypothesis_verification","score":0.0,"tokens":2117532,"cost_usd":3.272039,"elapsed_time":1207.36,"gpr":0.2,"tpe":0.0471,"ee":0.0952},{"task_id":"task_394_hypothesis_verification_medium_medium007","category":"hypothesis_verification","score":0.0,"tokens":1840847,"cost_usd":2.699145,"elapsed_time":864.8,"gpr":0.1111,"tpe":0.3138,"ee":0.1458},{"task_id":"task_395_hypothesis_verification_medium_medium008","category":"hypothesis_verification","score":0.0,"tokens":1097211,"cost_usd":1.812115,"elapsed_time":1022.77,"gpr":0.2,"tpe":0.2824,"ee":0.2},{"task_id":"task_396_hypothesis_verification_medium_medium009","category":"hypothesis_verification","score":0.0,"tokens":2012871,"cost_usd":3.022109,"elapsed_time":1207.39,"gpr":0.0,"tpe":0.0,"ee":0.087},{"task_id":"task_397_hypothesis_verification_medium_medium010","category":"hypothesis_verification","score":1.0,"tokens":804216,"cost_usd":1.512599,"elapsed_time":633.58,"gpr":1.0,"tpe":0.5069,"ee":0.25},{"task_id":"task_398_hypothesis_verification_medium_medium011","category":"hypothesis_verification","score":0.0,"tokens":1717517,"cost_usd":2.586787,"elapsed_time":902.96,"gpr":0.3333,"tpe":0.1677,"ee":0.1364},{"task_id":"task_399_hypothesis_verification_medium_medium012","category":"hypothesis_verification","score":1.0,"tokens":764012,"cost_usd":1.390953,"elapsed_time":347.33,"gpr":1.0,"tpe":0.2109,"ee":0.2609},{"task_id":"task_400_hypothesis_verification_medium_medium013","category":"hypothesis_verification","score":0.0,"tokens":1052531,"cost_usd":1.917509,"elapsed_time":943.98,"gpr":0.1667,"tpe":0.1094,"ee":0.1842},{"task_id":"task_401_hypothesis_verification_medium_medium014","category":"hypothesis_verification","score":1.0,"tokens":829813,"cost_usd":1.19463,"elapsed_time":533.93,"gpr":1.0,"tpe":0.5352,"ee":0.3333},{"task_id":"task_402_industry_planning_hard_hard001","category":"industry_planning","score":0.5,"tokens":4774682,"cost_usd":6.809522,"elapsed_time":1202.62,"gpr":1.0,"tpe":0.6835,"ee":0.1029},{"task_id":"task_403_industry_planning_hard_hard002","category":"industry_planning","score":0.0,"tokens":3762160,"cost_usd":5.010203,"elapsed_time":1207.35,"gpr":0.5,"tpe":0.2292,"ee":0.0843},{"task_id":"task_404_industry_planning_hard_hard003","category":"industry_planning","score":0.0,"tokens":2805887,"cost_usd":4.803895,"elapsed_time":1207.4,"gpr":0.2727,"tpe":0.4326,"ee":0.129},{"task_id":"task_405_industry_planning_hard_hard004","category":"industry_planning","score":0.0,"tokens":3375319,"cost_usd":4.958382,"elapsed_time":1202.8,"gpr":0.0833,"tpe":0.3487,"ee":0.0857},{"task_id":"task_406_industry_planning_hard_hard005","category":"industry_planning","score":0.0,"tokens":6033761,"cost_usd":9.313157,"elapsed_time":1207.43,"gpr":0.5556,"tpe":0.4892,"ee":0.1395},{"task_id":"task_407_industry_planning_hard_hard006","category":"industry_planning","score":0.0,"tokens":2431944,"cost_usd":3.641094,"elapsed_time":1207.44,"gpr":0.9167,"tpe":0.2536,"ee":0.0833},{"task_id":"task_408_industry_planning_hard_hard007","category":"industry_planning","score":0.0,"tokens":1219495,"cost_usd":2.029929,"elapsed_time":656.81,"gpr":0.1,"tpe":1.0,"ee":0.1463},{"task_id":"task_409_industry_planning_hard_hard008","category":"industry_planning","score":0.0,"tokens":2926768,"cost_usd":4.412703,"elapsed_time":1207.43,"gpr":0.375,"tpe":0.7829,"ee":0.0759},{"task_id":"task_410_industry_planning_hard_hard009","category":"industry_planning","score":0.0,"tokens":2065192,"cost_usd":3.444868,"elapsed_time":842.02,"gpr":0.1,"tpe":0.0985,"ee":0.1034},{"task_id":"task_411_industry_planning_hard_hard010","category":"industry_planning","score":0.0,"tokens":3185070,"cost_usd":3.940991,"elapsed_time":959.0,"gpr":0.7778,"tpe":0.0931,"ee":0.0822},{"task_id":"task_412_industry_planning_hard_hard011","category":"industry_planning","score":1.0,"tokens":5238283,"cost_usd":7.610215,"elapsed_time":1103.21,"gpr":1.0,"tpe":0.0283,"ee":0.08},{"task_id":"task_413_industry_planning_hard_hard012","category":"industry_planning","score":0.0,"tokens":3377970,"cost_usd":4.788805,"elapsed_time":1207.69,"gpr":0.0,"tpe":0.0,"ee":0.0645},{"task_id":"task_414_industry_planning_hard_hard013","category":"industry_planning","score":0.0,"tokens":2373583,"cost_usd":3.348104,"elapsed_time":965.06,"gpr":0.0,"tpe":0.0,"ee":0.1129},{"task_id":"task_415_industry_planning_hard_hard014","category":"industry_planning","score":0.0,"tokens":6631296,"cost_usd":8.851348,"elapsed_time":1207.45,"gpr":0.0,"tpe":0.0,"ee":0.049},{"task_id":"task_416_industry_planning_medium_medium001","category":"industry_planning","score":0.0,"tokens":7078563,"cost_usd":9.93338,"elapsed_time":1207.41,"gpr":0.5,"tpe":0.0098,"ee":0.0941},{"task_id":"task_417_industry_planning_medium_medium002","category":"industry_planning","score":1.0,"tokens":1309715,"cost_usd":2.525154,"elapsed_time":812.92,"gpr":1.0,"tpe":0.1402,"ee":0.2105},{"task_id":"task_418_industry_planning_medium_medium003","category":"industry_planning","score":0.0,"tokens":376664,"cost_usd":0.647295,"elapsed_time":330.38,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_419_industry_planning_medium_medium004","category":"industry_planning","score":1.0,"tokens":3044484,"cost_usd":3.949929,"elapsed_time":526.52,"gpr":1.0,"tpe":0.2484,"ee":0.1628},{"task_id":"task_420_industry_planning_medium_medium005","category":"industry_planning","score":0.0,"tokens":5290889,"cost_usd":7.588481,"elapsed_time":1207.38,"gpr":0.1111,"tpe":0.0001,"ee":0.0583},{"task_id":"task_421_industry_planning_medium_medium006","category":"industry_planning","score":0.0,"tokens":3153155,"cost_usd":4.907157,"elapsed_time":1207.78,"gpr":0.1111,"tpe":0.0004,"ee":0.0909},{"task_id":"task_422_industry_planning_medium_medium007","category":"industry_planning","score":0.0,"tokens":1918462,"cost_usd":3.315278,"elapsed_time":748.14,"gpr":0.25,"tpe":0.0582,"ee":0.1356},{"task_id":"task_423_industry_planning_medium_medium008","category":"industry_planning","score":1.0,"tokens":1514679,"cost_usd":2.653248,"elapsed_time":725.04,"gpr":1.0,"tpe":0.0097,"ee":0.1538},{"task_id":"task_424_industry_planning_medium_medium009","category":"industry_planning","score":0.0,"tokens":1297689,"cost_usd":2.385663,"elapsed_time":707.57,"gpr":0.2,"tpe":0.1501,"ee":0.1702},{"task_id":"task_425_industry_planning_medium_medium010","category":"industry_planning","score":1.0,"tokens":1568763,"cost_usd":2.627658,"elapsed_time":634.37,"gpr":0.9,"tpe":0.1023,"ee":0.1455},{"task_id":"task_426_industry_planning_medium_medium011","category":"industry_planning","score":1.0,"tokens":749012,"cost_usd":1.346513,"elapsed_time":485.35,"gpr":1.0,"tpe":0.1812,"ee":0.2581},{"task_id":"task_427_industry_planning_medium_medium012","category":"industry_planning","score":0.0,"tokens":2329665,"cost_usd":3.881037,"elapsed_time":953.85,"gpr":0.2222,"tpe":0.0048,"ee":0.1013},{"task_id":"task_428_industry_planning_medium_medium013","category":"industry_planning","score":1.0,"tokens":2371180,"cost_usd":4.365057,"elapsed_time":851.24,"gpr":0.1667,"tpe":0.1094,"ee":0.1143},{"task_id":"task_429_industry_planning_medium_medium014","category":"industry_planning","score":0.0,"tokens":3042616,"cost_usd":4.952755,"elapsed_time":892.92,"gpr":0.0,"tpe":0.0,"ee":0.0968},{"task_id":"task_430_international_comparison_hard_hard001","category":"international_comparison","score":0.0,"tokens":4190527,"cost_usd":3.241159,"elapsed_time":1207.98,"gpr":0.0,"tpe":0.0,"ee":0.0652},{"task_id":"task_431_international_comparison_hard_hard002","category":"international_comparison","score":0.0,"tokens":8744318,"cost_usd":6.951783,"elapsed_time":1208.16,"gpr":0.0,"tpe":0.0,"ee":0.0857},{"task_id":"task_432_international_comparison_hard_hard003","category":"international_comparison","score":0.0,"tokens":3890114,"cost_usd":2.440527,"elapsed_time":1208.13,"gpr":0.1111,"tpe":0.0079,"ee":0.0779},{"task_id":"task_433_international_comparison_hard_hard004","category":"international_comparison","score":0.0,"tokens":8230827,"cost_usd":6.678931,"elapsed_time":1208.01,"gpr":0.0,"tpe":0.0,"ee":0.0833},{"task_id":"task_434_international_comparison_hard_hard005","category":"international_comparison","score":0.0,"tokens":4268110,"cost_usd":4.666652,"elapsed_time":1176.06,"gpr":0.1818,"tpe":0.012,"ee":0.12},{"task_id":"task_435_international_comparison_hard_hard006","category":"international_comparison","score":0.0,"tokens":1449028,"cost_usd":1.516728,"elapsed_time":896.54,"gpr":0.1667,"tpe":0.0148,"ee":0.1321},{"task_id":"task_436_international_comparison_hard_hard007","category":"international_comparison","score":0.0,"tokens":595337,"cost_usd":0.54553,"elapsed_time":1208.11,"gpr":0.0909,"tpe":0.5314,"ee":0.2333},{"task_id":"task_437_international_comparison_hard_hard008","category":"international_comparison","score":0.5,"tokens":4131713,"cost_usd":3.6912,"elapsed_time":646.71,"gpr":0.1667,"tpe":0.0582,"ee":0.1333},{"task_id":"task_438_international_comparison_hard_hard009","category":"international_comparison","score":0.0,"tokens":4054084,"cost_usd":3.332594,"elapsed_time":1208.15,"gpr":0.3333,"tpe":0.0225,"ee":0.1273},{"task_id":"task_439_international_comparison_hard_hard010","category":"international_comparison","score":0.0,"tokens":4145895,"cost_usd":3.277742,"elapsed_time":1208.2,"gpr":1.0,"tpe":0.0437,"ee":0.0619},{"task_id":"task_440_international_comparison_hard_hard011","category":"international_comparison","score":0.0,"tokens":2181273,"cost_usd":1.586471,"elapsed_time":1207.42,"gpr":1.0,"tpe":0.0019,"ee":0.0909},{"task_id":"task_441_international_comparison_hard_hard012","category":"international_comparison","score":0.0,"tokens":1428735,"cost_usd":1.200876,"elapsed_time":1207.68,"gpr":0.1429,"tpe":0.0038,"ee":0.1},{"task_id":"task_442_international_comparison_hard_hard013","category":"international_comparison","score":0.0,"tokens":1126050,"cost_usd":1.08659,"elapsed_time":1207.39,"gpr":0.0,"tpe":0.0,"ee":0.1364},{"task_id":"task_443_international_comparison_hard_hard014","category":"international_comparison","score":0.0,"tokens":1187809,"cost_usd":1.348431,"elapsed_time":1207.37,"gpr":0.0,"tpe":0.0,"ee":0.2609},{"task_id":"task_444_international_comparison_medium_medium001","category":"international_comparison","score":0.0,"tokens":3164910,"cost_usd":2.398728,"elapsed_time":1207.54,"gpr":0.2,"tpe":0.0087,"ee":0.08},{"task_id":"task_445_international_comparison_medium_medium002","category":"international_comparison","score":0.0,"tokens":1908742,"cost_usd":1.833353,"elapsed_time":1207.43,"gpr":0.3333,"tpe":0.1636,"ee":0.0667},{"task_id":"task_446_international_comparison_medium_medium003","category":"international_comparison","score":0.0,"tokens":2286601,"cost_usd":2.20077,"elapsed_time":1207.4,"gpr":0.0,"tpe":0.0,"ee":0.0952},{"task_id":"task_447_international_comparison_medium_medium004","category":"international_comparison","score":0.0,"tokens":6804013,"cost_usd":5.421439,"elapsed_time":1207.44,"gpr":0.1667,"tpe":0.0886,"ee":0.0519},{"task_id":"task_448_international_comparison_medium_medium005","category":"international_comparison","score":0.0,"tokens":8473674,"cost_usd":8.321376,"elapsed_time":1207.39,"gpr":0.0,"tpe":0.0,"ee":0.1053},{"task_id":"task_449_international_comparison_medium_medium006","category":"international_comparison","score":0.0,"tokens":1607388,"cost_usd":1.452647,"elapsed_time":1207.29,"gpr":0.1667,"tpe":0.0309,"ee":0.0909},{"task_id":"task_450_international_comparison_medium_medium007","category":"international_comparison","score":0.0,"tokens":666645,"cost_usd":0.610046,"elapsed_time":1207.38,"gpr":0.3333,"tpe":0.4783,"ee":0.1053},{"task_id":"task_451_international_comparison_medium_medium008","category":"international_comparison","score":0.0,"tokens":2767549,"cost_usd":1.900984,"elapsed_time":1207.39,"gpr":0.0,"tpe":0.0,"ee":0.0741},{"task_id":"task_452_international_comparison_medium_medium009","category":"international_comparison","score":0.0,"tokens":2468519,"cost_usd":1.35809,"elapsed_time":1207.45,"gpr":0.0,"tpe":0.0,"ee":0.0862},{"task_id":"task_453_international_comparison_medium_medium010","category":"international_comparison","score":0.0,"tokens":2231177,"cost_usd":2.148134,"elapsed_time":1207.45,"gpr":0.0,"tpe":0.0,"ee":0.0781},{"task_id":"task_454_international_comparison_medium_medium011","category":"international_comparison","score":0.0,"tokens":4787791,"cost_usd":2.645568,"elapsed_time":1207.33,"gpr":0.3333,"tpe":0.0368,"ee":0.0471},{"task_id":"task_455_international_comparison_medium_medium012","category":"international_comparison","score":1.0,"tokens":3811223,"cost_usd":2.468744,"elapsed_time":862.12,"gpr":1.0,"tpe":0.0175,"ee":0.0735},{"task_id":"task_456_international_comparison_medium_medium013","category":"international_comparison","score":1.0,"tokens":1393401,"cost_usd":1.067736,"elapsed_time":714.68,"gpr":1.0,"tpe":0.0162,"ee":0.0889},{"task_id":"task_457_international_comparison_medium_medium014","category":"international_comparison","score":0.5,"tokens":1907680,"cost_usd":1.692046,"elapsed_time":1055.27,"gpr":0.5,"tpe":0.0469,"ee":0.0806},{"task_id":"task_458_international_comparison_medium_medium015","category":"international_comparison","score":0.0,"tokens":4483810,"cost_usd":2.86349,"elapsed_time":1207.33,"gpr":0.2857,"tpe":0.0309,"ee":0.0676},{"task_id":"task_459_international_comparison_medium_medium016","category":"international_comparison","score":1.0,"tokens":2300316,"cost_usd":1.718566,"elapsed_time":522.27,"gpr":1.0,"tpe":0.0651,"ee":0.122},{"task_id":"task_460_international_comparison_medium_medium017","category":"international_comparison","score":0.0,"tokens":3334319,"cost_usd":2.683141,"elapsed_time":1207.41,"gpr":0.0,"tpe":0.0,"ee":0.0515},{"task_id":"task_461_international_comparison_medium_medium018","category":"international_comparison","score":0.0,"tokens":4461835,"cost_usd":3.72041,"elapsed_time":1207.37,"gpr":0.0,"tpe":0.0,"ee":0.0714},{"task_id":"task_462_international_comparison_medium_medium019","category":"international_comparison","score":0.0,"tokens":1860465,"cost_usd":1.770745,"elapsed_time":1207.41,"gpr":0.6667,"tpe":0.0186,"ee":0.0784},{"task_id":"task_463_international_comparison_medium_medium020","category":"international_comparison","score":0.0,"tokens":9800021,"cost_usd":5.649096,"elapsed_time":1207.39,"gpr":0.0,"tpe":0.0,"ee":0.0676},{"task_id":"task_464_international_comparison_medium_medium021","category":"international_comparison","score":0.0,"tokens":9514144,"cost_usd":6.14474,"elapsed_time":1207.36,"gpr":0.2857,"tpe":0.0003,"ee":0.0562},{"task_id":"task_465_international_comparison_medium_medium022","category":"international_comparison","score":0.0,"tokens":2511421,"cost_usd":2.746467,"elapsed_time":1207.61,"gpr":0.6364,"tpe":0.4657,"ee":0.0877},{"task_id":"task_466_international_comparison_medium_medium023","category":"international_comparison","score":0.0,"tokens":9098243,"cost_usd":7.312405,"elapsed_time":1207.42,"gpr":0.0,"tpe":0.0,"ee":0.1},{"task_id":"task_467_international_comparison_medium_medium024","category":"international_comparison","score":0.0,"tokens":931290,"cost_usd":1.089367,"elapsed_time":1058.58,"gpr":0.5833,"tpe":0.1886,"ee":0.1136},{"task_id":"task_468_international_comparison_medium_medium025","category":"international_comparison","score":0.0,"tokens":15752159,"cost_usd":11.797438,"elapsed_time":1207.4,"gpr":0.1667,"tpe":0.0523,"ee":0.0588},{"task_id":"task_469_risk_assessment_hard_hard001","category":"risk_assessment","score":1.0,"tokens":5740156,"cost_usd":7.36307,"elapsed_time":606.58,"gpr":1.0,"tpe":0.1786,"ee":0.1556},{"task_id":"task_470_risk_assessment_hard_hard002","category":"risk_assessment","score":1.0,"tokens":4729854,"cost_usd":7.547217,"elapsed_time":1089.69,"gpr":1.0,"tpe":0.1201,"ee":0.0795},{"task_id":"task_471_risk_assessment_hard_hard003","category":"risk_assessment","score":0.0,"tokens":4301306,"cost_usd":6.401549,"elapsed_time":1207.39,"gpr":0.9091,"tpe":0.1291,"ee":0.0986},{"task_id":"task_472_risk_assessment_hard_hard004","category":"risk_assessment","score":0.0,"tokens":5112572,"cost_usd":7.735975,"elapsed_time":1207.5,"gpr":1.0,"tpe":0.202,"ee":0.0824},{"task_id":"task_473_risk_assessment_hard_hard005","category":"risk_assessment","score":1.0,"tokens":2454651,"cost_usd":4.267338,"elapsed_time":1019.42,"gpr":1.0,"tpe":0.0126,"ee":0.1},{"task_id":"task_474_risk_assessment_hard_hard006","category":"risk_assessment","score":1.0,"tokens":2324246,"cost_usd":3.84504,"elapsed_time":995.57,"gpr":1.0,"tpe":0.0168,"ee":0.0972},{"task_id":"task_475_risk_assessment_hard_hard007","category":"risk_assessment","score":1.0,"tokens":2134556,"cost_usd":3.50977,"elapsed_time":908.2,"gpr":1.0,"tpe":0.1989,"ee":0.0886},{"task_id":"task_476_risk_assessment_hard_hard008","category":"risk_assessment","score":0.0,"tokens":666707,"cost_usd":1.169387,"elapsed_time":502.94,"gpr":0.5,"tpe":0.4305,"ee":0.2059},{"task_id":"task_477_risk_assessment_hard_hard009","category":"risk_assessment","score":0.0,"tokens":3110168,"cost_usd":4.742401,"elapsed_time":861.07,"gpr":0.0,"tpe":0.0,"ee":0.1207},{"task_id":"task_478_risk_assessment_hard_hard010","category":"risk_assessment","score":0.0,"tokens":4871138,"cost_usd":7.156004,"elapsed_time":1207.38,"gpr":0.0,"tpe":0.0,"ee":0.0851},{"task_id":"task_479_risk_assessment_hard_hard011","category":"risk_assessment","score":0.0,"tokens":1714568,"cost_usd":2.666373,"elapsed_time":739.03,"gpr":0.0,"tpe":0.0,"ee":0.1273},{"task_id":"task_480_risk_assessment_hard_hard012","category":"risk_assessment","score":0.0,"tokens":3871064,"cost_usd":5.867559,"elapsed_time":1207.54,"gpr":0.25,"tpe":0.0164,"ee":0.1094},{"task_id":"task_481_risk_assessment_hard_hard013","category":"risk_assessment","score":0.0,"tokens":9632831,"cost_usd":12.736679,"elapsed_time":1207.52,"gpr":0.0,"tpe":0.0,"ee":0.1139},{"task_id":"task_482_risk_assessment_medium_medium001","category":"risk_assessment","score":0.0,"tokens":1665997,"cost_usd":2.233629,"elapsed_time":1006.73,"gpr":0.0,"tpe":0.0,"ee":0.0952},{"task_id":"task_483_risk_assessment_medium_medium002","category":"risk_assessment","score":0.0,"tokens":2253233,"cost_usd":3.55992,"elapsed_time":1207.53,"gpr":0.25,"tpe":0.002,"ee":0.0845},{"task_id":"task_484_risk_assessment_medium_medium003","category":"risk_assessment","score":0.0,"tokens":5729713,"cost_usd":8.375373,"elapsed_time":1207.5,"gpr":0.0,"tpe":0.0,"ee":0.0882},{"task_id":"task_485_risk_assessment_medium_medium004","category":"risk_assessment","score":0.5,"tokens":2773751,"cost_usd":4.257434,"elapsed_time":1207.52,"gpr":0.4,"tpe":0.1239,"ee":0.0781},{"task_id":"task_486_risk_assessment_medium_medium005","category":"risk_assessment","score":0.5,"tokens":1426297,"cost_usd":2.545211,"elapsed_time":750.36,"gpr":0.125,"tpe":0.3487,"ee":0.125},{"task_id":"task_487_risk_assessment_medium_medium006","category":"risk_assessment","score":0.5,"tokens":1065562,"cost_usd":1.961057,"elapsed_time":347.58,"gpr":0.4545,"tpe":0.641,"ee":0.1852},{"task_id":"task_488_risk_assessment_medium_medium007","category":"risk_assessment","score":0.33,"tokens":1719551,"cost_usd":3.012946,"elapsed_time":1059.0,"gpr":0.1667,"tpe":0.0161,"ee":0.1111},{"task_id":"task_489_risk_assessment_medium_medium008","category":"risk_assessment","score":0.333,"tokens":781192,"cost_usd":1.321121,"elapsed_time":911.38,"gpr":0.0625,"tpe":0.5314,"ee":0.2},{"task_id":"task_490_risk_assessment_medium_medium009","category":"risk_assessment","score":1.0,"tokens":916604,"cost_usd":1.579332,"elapsed_time":830.08,"gpr":1.0,"tpe":0.2647,"ee":0.1538},{"task_id":"task_491_risk_assessment_medium_medium010","category":"risk_assessment","score":0.0,"tokens":2494454,"cost_usd":3.745941,"elapsed_time":1207.37,"gpr":0.6,"tpe":0.0088,"ee":0.08},{"task_id":"task_492_risk_assessment_medium_medium011","category":"risk_assessment","score":0.0,"tokens":1791142,"cost_usd":2.642732,"elapsed_time":1009.08,"gpr":0.0909,"tpe":0.0382,"ee":0.1071}]},{"model":"minimax-m2.7","acc":0.4126,"ee":0.2555,"gpr":0.3323,"tpe":0.2603,"efficiency":{"total_tokens":552907365,"total_cost_usd":61.15,"total_requests":13806},"category_scores":{"comprehensive_decision":{"acc":0.381,"ee":0.2272,"gpr":0.2954,"tpe":0.2452,"count":70},"enterprise_industry_analysis":{"acc":0.5487,"ee":0.2663,"gpr":0.373,"tpe":0.2367,"count":226},"enterprise_industry_policy_analysis":{"acc":0.3158,"ee":0.2561,"gpr":0.3462,"tpe":0.2938,"count":76},"hypothesis_verification":{"acc":0.069,"ee":0.3154,"gpr":0.1788,"tpe":0.3493,"count":29},"industry_planning":{"acc":0.3214,"ee":0.1727,"gpr":0.272,"tpe":0.305,"count":28},"international_comparison":{"acc":0.2179,"ee":0.2111,"gpr":0.3902,"tpe":0.1923,"count":39},"risk_assessment":{"acc":0.3681,"ee":0.2493,"gpr":0.3341,"tpe":0.296,"count":24}},"tasks":[{"task_id":"task_001_comprehensive_decision_easy_easy001","category":"comprehensive_decision","score":0.0,"tokens":141280,"cost_usd":0.018903,"elapsed_time":67.15,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_002_comprehensive_decision_easy_easy002","category":"comprehensive_decision","score":1.0,"tokens":164882,"cost_usd":0.023021,"elapsed_time":88.76,"gpr":1.0,"tpe":0.4168,"ee":0.1667},{"task_id":"task_003_comprehensive_decision_easy_easy003","category":"comprehensive_decision","score":1.0,"tokens":217303,"cost_usd":0.040506,"elapsed_time":70.07,"gpr":1.0,"tpe":0.6011,"ee":0.3333},{"task_id":"task_004_comprehensive_decision_easy_easy004","category":"comprehensive_decision","score":1.0,"tokens":150200,"cost_usd":0.032468,"elapsed_time":68.12,"gpr":1.0,"tpe":0.7366,"ee":0.25},{"task_id":"task_005_comprehensive_decision_easy_easy005","category":"comprehensive_decision","score":0.0,"tokens":1684350,"cost_usd":0.161177,"elapsed_time":396.63,"gpr":0.3333,"tpe":0.6561,"ee":0.0577},{"task_id":"task_006_comprehensive_decision_easy_easy006","category":"comprehensive_decision","score":0.0,"tokens":1772115,"cost_usd":0.203229,"elapsed_time":316.31,"gpr":0.0,"tpe":0.0,"ee":0.0789},{"task_id":"task_007_comprehensive_decision_hard_hard001","category":"comprehensive_decision","score":0.0,"tokens":2101193,"cost_usd":0.335079,"elapsed_time":231.51,"gpr":0.125,"tpe":1.0,"ee":0.3333},{"task_id":"task_008_comprehensive_decision_hard_hard002","category":"comprehensive_decision","score":1.0,"tokens":1056906,"cost_usd":0.142967,"elapsed_time":340.54,"gpr":1.0,"tpe":0.1981,"ee":0.3043},{"task_id":"task_009_comprehensive_decision_hard_hard003","category":"comprehensive_decision","score":0.0,"tokens":1007068,"cost_usd":0.117198,"elapsed_time":292.5,"gpr":0.5714,"tpe":0.3982,"ee":0.2593},{"task_id":"task_010_comprehensive_decision_hard_hard004","category":"comprehensive_decision","score":0.0,"tokens":1761307,"cost_usd":0.182542,"elapsed_time":451.37,"gpr":0.125,"tpe":0.3487,"ee":0.1667},{"task_id":"task_011_comprehensive_decision_hard_hard005","category":"comprehensive_decision","score":0.0,"tokens":3371961,"cost_usd":0.3137,"elapsed_time":790.37,"gpr":0.1429,"tpe":1.0,"ee":0.0946},{"task_id":"task_012_comprehensive_decision_hard_hard006","category":"comprehensive_decision","score":0.0,"tokens":6053749,"cost_usd":0.646588,"elapsed_time":1039.32,"gpr":0.1667,"tpe":0.5314,"ee":0.1077},{"task_id":"task_013_comprehensive_decision_hard_hard007","category":"comprehensive_decision","score":0.0,"tokens":1459229,"cost_usd":0.156783,"elapsed_time":467.99,"gpr":0.6667,"tpe":0.1012,"ee":0.1667},{"task_id":"task_014_comprehensive_decision_hard_hard008","category":"comprehensive_decision","score":0.0,"tokens":792234,"cost_usd":0.090836,"elapsed_time":310.32,"gpr":0.0,"tpe":0.0,"ee":0.2593},{"task_id":"task_015_comprehensive_decision_hard_hard009","category":"comprehensive_decision","score":0.0,"tokens":3300531,"cost_usd":0.534091,"elapsed_time":537.31,"gpr":0.0,"tpe":0.0,"ee":0.2353},{"task_id":"task_016_comprehensive_decision_hard_hard010","category":"comprehensive_decision","score":0.0,"tokens":3654600,"cost_usd":0.323764,"elapsed_time":678.22,"gpr":0.1429,"tpe":0.0225,"ee":0.1296},{"task_id":"task_017_comprehensive_decision_hard_hard011","category":"comprehensive_decision","score":0.0,"tokens":763576,"cost_usd":0.129935,"elapsed_time":190.03,"gpr":0.0,"tpe":0.0,"ee":0.3684},{"task_id":"task_018_comprehensive_decision_hard_hard012","category":"comprehensive_decision","score":0.0,"tokens":734642,"cost_usd":0.10764,"elapsed_time":314.51,"gpr":0.1667,"tpe":1.0,"ee":0.2692},{"task_id":"task_019_comprehensive_decision_hard_hard013","category":"comprehensive_decision","score":0.5,"tokens":687480,"cost_usd":0.076212,"elapsed_time":287.61,"gpr":1.0,"tpe":0.1158,"ee":0.2414},{"task_id":"task_020_comprehensive_decision_hard_hard014","category":"comprehensive_decision","score":0.0,"tokens":806674,"cost_usd":0.098558,"elapsed_time":361.87,"gpr":0.2857,"tpe":0.3874,"ee":0.2414},{"task_id":"task_021_comprehensive_decision_hard_hard015","category":"comprehensive_decision","score":0.0,"tokens":2811650,"cost_usd":0.267791,"elapsed_time":450.18,"gpr":0.4286,"tpe":0.0718,"ee":0.2162},{"task_id":"task_022_comprehensive_decision_hard_hard016","category":"comprehensive_decision","score":0.0,"tokens":1271660,"cost_usd":0.136035,"elapsed_time":443.7,"gpr":0.625,"tpe":0.0325,"ee":0.1364},{"task_id":"task_023_comprehensive_decision_hard_hard017","category":"comprehensive_decision","score":0.0,"tokens":1251212,"cost_usd":0.178783,"elapsed_time":404.37,"gpr":1.0,"tpe":0.4895,"ee":0.3529},{"task_id":"task_024_comprehensive_decision_hard_hard018","category":"comprehensive_decision","score":1.0,"tokens":1994000,"cost_usd":0.199672,"elapsed_time":407.61,"gpr":1.0,"tpe":0.3334,"ee":0.3182},{"task_id":"task_025_comprehensive_decision_hard_hard019","category":"comprehensive_decision","score":0.67,"tokens":201880,"cost_usd":0.03929,"elapsed_time":149.85,"gpr":0.7143,"tpe":0.6561,"ee":0.5455},{"task_id":"task_026_comprehensive_decision_medium_medium001","category":"comprehensive_decision","score":0.5,"tokens":1244081,"cost_usd":0.14147,"elapsed_time":456.4,"gpr":0.875,"tpe":0.2529,"ee":0.1429},{"task_id":"task_027_comprehensive_decision_medium_medium002","category":"comprehensive_decision","score":1.0,"tokens":1397651,"cost_usd":0.140707,"elapsed_time":440.95,"gpr":1.0,"tpe":0.4586,"ee":0.125},{"task_id":"task_028_comprehensive_decision_medium_medium003","category":"comprehensive_decision","score":0.0,"tokens":945348,"cost_usd":0.101861,"elapsed_time":404.48,"gpr":0.4286,"tpe":0.3223,"ee":0.1316},{"task_id":"task_029_comprehensive_decision_medium_medium004","category":"comprehensive_decision","score":0.5,"tokens":878126,"cost_usd":0.101436,"elapsed_time":399.68,"gpr":0.875,"tpe":0.2312,"ee":0.1515},{"task_id":"task_030_comprehensive_decision_medium_medium005","category":"comprehensive_decision","score":1.0,"tokens":1183051,"cost_usd":0.100246,"elapsed_time":244.72,"gpr":1.0,"tpe":0.0384,"ee":0.125},{"task_id":"task_031_comprehensive_decision_medium_medium006","category":"comprehensive_decision","score":1.0,"tokens":214895,"cost_usd":0.058656,"elapsed_time":57.08,"gpr":1.0,"tpe":0.855,"ee":0.6667},{"task_id":"task_032_comprehensive_decision_medium_medium007","category":"comprehensive_decision","score":0.0,"tokens":1430414,"cost_usd":0.141357,"elapsed_time":309.52,"gpr":0.0,"tpe":0.0,"ee":0.1724},{"task_id":"task_033_comprehensive_decision_medium_medium008","category":"comprehensive_decision","score":0.0,"tokens":1296758,"cost_usd":0.116404,"elapsed_time":347.15,"gpr":0.0,"tpe":0.0,"ee":0.119},{"task_id":"task_034_comprehensive_decision_medium_medium009","category":"comprehensive_decision","score":1.0,"tokens":300827,"cost_usd":0.033323,"elapsed_time":122.42,"gpr":1.0,"tpe":0.2527,"ee":0.2381},{"task_id":"task_035_comprehensive_decision_medium_medium010","category":"comprehensive_decision","score":0.0,"tokens":83298,"cost_usd":0.014715,"elapsed_time":48.02,"gpr":1.0,"tpe":1.0,"ee":0.8333},{"task_id":"task_036_comprehensive_decision_medium_medium011","category":"comprehensive_decision","score":0.0,"tokens":3344729,"cost_usd":0.291247,"elapsed_time":602.53,"gpr":0.3333,"tpe":0.0108,"ee":0.0658},{"task_id":"task_037_comprehensive_decision_medium_medium012","category":"comprehensive_decision","score":1.0,"tokens":763556,"cost_usd":0.083319,"elapsed_time":191.44,"gpr":1.0,"tpe":0.1807,"ee":0.2},{"task_id":"task_038_comprehensive_decision_medium_medium013","category":"comprehensive_decision","score":0.5,"tokens":2055707,"cost_usd":0.17252,"elapsed_time":409.64,"gpr":0.6,"tpe":0.1708,"ee":0.0781},{"task_id":"task_039_comprehensive_decision_medium_medium014","category":"comprehensive_decision","score":1.0,"tokens":1771123,"cost_usd":0.174803,"elapsed_time":335.56,"gpr":1.0,"tpe":0.2562,"ee":0.1724},{"task_id":"task_040_comprehensive_decision_medium_medium015","category":"comprehensive_decision","score":1.0,"tokens":602516,"cost_usd":0.107669,"elapsed_time":311.63,"gpr":1.0,"tpe":0.5265,"ee":0.2941},{"task_id":"task_041_comprehensive_decision_medium_medium016","category":"comprehensive_decision","score":0.0,"tokens":6984992,"cost_usd":0.566496,"elapsed_time":1207.55,"gpr":0.0,"tpe":0.0,"ee":0.0424},{"task_id":"task_042_comprehensive_decision_medium_medium017","category":"comprehensive_decision","score":1.0,"tokens":2578671,"cost_usd":0.218078,"elapsed_time":345.65,"gpr":1.0,"tpe":0.0058,"ee":0.0909},{"task_id":"task_043_comprehensive_decision_medium_medium018","category":"comprehensive_decision","score":1.0,"tokens":914281,"cost_usd":0.103794,"elapsed_time":145.05,"gpr":1.0,"tpe":0.3608,"ee":0.3125},{"task_id":"task_044_comprehensive_decision_medium_medium019","category":"comprehensive_decision","score":0.0,"tokens":990859,"cost_usd":0.110316,"elapsed_time":229.29,"gpr":0.0,"tpe":0.0,"ee":0.1282},{"task_id":"task_045_comprehensive_decision_medium_medium020","category":"comprehensive_decision","score":0.0,"tokens":1825566,"cost_usd":0.151834,"elapsed_time":283.73,"gpr":0.0,"tpe":0.0,"ee":0.098},{"task_id":"task_046_comprehensive_decision_medium_medium021","category":"comprehensive_decision","score":0.0,"tokens":1590948,"cost_usd":0.139085,"elapsed_time":253.41,"gpr":0.0,"tpe":0.0,"ee":0.1515},{"task_id":"task_047_comprehensive_decision_medium_medium022","category":"comprehensive_decision","score":1.0,"tokens":9722167,"cost_usd":0.91101,"elapsed_time":815.6,"gpr":1.0,"tpe":0.0003,"ee":0.0617},{"task_id":"task_048_comprehensive_decision_medium_medium023","category":"comprehensive_decision","score":1.0,"tokens":1989739,"cost_usd":0.206167,"elapsed_time":222.52,"gpr":1.0,"tpe":0.0488,"ee":0.1176},{"task_id":"task_049_comprehensive_decision_medium_medium024","category":"comprehensive_decision","score":0.0,"tokens":282118,"cost_usd":0.041708,"elapsed_time":151.88,"gpr":0.0,"tpe":0.0,"ee":0.3571},{"task_id":"task_050_comprehensive_decision_medium_medium025","category":"comprehensive_decision","score":0.0,"tokens":769419,"cost_usd":0.090967,"elapsed_time":368.75,"gpr":0.4,"tpe":0.4783,"ee":0.1515},{"task_id":"task_051_comprehensive_decision_medium_medium026","category":"comprehensive_decision","score":1.0,"tokens":597798,"cost_usd":0.081102,"elapsed_time":126.87,"gpr":1.0,"tpe":0.1501,"ee":0.2174},{"task_id":"task_052_comprehensive_decision_medium_medium027","category":"comprehensive_decision","score":0.0,"tokens":153019,"cost_usd":0.029711,"elapsed_time":140.24,"gpr":0.0,"tpe":0.0,"ee":0.5556},{"task_id":"task_053_comprehensive_decision_medium_medium028","category":"comprehensive_decision","score":0.0,"tokens":710508,"cost_usd":0.099457,"elapsed_time":373.77,"gpr":0.5,"tpe":0.0798,"ee":0.1724},{"task_id":"task_054_comprehensive_decision_medium_medium029","category":"comprehensive_decision","score":0.0,"tokens":670608,"cost_usd":0.090687,"elapsed_time":87.34,"gpr":0.5,"tpe":1.0,"ee":0.4545},{"task_id":"task_055_comprehensive_decision_medium_medium030","category":"comprehensive_decision","score":1.0,"tokens":479088,"cost_usd":0.056528,"elapsed_time":214.64,"gpr":1.0,"tpe":0.162,"ee":0.2083},{"task_id":"task_056_comprehensive_decision_medium_medium031","category":"comprehensive_decision","score":1.0,"tokens":734325,"cost_usd":0.069924,"elapsed_time":177.99,"gpr":1.0,"tpe":0.273,"ee":0.1667},{"task_id":"task_057_comprehensive_decision_medium_medium032","category":"comprehensive_decision","score":0.0,"tokens":464867,"cost_usd":0.045456,"elapsed_time":155.87,"gpr":0.0,"tpe":0.0,"ee":0.1786},{"task_id":"task_058_comprehensive_decision_medium_medium033","category":"comprehensive_decision","score":1.0,"tokens":437314,"cost_usd":0.05264,"elapsed_time":152.54,"gpr":1.0,"tpe":0.1958,"ee":0.2174},{"task_id":"task_059_comprehensive_decision_medium_medium034","category":"comprehensive_decision","score":0.0,"tokens":698404,"cost_usd":0.073401,"elapsed_time":335.05,"gpr":0.0,"tpe":0.0,"ee":0.1562},{"task_id":"task_060_comprehensive_decision_medium_medium035","category":"comprehensive_decision","score":0.0,"tokens":2905964,"cost_usd":0.402234,"elapsed_time":281.96,"gpr":0.0,"tpe":0.0,"ee":0.1351},{"task_id":"task_061_comprehensive_decision_medium_medium036","category":"comprehensive_decision","score":1.0,"tokens":631564,"cost_usd":0.129354,"elapsed_time":360.24,"gpr":1.0,"tpe":0.2929,"ee":0.2778},{"task_id":"task_062_comprehensive_decision_medium_medium037","category":"comprehensive_decision","score":0.0,"tokens":909702,"cost_usd":0.147841,"elapsed_time":837.66,"gpr":1.0,"tpe":0.5314,"ee":0.1923},{"task_id":"task_063_comprehensive_decision_medium_medium038","category":"comprehensive_decision","score":0.0,"tokens":556801,"cost_usd":0.05945,"elapsed_time":217.49,"gpr":0.3333,"tpe":0.1351,"ee":0.129},{"task_id":"task_064_comprehensive_decision_medium_medium039","category":"comprehensive_decision","score":0.0,"tokens":229765,"cost_usd":0.026066,"elapsed_time":100.23,"gpr":0.0,"tpe":0.0,"ee":0.2353},{"task_id":"task_065_comprehensive_decision_medium_medium040","category":"comprehensive_decision","score":1.0,"tokens":2058283,"cost_usd":0.185601,"elapsed_time":497.69,"gpr":1.0,"tpe":0.0225,"ee":0.1},{"task_id":"task_066_comprehensive_decision_medium_medium041","category":"comprehensive_decision","score":1.0,"tokens":182130,"cost_usd":0.025486,"elapsed_time":132.59,"gpr":1.0,"tpe":0.4305,"ee":0.3846},{"task_id":"task_067_comprehensive_decision_medium_medium042","category":"comprehensive_decision","score":0.0,"tokens":2752464,"cost_usd":0.275879,"elapsed_time":336.49,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_068_comprehensive_decision_medium_medium043","category":"comprehensive_decision","score":0.0,"tokens":141578,"cost_usd":0.022874,"elapsed_time":66.91,"gpr":0.0,"tpe":0.0,"ee":0.4444},{"task_id":"task_069_comprehensive_decision_medium_medium044","category":"comprehensive_decision","score":0.0,"tokens":663118,"cost_usd":0.07859,"elapsed_time":208.36,"gpr":0.25,"tpe":0.2542,"ee":0.1852},{"task_id":"task_070_comprehensive_decision_medium_medium045","category":"comprehensive_decision","score":1.0,"tokens":1628005,"cost_usd":0.166673,"elapsed_time":312.87,"gpr":1.0,"tpe":0.0117,"ee":0.1042},{"task_id":"task_071_enterprise_industry_analysis_easy_easy001","category":"enterprise_industry_analysis","score":1.0,"tokens":200129,"cost_usd":0.026443,"elapsed_time":77.48,"gpr":1.0,"tpe":0.3636,"ee":0.2143},{"task_id":"task_072_enterprise_industry_analysis_easy_easy002","category":"enterprise_industry_analysis","score":0.0,"tokens":287665,"cost_usd":0.031223,"elapsed_time":101.85,"gpr":1.0,"tpe":0.2821,"ee":0.1429},{"task_id":"task_073_enterprise_industry_analysis_easy_easy003","category":"enterprise_industry_analysis","score":0.0,"tokens":215820,"cost_usd":0.03059,"elapsed_time":88.5,"gpr":0.0,"tpe":0.0,"ee":0.1875},{"task_id":"task_074_enterprise_industry_analysis_easy_easy004","category":"enterprise_industry_analysis","score":0.0,"tokens":925451,"cost_usd":0.090009,"elapsed_time":211.8,"gpr":0.0,"tpe":0.0,"ee":0.075},{"task_id":"task_075_enterprise_industry_analysis_easy_easy005","category":"enterprise_industry_analysis","score":1.0,"tokens":1896395,"cost_usd":0.15993,"elapsed_time":222.97,"gpr":1.0,"tpe":0.0457,"ee":0.0909},{"task_id":"task_076_enterprise_industry_analysis_easy_easy006","category":"enterprise_industry_analysis","score":0.0,"tokens":279868,"cost_usd":0.030465,"elapsed_time":99.67,"gpr":0.0,"tpe":0.0,"ee":0.1667},{"task_id":"task_077_enterprise_industry_analysis_easy_easy007","category":"enterprise_industry_analysis","score":1.0,"tokens":330722,"cost_usd":0.044648,"elapsed_time":71.5,"gpr":1.0,"tpe":0.5137,"ee":0.3},{"task_id":"task_078_enterprise_industry_analysis_easy_easy008","category":"enterprise_industry_analysis","score":0.0,"tokens":980118,"cost_usd":0.100366,"elapsed_time":155.91,"gpr":0.6667,"tpe":0.1155,"ee":0.1111},{"task_id":"task_079_enterprise_industry_analysis_easy_easy009","category":"enterprise_industry_analysis","score":1.0,"tokens":454541,"cost_usd":0.046461,"elapsed_time":134.71,"gpr":1.0,"tpe":0.1175,"ee":0.125},{"task_id":"task_080_enterprise_industry_analysis_easy_easy010","category":"enterprise_industry_analysis","score":1.0,"tokens":1193919,"cost_usd":0.116286,"elapsed_time":178.19,"gpr":1.0,"tpe":0.0625,"ee":0.1},{"task_id":"task_081_enterprise_industry_analysis_easy_easy011","category":"enterprise_industry_analysis","score":1.0,"tokens":290980,"cost_usd":0.040845,"elapsed_time":87.95,"gpr":1.0,"tpe":0.2824,"ee":0.2},{"task_id":"task_082_enterprise_industry_analysis_easy_easy012","category":"enterprise_industry_analysis","score":1.0,"tokens":256683,"cost_usd":0.031414,"elapsed_time":103.02,"gpr":1.0,"tpe":0.1853,"ee":0.1579},{"task_id":"task_083_enterprise_industry_analysis_easy_easy013","category":"enterprise_industry_analysis","score":0.0,"tokens":555281,"cost_usd":0.053295,"elapsed_time":178.77,"gpr":0.0,"tpe":0.0,"ee":0.0938},{"task_id":"task_084_enterprise_industry_analysis_easy_easy014","category":"enterprise_industry_analysis","score":1.0,"tokens":390849,"cost_usd":0.039583,"elapsed_time":116.74,"gpr":1.0,"tpe":0.0886,"ee":0.1154},{"task_id":"task_085_enterprise_industry_analysis_easy_easy015","category":"enterprise_industry_analysis","score":1.0,"tokens":206361,"cost_usd":0.024106,"elapsed_time":71.62,"gpr":1.0,"tpe":0.3371,"ee":0.2143},{"task_id":"task_086_enterprise_industry_analysis_easy_easy016","category":"enterprise_industry_analysis","score":0.0,"tokens":113200,"cost_usd":0.016825,"elapsed_time":53.48,"gpr":1.0,"tpe":0.6342,"ee":0.375},{"task_id":"task_087_enterprise_industry_analysis_easy_easy017","category":"enterprise_industry_analysis","score":0.0,"tokens":2958100,"cost_usd":0.537311,"elapsed_time":278.13,"gpr":0.0,"tpe":0.0,"ee":0.0882},{"task_id":"task_088_enterprise_industry_analysis_easy_easy018","category":"enterprise_industry_analysis","score":0.0,"tokens":777335,"cost_usd":0.080251,"elapsed_time":217.18,"gpr":0.0,"tpe":0.0,"ee":0.0789},{"task_id":"task_089_enterprise_industry_analysis_easy_easy019","category":"enterprise_industry_analysis","score":1.0,"tokens":247216,"cost_usd":0.026874,"elapsed_time":91.18,"gpr":1.0,"tpe":0.2059,"ee":0.1667},{"task_id":"task_090_enterprise_industry_analysis_easy_easy020","category":"enterprise_industry_analysis","score":0.0,"tokens":842820,"cost_usd":0.098655,"elapsed_time":147.84,"gpr":0.0,"tpe":0.0,"ee":0.1304},{"task_id":"task_091_enterprise_industry_analysis_easy_easy021","category":"enterprise_industry_analysis","score":1.0,"tokens":288036,"cost_usd":0.033529,"elapsed_time":102.8,"gpr":1.0,"tpe":0.1507,"ee":0.1364},{"task_id":"task_092_enterprise_industry_analysis_easy_easy022","category":"enterprise_industry_analysis","score":1.0,"tokens":416198,"cost_usd":0.041921,"elapsed_time":127.91,"gpr":1.0,"tpe":0.1058,"ee":0.12},{"task_id":"task_093_enterprise_industry_analysis_easy_easy023","category":"enterprise_industry_analysis","score":1.0,"tokens":207309,"cost_usd":0.024974,"elapsed_time":87.04,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_094_enterprise_industry_analysis_easy_easy024","category":"enterprise_industry_analysis","score":1.0,"tokens":259332,"cost_usd":0.027226,"elapsed_time":91.66,"gpr":1.0,"tpe":0.186,"ee":0.15},{"task_id":"task_095_enterprise_industry_analysis_easy_easy025","category":"enterprise_industry_analysis","score":1.0,"tokens":346474,"cost_usd":0.040494,"elapsed_time":148.91,"gpr":1.0,"tpe":0.1501,"ee":0.1429},{"task_id":"task_096_enterprise_industry_analysis_easy_easy026","category":"enterprise_industry_analysis","score":1.0,"tokens":150977,"cost_usd":0.018835,"elapsed_time":62.06,"gpr":1.0,"tpe":0.4498,"ee":0.25},{"task_id":"task_097_enterprise_industry_analysis_easy_easy027","category":"enterprise_industry_analysis","score":1.0,"tokens":653697,"cost_usd":0.086835,"elapsed_time":93.46,"gpr":1.0,"tpe":0.3034,"ee":0.2},{"task_id":"task_098_enterprise_industry_analysis_easy_easy028","category":"enterprise_industry_analysis","score":0.0,"tokens":514063,"cost_usd":0.058288,"elapsed_time":147.92,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_099_enterprise_industry_analysis_easy_easy029","category":"enterprise_industry_analysis","score":1.0,"tokens":224126,"cost_usd":0.031125,"elapsed_time":101.18,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_100_enterprise_industry_analysis_easy_easy030","category":"enterprise_industry_analysis","score":1.0,"tokens":507866,"cost_usd":0.054211,"elapsed_time":154.66,"gpr":1.0,"tpe":0.0582,"ee":0.1},{"task_id":"task_101_enterprise_industry_analysis_easy_easy031","category":"enterprise_industry_analysis","score":1.0,"tokens":542510,"cost_usd":0.057537,"elapsed_time":150.7,"gpr":1.0,"tpe":0.0627,"ee":0.1},{"task_id":"task_102_enterprise_industry_analysis_easy_easy032","category":"enterprise_industry_analysis","score":1.0,"tokens":414801,"cost_usd":0.045587,"elapsed_time":143.85,"gpr":1.0,"tpe":0.0857,"ee":0.1111},{"task_id":"task_103_enterprise_industry_analysis_easy_easy033","category":"enterprise_industry_analysis","score":0.0,"tokens":360625,"cost_usd":0.045729,"elapsed_time":159.21,"gpr":0.3333,"tpe":0.1501,"ee":0.1304},{"task_id":"task_104_enterprise_industry_analysis_easy_easy034","category":"enterprise_industry_analysis","score":1.0,"tokens":576193,"cost_usd":0.063471,"elapsed_time":233.34,"gpr":1.0,"tpe":0.0309,"ee":0.0833},{"task_id":"task_105_enterprise_industry_analysis_easy_easy035","category":"enterprise_industry_analysis","score":0.0,"tokens":228173,"cost_usd":0.04182,"elapsed_time":64.54,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_106_enterprise_industry_analysis_easy_easy036","category":"enterprise_industry_analysis","score":1.0,"tokens":97410,"cost_usd":0.018813,"elapsed_time":42.62,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_107_enterprise_industry_analysis_easy_easy037","category":"enterprise_industry_analysis","score":1.0,"tokens":204870,"cost_usd":0.022323,"elapsed_time":78.25,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_108_enterprise_industry_analysis_easy_easy038","category":"enterprise_industry_analysis","score":0.0,"tokens":104346,"cost_usd":0.014556,"elapsed_time":45.11,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_109_enterprise_industry_analysis_easy_easy039","category":"enterprise_industry_analysis","score":1.0,"tokens":81199,"cost_usd":0.012785,"elapsed_time":42.33,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_110_enterprise_industry_analysis_easy_easy040","category":"enterprise_industry_analysis","score":0.0,"tokens":410847,"cost_usd":0.056603,"elapsed_time":66.67,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_111_enterprise_industry_analysis_easy_easy041","category":"enterprise_industry_analysis","score":1.0,"tokens":339195,"cost_usd":0.041438,"elapsed_time":89.13,"gpr":1.0,"tpe":0.2288,"ee":0.1765},{"task_id":"task_112_enterprise_industry_analysis_easy_easy042","category":"enterprise_industry_analysis","score":1.0,"tokens":57171,"cost_usd":0.009691,"elapsed_time":33.46,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_113_enterprise_industry_analysis_easy_easy043","category":"enterprise_industry_analysis","score":1.0,"tokens":130412,"cost_usd":0.015941,"elapsed_time":51.33,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_114_enterprise_industry_analysis_easy_easy044","category":"enterprise_industry_analysis","score":1.0,"tokens":46377,"cost_usd":0.009252,"elapsed_time":30.22,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_115_enterprise_industry_analysis_easy_easy045","category":"enterprise_industry_analysis","score":1.0,"tokens":302938,"cost_usd":0.03322,"elapsed_time":92.63,"gpr":1.0,"tpe":0.2059,"ee":0.1667},{"task_id":"task_116_enterprise_industry_analysis_easy_easy046","category":"enterprise_industry_analysis","score":1.0,"tokens":196090,"cost_usd":0.024538,"elapsed_time":89.97,"gpr":1.0,"tpe":0.3138,"ee":0.2143},{"task_id":"task_117_enterprise_industry_analysis_easy_easy047","category":"enterprise_industry_analysis","score":1.0,"tokens":329519,"cost_usd":0.038723,"elapsed_time":122.65,"gpr":1.0,"tpe":0.1216,"ee":0.1304},{"task_id":"task_118_enterprise_industry_analysis_easy_easy048","category":"enterprise_industry_analysis","score":1.0,"tokens":80409,"cost_usd":0.012047,"elapsed_time":39.13,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_119_enterprise_industry_analysis_easy_easy049","category":"enterprise_industry_analysis","score":0.0,"tokens":309611,"cost_usd":0.054059,"elapsed_time":67.04,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_120_enterprise_industry_analysis_easy_easy050","category":"enterprise_industry_analysis","score":1.0,"tokens":45993,"cost_usd":0.009462,"elapsed_time":30.6,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_121_enterprise_industry_analysis_easy_easy051","category":"enterprise_industry_analysis","score":0.0,"tokens":91862,"cost_usd":0.011392,"elapsed_time":40.56,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_122_enterprise_industry_analysis_easy_easy052","category":"enterprise_industry_analysis","score":1.0,"tokens":57813,"cost_usd":0.010406,"elapsed_time":31.23,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_123_enterprise_industry_analysis_easy_easy053","category":"enterprise_industry_analysis","score":1.0,"tokens":226676,"cost_usd":0.029468,"elapsed_time":76.36,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_124_enterprise_industry_analysis_easy_easy054","category":"enterprise_industry_analysis","score":1.0,"tokens":94014,"cost_usd":0.012303,"elapsed_time":43.4,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_125_enterprise_industry_analysis_easy_easy055","category":"enterprise_industry_analysis","score":0.0,"tokens":225062,"cost_usd":0.035366,"elapsed_time":56.27,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_126_enterprise_industry_analysis_easy_easy056","category":"enterprise_industry_analysis","score":1.0,"tokens":1518296,"cost_usd":0.138552,"elapsed_time":320.6,"gpr":1.0,"tpe":0.0042,"ee":0.0545},{"task_id":"task_127_enterprise_industry_analysis_easy_easy057","category":"enterprise_industry_analysis","score":0.0,"tokens":128660,"cost_usd":0.017055,"elapsed_time":55.97,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_128_enterprise_industry_analysis_easy_easy058","category":"enterprise_industry_analysis","score":0.0,"tokens":1135620,"cost_usd":0.203033,"elapsed_time":126.36,"gpr":0.0,"tpe":0.0,"ee":0.2143},{"task_id":"task_129_enterprise_industry_analysis_easy_easy059","category":"enterprise_industry_analysis","score":0.0,"tokens":68254,"cost_usd":0.016518,"elapsed_time":43.68,"gpr":0.0,"tpe":0.0,"ee":0.75},{"task_id":"task_130_enterprise_industry_analysis_easy_easy060","category":"enterprise_industry_analysis","score":1.0,"tokens":340075,"cost_usd":0.03943,"elapsed_time":159.42,"gpr":1.0,"tpe":0.1216,"ee":0.1304},{"task_id":"task_131_enterprise_industry_analysis_easy_easy061","category":"enterprise_industry_analysis","score":1.0,"tokens":550359,"cost_usd":0.080273,"elapsed_time":146.05,"gpr":1.0,"tpe":0.1668,"ee":0.15},{"task_id":"task_132_enterprise_industry_analysis_easy_easy062","category":"enterprise_industry_analysis","score":0.0,"tokens":640672,"cost_usd":0.10046,"elapsed_time":130.02,"gpr":0.6667,"tpe":0.2981,"ee":0.1667},{"task_id":"task_133_enterprise_industry_analysis_easy_easy063","category":"enterprise_industry_analysis","score":1.0,"tokens":125325,"cost_usd":0.017096,"elapsed_time":61.01,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_134_enterprise_industry_analysis_easy_easy064","category":"enterprise_industry_analysis","score":1.0,"tokens":195353,"cost_usd":0.032015,"elapsed_time":78.34,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_135_enterprise_industry_analysis_easy_easy065","category":"enterprise_industry_analysis","score":0.0,"tokens":261560,"cost_usd":0.029115,"elapsed_time":117.52,"gpr":0.6667,"tpe":0.3156,"ee":0.1765},{"task_id":"task_136_enterprise_industry_analysis_easy_easy066","category":"enterprise_industry_analysis","score":1.0,"tokens":362266,"cost_usd":0.059198,"elapsed_time":82.42,"gpr":1.0,"tpe":0.4161,"ee":0.25},{"task_id":"task_137_enterprise_industry_analysis_easy_easy067","category":"enterprise_industry_analysis","score":1.0,"tokens":100475,"cost_usd":0.018065,"elapsed_time":62.15,"gpr":1.0,"tpe":0.6828,"ee":0.375},{"task_id":"task_138_enterprise_industry_analysis_easy_easy068","category":"enterprise_industry_analysis","score":1.0,"tokens":263848,"cost_usd":0.044056,"elapsed_time":59.51,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_139_enterprise_industry_analysis_easy_easy069","category":"enterprise_industry_analysis","score":0.0,"tokens":467913,"cost_usd":0.049799,"elapsed_time":173.5,"gpr":1.0,"tpe":0.1704,"ee":0.1071},{"task_id":"task_140_enterprise_industry_analysis_easy_easy070","category":"enterprise_industry_analysis","score":0.0,"tokens":523153,"cost_usd":0.066176,"elapsed_time":284.74,"gpr":1.0,"tpe":0.6243,"ee":0.1154},{"task_id":"task_141_enterprise_industry_analysis_easy_easy071","category":"enterprise_industry_analysis","score":1.0,"tokens":891406,"cost_usd":0.092872,"elapsed_time":112.91,"gpr":1.0,"tpe":0.1501,"ee":0.1429},{"task_id":"task_142_enterprise_industry_analysis_easy_easy072","category":"enterprise_industry_analysis","score":1.0,"tokens":262304,"cost_usd":0.02874,"elapsed_time":90.04,"gpr":1.0,"tpe":0.3792,"ee":0.1765},{"task_id":"task_143_enterprise_industry_analysis_easy_easy073","category":"enterprise_industry_analysis","score":0.0,"tokens":114039,"cost_usd":0.021065,"elapsed_time":46.98,"gpr":0.3333,"tpe":0.729,"ee":0.4286},{"task_id":"task_144_enterprise_industry_analysis_easy_easy074","category":"enterprise_industry_analysis","score":1.0,"tokens":62659,"cost_usd":0.011482,"elapsed_time":35.54,"gpr":1.0,"tpe":0.87,"ee":0.6},{"task_id":"task_145_enterprise_industry_analysis_easy_easy075","category":"enterprise_industry_analysis","score":1.0,"tokens":110612,"cost_usd":0.014237,"elapsed_time":60.38,"gpr":1.0,"tpe":0.6631,"ee":0.3333},{"task_id":"task_146_enterprise_industry_analysis_easy_easy076","category":"enterprise_industry_analysis","score":1.0,"tokens":95071,"cost_usd":0.013709,"elapsed_time":45.23,"gpr":1.0,"tpe":0.6342,"ee":0.375},{"task_id":"task_147_enterprise_industry_analysis_easy_easy077","category":"enterprise_industry_analysis","score":1.0,"tokens":76039,"cost_usd":0.013223,"elapsed_time":43.59,"gpr":1.0,"tpe":0.9097,"ee":0.5},{"task_id":"task_148_enterprise_industry_analysis_easy_easy078","category":"enterprise_industry_analysis","score":1.0,"tokens":69746,"cost_usd":0.013198,"elapsed_time":44.34,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_149_enterprise_industry_analysis_easy_easy079","category":"enterprise_industry_analysis","score":1.0,"tokens":57124,"cost_usd":0.010023,"elapsed_time":30.55,"gpr":1.0,"tpe":0.87,"ee":0.6},{"task_id":"task_150_enterprise_industry_analysis_easy_easy080","category":"enterprise_industry_analysis","score":1.0,"tokens":74588,"cost_usd":0.015295,"elapsed_time":38.21,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_151_enterprise_industry_analysis_easy_easy081","category":"enterprise_industry_analysis","score":0.0,"tokens":134234,"cost_usd":0.021213,"elapsed_time":51.81,"gpr":0.3333,"tpe":0.5905,"ee":0.3333},{"task_id":"task_152_enterprise_industry_analysis_easy_easy082","category":"enterprise_industry_analysis","score":1.0,"tokens":181101,"cost_usd":0.019981,"elapsed_time":72.49,"gpr":1.0,"tpe":0.35,"ee":0.2143},{"task_id":"task_153_enterprise_industry_analysis_easy_easy083","category":"enterprise_industry_analysis","score":1.0,"tokens":472283,"cost_usd":0.048066,"elapsed_time":139.87,"gpr":1.0,"tpe":0.1151,"ee":0.1154},{"task_id":"task_154_enterprise_industry_analysis_easy_easy084","category":"enterprise_industry_analysis","score":1.0,"tokens":127899,"cost_usd":0.018571,"elapsed_time":59.89,"gpr":1.0,"tpe":0.5511,"ee":0.3333},{"task_id":"task_155_enterprise_industry_analysis_easy_easy085","category":"enterprise_industry_analysis","score":1.0,"tokens":161691,"cost_usd":0.035528,"elapsed_time":51.85,"gpr":1.0,"tpe":0.6631,"ee":0.3333},{"task_id":"task_156_enterprise_industry_analysis_easy_easy086","category":"enterprise_industry_analysis","score":1.0,"tokens":348247,"cost_usd":0.046501,"elapsed_time":111.67,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_157_enterprise_industry_analysis_easy_easy088","category":"enterprise_industry_analysis","score":1.0,"tokens":281489,"cost_usd":0.035889,"elapsed_time":94.63,"gpr":1.0,"tpe":0.2288,"ee":0.125},{"task_id":"task_158_enterprise_industry_analysis_easy_easy090","category":"enterprise_industry_analysis","score":0.0,"tokens":410851,"cost_usd":0.049114,"elapsed_time":133.53,"gpr":0.6667,"tpe":0.4329,"ee":0.1579},{"task_id":"task_159_enterprise_industry_analysis_easy_easy091","category":"enterprise_industry_analysis","score":0.0,"tokens":148819,"cost_usd":0.015479,"elapsed_time":61.4,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_160_enterprise_industry_analysis_easy_easy092","category":"enterprise_industry_analysis","score":0.0,"tokens":511952,"cost_usd":0.063532,"elapsed_time":104.09,"gpr":0.0,"tpe":0.0,"ee":0.1765},{"task_id":"task_161_enterprise_industry_analysis_easy_easy093","category":"enterprise_industry_analysis","score":0.0,"tokens":180153,"cost_usd":0.019764,"elapsed_time":67.43,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_162_enterprise_industry_analysis_easy_easy094","category":"enterprise_industry_analysis","score":1.0,"tokens":115462,"cost_usd":0.016412,"elapsed_time":57.85,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_163_enterprise_industry_analysis_easy_easy095","category":"enterprise_industry_analysis","score":1.0,"tokens":68677,"cost_usd":0.008162,"elapsed_time":32.04,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_164_enterprise_industry_analysis_easy_easy096","category":"enterprise_industry_analysis","score":1.0,"tokens":194216,"cost_usd":0.024872,"elapsed_time":93.99,"gpr":1.0,"tpe":0.3138,"ee":0.2143},{"task_id":"task_165_enterprise_industry_analysis_easy_easy097","category":"enterprise_industry_analysis","score":1.0,"tokens":205294,"cost_usd":0.021524,"elapsed_time":78.37,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_166_enterprise_industry_analysis_easy_easy098","category":"enterprise_industry_analysis","score":0.0,"tokens":54786,"cost_usd":0.011496,"elapsed_time":29.47,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_167_enterprise_industry_analysis_easy_easy099","category":"enterprise_industry_analysis","score":0.0,"tokens":147939,"cost_usd":0.015883,"elapsed_time":57.95,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_168_enterprise_industry_analysis_easy_easy100","category":"enterprise_industry_analysis","score":1.0,"tokens":115784,"cost_usd":0.025133,"elapsed_time":54.64,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_169_enterprise_industry_analysis_easy_easy101","category":"enterprise_industry_analysis","score":1.0,"tokens":56227,"cost_usd":0.010282,"elapsed_time":37.35,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_170_enterprise_industry_analysis_easy_easy102","category":"enterprise_industry_analysis","score":1.0,"tokens":154165,"cost_usd":0.017527,"elapsed_time":57.38,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_171_enterprise_industry_analysis_easy_easy103","category":"enterprise_industry_analysis","score":1.0,"tokens":116160,"cost_usd":0.013124,"elapsed_time":47.0,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_172_enterprise_industry_analysis_easy_easy104","category":"enterprise_industry_analysis","score":1.0,"tokens":66967,"cost_usd":0.01173,"elapsed_time":34.43,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_173_enterprise_industry_analysis_easy_easy105","category":"enterprise_industry_analysis","score":1.0,"tokens":72497,"cost_usd":0.012311,"elapsed_time":36.87,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_174_enterprise_industry_analysis_easy_easy106","category":"enterprise_industry_analysis","score":1.0,"tokens":82472,"cost_usd":0.01279,"elapsed_time":46.46,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_175_enterprise_industry_analysis_easy_easy107","category":"enterprise_industry_analysis","score":1.0,"tokens":58389,"cost_usd":0.010981,"elapsed_time":36.08,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_176_enterprise_industry_analysis_easy_easy108","category":"enterprise_industry_analysis","score":0.0,"tokens":474051,"cost_usd":0.074319,"elapsed_time":92.52,"gpr":0.0,"tpe":0.0,"ee":0.1875},{"task_id":"task_177_enterprise_industry_analysis_easy_easy109","category":"enterprise_industry_analysis","score":0.0,"tokens":32331,"cost_usd":0.007358,"elapsed_time":23.05,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_178_enterprise_industry_analysis_easy_easy110","category":"enterprise_industry_analysis","score":0.0,"tokens":43951,"cost_usd":0.007335,"elapsed_time":27.36,"gpr":0.0,"tpe":0.0,"ee":0.75},{"task_id":"task_179_enterprise_industry_analysis_easy_easy111","category":"enterprise_industry_analysis","score":1.0,"tokens":45701,"cost_usd":0.009248,"elapsed_time":29.51,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_180_enterprise_industry_analysis_easy_easy112","category":"enterprise_industry_analysis","score":0.0,"tokens":220881,"cost_usd":0.040633,"elapsed_time":51.86,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_181_enterprise_industry_analysis_easy_easy113","category":"enterprise_industry_analysis","score":0.0,"tokens":82637,"cost_usd":0.012686,"elapsed_time":42.97,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_182_enterprise_industry_analysis_easy_easy114","category":"enterprise_industry_analysis","score":0.0,"tokens":240425,"cost_usd":0.046844,"elapsed_time":96.05,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_183_enterprise_industry_analysis_easy_easy115","category":"enterprise_industry_analysis","score":1.0,"tokens":165853,"cost_usd":0.043981,"elapsed_time":45.55,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_184_enterprise_industry_analysis_easy_easy116","category":"enterprise_industry_analysis","score":0.0,"tokens":195074,"cost_usd":0.048216,"elapsed_time":72.49,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_185_enterprise_industry_analysis_easy_easy117","category":"enterprise_industry_analysis","score":1.0,"tokens":143623,"cost_usd":0.020157,"elapsed_time":60.22,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_186_enterprise_industry_analysis_medium_medium001","category":"enterprise_industry_analysis","score":1.0,"tokens":450101,"cost_usd":0.046468,"elapsed_time":151.28,"gpr":1.0,"tpe":0.0985,"ee":0.1538},{"task_id":"task_187_enterprise_industry_analysis_medium_medium002","category":"enterprise_industry_analysis","score":0.0,"tokens":1125521,"cost_usd":0.095241,"elapsed_time":259.35,"gpr":0.25,"tpe":0.0278,"ee":0.0702},{"task_id":"task_188_enterprise_industry_analysis_medium_medium003","category":"enterprise_industry_analysis","score":0.0,"tokens":1308671,"cost_usd":0.135346,"elapsed_time":440.45,"gpr":0.25,"tpe":0.0985,"ee":0.069},{"task_id":"task_189_enterprise_industry_analysis_medium_medium004","category":"enterprise_industry_analysis","score":1.0,"tokens":575056,"cost_usd":0.058893,"elapsed_time":179.58,"gpr":1.0,"tpe":0.0257,"ee":0.1026},{"task_id":"task_190_enterprise_industry_analysis_medium_medium005","category":"enterprise_industry_analysis","score":1.0,"tokens":447790,"cost_usd":0.045535,"elapsed_time":145.81,"gpr":1.0,"tpe":0.0985,"ee":0.1538},{"task_id":"task_191_enterprise_industry_analysis_medium_medium006","category":"enterprise_industry_analysis","score":1.0,"tokens":582038,"cost_usd":0.051531,"elapsed_time":185.25,"gpr":1.0,"tpe":0.0309,"ee":0.1081},{"task_id":"task_192_enterprise_industry_analysis_medium_medium007","category":"enterprise_industry_analysis","score":0.0,"tokens":1556549,"cost_usd":0.128797,"elapsed_time":364.81,"gpr":0.0,"tpe":0.0,"ee":0.0635},{"task_id":"task_193_enterprise_industry_analysis_medium_medium008","category":"enterprise_industry_analysis","score":1.0,"tokens":973500,"cost_usd":0.091262,"elapsed_time":165.06,"gpr":1.0,"tpe":0.0886,"ee":0.1481},{"task_id":"task_194_enterprise_industry_analysis_medium_medium009","category":"enterprise_industry_analysis","score":0.0,"tokens":759405,"cost_usd":0.088114,"elapsed_time":165.07,"gpr":0.0,"tpe":0.0,"ee":0.125},{"task_id":"task_195_enterprise_industry_analysis_medium_medium010","category":"enterprise_industry_analysis","score":0.0,"tokens":217985,"cost_usd":0.030873,"elapsed_time":96.08,"gpr":0.4,"tpe":0.3138,"ee":0.2667},{"task_id":"task_196_enterprise_industry_analysis_medium_medium011","category":"enterprise_industry_analysis","score":0.0,"tokens":411882,"cost_usd":0.055886,"elapsed_time":167.93,"gpr":0.0,"tpe":0.0,"ee":0.1739},{"task_id":"task_197_enterprise_industry_analysis_medium_medium012","category":"enterprise_industry_analysis","score":1.0,"tokens":379631,"cost_usd":0.045444,"elapsed_time":95.84,"gpr":1.0,"tpe":0.1853,"ee":0.2},{"task_id":"task_198_enterprise_industry_analysis_medium_medium013","category":"enterprise_industry_analysis","score":1.0,"tokens":677789,"cost_usd":0.06802,"elapsed_time":190.82,"gpr":1.0,"tpe":0.0424,"ee":0.1176},{"task_id":"task_199_enterprise_industry_analysis_medium_medium014","category":"enterprise_industry_analysis","score":0.0,"tokens":1052344,"cost_usd":0.10305,"elapsed_time":314.04,"gpr":0.6,"tpe":0.5057,"ee":0.1429},{"task_id":"task_200_enterprise_industry_analysis_medium_medium015","category":"enterprise_industry_analysis","score":1.0,"tokens":282196,"cost_usd":0.034157,"elapsed_time":111.31,"gpr":1.0,"tpe":0.1853,"ee":0.2381},{"task_id":"task_201_enterprise_industry_analysis_medium_medium016","category":"enterprise_industry_analysis","score":1.0,"tokens":557877,"cost_usd":0.065919,"elapsed_time":176.97,"gpr":1.0,"tpe":0.0857,"ee":0.1667},{"task_id":"task_202_enterprise_industry_analysis_medium_medium017","category":"enterprise_industry_analysis","score":0.0,"tokens":271989,"cost_usd":0.034243,"elapsed_time":129.45,"gpr":0.4,"tpe":0.3138,"ee":0.3125},{"task_id":"task_203_enterprise_industry_analysis_medium_medium018","category":"enterprise_industry_analysis","score":0.0,"tokens":534274,"cost_usd":0.059949,"elapsed_time":250.03,"gpr":0.4,"tpe":0.4725,"ee":0.1613},{"task_id":"task_204_enterprise_industry_analysis_medium_medium019","category":"enterprise_industry_analysis","score":0.0,"tokens":470617,"cost_usd":0.057052,"elapsed_time":245.47,"gpr":0.8,"tpe":0.4449,"ee":0.1852},{"task_id":"task_205_enterprise_industry_analysis_medium_medium020","category":"enterprise_industry_analysis","score":0.0,"tokens":531547,"cost_usd":0.063742,"elapsed_time":279.08,"gpr":0.8,"tpe":0.6669,"ee":0.1667},{"task_id":"task_206_enterprise_industry_analysis_medium_medium021","category":"enterprise_industry_analysis","score":0.0,"tokens":461098,"cost_usd":0.058915,"elapsed_time":218.97,"gpr":0.0,"tpe":0.0,"ee":0.2083},{"task_id":"task_207_enterprise_industry_analysis_medium_medium022","category":"enterprise_industry_analysis","score":0.0,"tokens":626116,"cost_usd":0.071433,"elapsed_time":274.63,"gpr":0.4,"tpe":0.11,"ee":0.1613},{"task_id":"task_208_enterprise_industry_analysis_medium_medium023","category":"enterprise_industry_analysis","score":0.0,"tokens":243102,"cost_usd":0.030029,"elapsed_time":104.81,"gpr":0.8,"tpe":0.5729,"ee":0.2941},{"task_id":"task_209_enterprise_industry_analysis_medium_medium024","category":"enterprise_industry_analysis","score":1.0,"tokens":1207122,"cost_usd":0.120433,"elapsed_time":276.11,"gpr":1.0,"tpe":0.4831,"ee":0.1562},{"task_id":"task_210_enterprise_industry_analysis_medium_medium025","category":"enterprise_industry_analysis","score":1.0,"tokens":446575,"cost_usd":0.056511,"elapsed_time":123.55,"gpr":1.0,"tpe":0.2999,"ee":0.2083},{"task_id":"task_211_enterprise_industry_analysis_medium_medium026","category":"enterprise_industry_analysis","score":0.0,"tokens":128246,"cost_usd":0.015557,"elapsed_time":55.55,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_212_enterprise_industry_analysis_medium_medium027","category":"enterprise_industry_analysis","score":1.0,"tokens":346537,"cost_usd":0.046543,"elapsed_time":125.84,"gpr":1.0,"tpe":0.3842,"ee":0.2381},{"task_id":"task_213_enterprise_industry_analysis_medium_medium028","category":"enterprise_industry_analysis","score":0.0,"tokens":1865977,"cost_usd":0.175011,"elapsed_time":240.78,"gpr":0.8,"tpe":0.2665,"ee":0.2083},{"task_id":"task_214_enterprise_industry_analysis_medium_medium029","category":"enterprise_industry_analysis","score":0.0,"tokens":398949,"cost_usd":0.049299,"elapsed_time":199.63,"gpr":0.8,"tpe":0.5779,"ee":0.2273},{"task_id":"task_215_enterprise_industry_analysis_medium_medium030","category":"enterprise_industry_analysis","score":0.0,"tokens":1142400,"cost_usd":0.132245,"elapsed_time":589.59,"gpr":1.0,"tpe":0.0052,"ee":0.0893},{"task_id":"task_216_enterprise_industry_analysis_medium_medium031","category":"enterprise_industry_analysis","score":0.0,"tokens":210045,"cost_usd":0.02969,"elapsed_time":113.21,"gpr":0.8,"tpe":0.6341,"ee":0.3125},{"task_id":"task_217_enterprise_industry_analysis_medium_medium032","category":"enterprise_industry_analysis","score":1.0,"tokens":1067787,"cost_usd":0.130918,"elapsed_time":257.58,"gpr":1.0,"tpe":0.1753,"ee":0.1471},{"task_id":"task_218_enterprise_industry_analysis_medium_medium033","category":"enterprise_industry_analysis","score":0.0,"tokens":257783,"cost_usd":0.034744,"elapsed_time":146.43,"gpr":0.4,"tpe":0.5105,"ee":0.2778},{"task_id":"task_219_enterprise_industry_analysis_medium_medium034","category":"enterprise_industry_analysis","score":1.0,"tokens":226021,"cost_usd":0.040984,"elapsed_time":90.83,"gpr":1.0,"tpe":0.3874,"ee":0.3571},{"task_id":"task_220_enterprise_industry_analysis_medium_medium035","category":"enterprise_industry_analysis","score":0.0,"tokens":690258,"cost_usd":0.076251,"elapsed_time":274.31,"gpr":0.6,"tpe":0.103,"ee":0.1351},{"task_id":"task_221_enterprise_industry_analysis_medium_medium036","category":"enterprise_industry_analysis","score":0.0,"tokens":525214,"cost_usd":0.051667,"elapsed_time":177.09,"gpr":0.6,"tpe":0.4387,"ee":0.1923},{"task_id":"task_222_enterprise_industry_analysis_medium_medium037","category":"enterprise_industry_analysis","score":1.0,"tokens":282998,"cost_usd":0.041228,"elapsed_time":170.83,"gpr":1.0,"tpe":0.2542,"ee":0.2778},{"task_id":"task_223_enterprise_industry_analysis_medium_medium038","category":"enterprise_industry_analysis","score":1.0,"tokens":564111,"cost_usd":0.060918,"elapsed_time":194.5,"gpr":1.0,"tpe":0.1026,"ee":0.1724},{"task_id":"task_224_enterprise_industry_analysis_medium_medium039","category":"enterprise_industry_analysis","score":1.0,"tokens":156506,"cost_usd":0.022966,"elapsed_time":70.18,"gpr":1.0,"tpe":0.7274,"ee":0.5},{"task_id":"task_225_enterprise_industry_analysis_medium_medium040","category":"enterprise_industry_analysis","score":0.0,"tokens":601042,"cost_usd":0.064359,"elapsed_time":228.01,"gpr":0.4,"tpe":0.1216,"ee":0.1786},{"task_id":"task_226_enterprise_industry_analysis_medium_medium041","category":"enterprise_industry_analysis","score":0.0,"tokens":1535894,"cost_usd":0.161224,"elapsed_time":315.51,"gpr":0.8,"tpe":0.1855,"ee":0.1282},{"task_id":"task_227_enterprise_industry_analysis_medium_medium042","category":"enterprise_industry_analysis","score":0.0,"tokens":6447862,"cost_usd":0.53991,"elapsed_time":1207.47,"gpr":0.8,"tpe":0.4358,"ee":0.0427},{"task_id":"task_228_enterprise_industry_analysis_medium_medium043","category":"enterprise_industry_analysis","score":0.0,"tokens":803051,"cost_usd":0.080695,"elapsed_time":263.11,"gpr":0.6,"tpe":0.6882,"ee":0.1429},{"task_id":"task_229_enterprise_industry_analysis_medium_medium044","category":"enterprise_industry_analysis","score":1.0,"tokens":633574,"cost_usd":0.074875,"elapsed_time":312.28,"gpr":0.8,"tpe":0.3099,"ee":0.1389},{"task_id":"task_230_enterprise_industry_analysis_medium_medium045","category":"enterprise_industry_analysis","score":0.0,"tokens":345783,"cost_usd":0.047917,"elapsed_time":220.3,"gpr":0.6,"tpe":0.2398,"ee":0.2381},{"task_id":"task_231_enterprise_industry_analysis_medium_medium046","category":"enterprise_industry_analysis","score":1.0,"tokens":393531,"cost_usd":0.048014,"elapsed_time":166.11,"gpr":1.0,"tpe":0.2288,"ee":0.2632},{"task_id":"task_232_enterprise_industry_analysis_medium_medium047","category":"enterprise_industry_analysis","score":0.0,"tokens":464925,"cost_usd":0.046285,"elapsed_time":178.45,"gpr":0.4,"tpe":0.1157,"ee":0.1562},{"task_id":"task_233_enterprise_industry_analysis_medium_medium048","category":"enterprise_industry_analysis","score":0.0,"tokens":592305,"cost_usd":0.070568,"elapsed_time":218.89,"gpr":0.2,"tpe":0.2288,"ee":0.1852},{"task_id":"task_234_enterprise_industry_analysis_medium_medium049","category":"enterprise_industry_analysis","score":1.0,"tokens":996965,"cost_usd":0.100873,"elapsed_time":397.64,"gpr":1.0,"tpe":0.1321,"ee":0.102},{"task_id":"task_235_enterprise_industry_analysis_medium_medium050","category":"enterprise_industry_analysis","score":0.0,"tokens":289006,"cost_usd":0.034796,"elapsed_time":158.78,"gpr":0.0,"tpe":0.0,"ee":0.2778},{"task_id":"task_236_enterprise_industry_analysis_medium_medium051","category":"enterprise_industry_analysis","score":0.0,"tokens":1337981,"cost_usd":0.127802,"elapsed_time":366.71,"gpr":0.8,"tpe":0.4244,"ee":0.1282},{"task_id":"task_237_enterprise_industry_analysis_medium_medium052","category":"enterprise_industry_analysis","score":1.0,"tokens":219605,"cost_usd":0.02821,"elapsed_time":106.72,"gpr":1.0,"tpe":0.4721,"ee":0.3125},{"task_id":"task_238_enterprise_industry_analysis_medium_medium053","category":"enterprise_industry_analysis","score":0.0,"tokens":377302,"cost_usd":0.051683,"elapsed_time":142.43,"gpr":0.6,"tpe":0.5028,"ee":0.2273},{"task_id":"task_239_enterprise_industry_analysis_medium_medium054","category":"enterprise_industry_analysis","score":1.0,"tokens":257399,"cost_usd":0.036441,"elapsed_time":97.13,"gpr":1.0,"tpe":0.5659,"ee":0.3333},{"task_id":"task_240_enterprise_industry_analysis_medium_medium055","category":"enterprise_industry_analysis","score":1.0,"tokens":666601,"cost_usd":0.085727,"elapsed_time":178.7,"gpr":1.0,"tpe":0.3605,"ee":0.3125},{"task_id":"task_241_enterprise_industry_analysis_medium_medium056","category":"enterprise_industry_analysis","score":1.0,"tokens":320975,"cost_usd":0.039567,"elapsed_time":184.32,"gpr":1.0,"tpe":0.1668,"ee":0.2273},{"task_id":"task_242_enterprise_industry_analysis_medium_medium057","category":"enterprise_industry_analysis","score":1.0,"tokens":333269,"cost_usd":0.034192,"elapsed_time":126.65,"gpr":1.0,"tpe":0.3503,"ee":0.2083},{"task_id":"task_243_enterprise_industry_analysis_medium_medium058","category":"enterprise_industry_analysis","score":0.0,"tokens":388307,"cost_usd":0.049702,"elapsed_time":179.7,"gpr":1.0,"tpe":0.3455,"ee":0.25},{"task_id":"task_244_enterprise_industry_analysis_medium_medium059","category":"enterprise_industry_analysis","score":1.0,"tokens":289801,"cost_usd":0.034376,"elapsed_time":146.97,"gpr":1.0,"tpe":0.3498,"ee":0.2632},{"task_id":"task_245_enterprise_industry_analysis_medium_medium060","category":"enterprise_industry_analysis","score":0.0,"tokens":1095985,"cost_usd":0.162413,"elapsed_time":145.0,"gpr":0.0,"tpe":0.0,"ee":0.2941},{"task_id":"task_246_enterprise_industry_analysis_medium_medium061","category":"enterprise_industry_analysis","score":1.0,"tokens":820992,"cost_usd":0.140174,"elapsed_time":114.46,"gpr":1.0,"tpe":0.6721,"ee":0.5},{"task_id":"task_247_enterprise_industry_analysis_medium_medium062","category":"enterprise_industry_analysis","score":1.0,"tokens":281513,"cost_usd":0.032804,"elapsed_time":122.67,"gpr":1.0,"tpe":0.2823,"ee":0.25},{"task_id":"task_248_enterprise_industry_analysis_medium_medium063","category":"enterprise_industry_analysis","score":0.0,"tokens":147124,"cost_usd":0.019637,"elapsed_time":76.13,"gpr":0.5,"tpe":0.729,"ee":0.3333},{"task_id":"task_249_enterprise_industry_analysis_medium_medium064","category":"enterprise_industry_analysis","score":0.0,"tokens":357028,"cost_usd":0.046285,"elapsed_time":145.03,"gpr":0.75,"tpe":0.3734,"ee":0.1905},{"task_id":"task_250_enterprise_industry_analysis_medium_medium065","category":"enterprise_industry_analysis","score":1.0,"tokens":119331,"cost_usd":0.016663,"elapsed_time":54.27,"gpr":1.0,"tpe":0.8151,"ee":0.4},{"task_id":"task_251_enterprise_industry_analysis_medium_medium066","category":"enterprise_industry_analysis","score":1.0,"tokens":164122,"cost_usd":0.020065,"elapsed_time":74.66,"gpr":1.0,"tpe":0.3874,"ee":0.3077},{"task_id":"task_252_enterprise_industry_analysis_medium_medium067","category":"enterprise_industry_analysis","score":1.0,"tokens":190581,"cost_usd":0.026025,"elapsed_time":79.24,"gpr":1.0,"tpe":0.6365,"ee":0.2857},{"task_id":"task_253_enterprise_industry_analysis_medium_medium068","category":"enterprise_industry_analysis","score":0.0,"tokens":436206,"cost_usd":0.054148,"elapsed_time":194.43,"gpr":0.75,"tpe":0.4018,"ee":0.1667},{"task_id":"task_254_enterprise_industry_analysis_medium_medium069","category":"enterprise_industry_analysis","score":1.0,"tokens":362691,"cost_usd":0.039155,"elapsed_time":137.91,"gpr":1.0,"tpe":0.2758,"ee":0.2},{"task_id":"task_255_enterprise_industry_analysis_medium_medium070","category":"enterprise_industry_analysis","score":0.0,"tokens":140718,"cost_usd":0.019977,"elapsed_time":95.85,"gpr":0.3333,"tpe":0.6561,"ee":0.4},{"task_id":"task_256_enterprise_industry_analysis_medium_medium071","category":"enterprise_industry_analysis","score":0.0,"tokens":372480,"cost_usd":0.041093,"elapsed_time":114.67,"gpr":0.5,"tpe":0.1863,"ee":0.1667},{"task_id":"task_257_enterprise_industry_analysis_medium_medium072","category":"enterprise_industry_analysis","score":1.0,"tokens":387295,"cost_usd":0.044717,"elapsed_time":126.9,"gpr":1.0,"tpe":0.1205,"ee":0.1481},{"task_id":"task_258_enterprise_industry_analysis_medium_medium073","category":"enterprise_industry_analysis","score":1.0,"tokens":634254,"cost_usd":0.067299,"elapsed_time":261.69,"gpr":1.0,"tpe":0.1243,"ee":0.1111},{"task_id":"task_259_enterprise_industry_analysis_medium_medium074","category":"enterprise_industry_analysis","score":0.0,"tokens":1331681,"cost_usd":0.134128,"elapsed_time":232.49,"gpr":0.75,"tpe":0.4125,"ee":0.1143},{"task_id":"task_260_enterprise_industry_analysis_medium_medium075","category":"enterprise_industry_analysis","score":1.0,"tokens":434253,"cost_usd":0.046374,"elapsed_time":162.64,"gpr":1.0,"tpe":0.3288,"ee":0.1667},{"task_id":"task_261_enterprise_industry_analysis_medium_medium076","category":"enterprise_industry_analysis","score":0.0,"tokens":971750,"cost_usd":0.099775,"elapsed_time":216.69,"gpr":0.5,"tpe":0.0552,"ee":0.1053},{"task_id":"task_262_enterprise_industry_analysis_medium_medium077","category":"enterprise_industry_analysis","score":1.0,"tokens":635633,"cost_usd":0.087781,"elapsed_time":183.51,"gpr":1.0,"tpe":0.1689,"ee":0.1667},{"task_id":"task_263_enterprise_industry_analysis_medium_medium078","category":"enterprise_industry_analysis","score":0.0,"tokens":439090,"cost_usd":0.045148,"elapsed_time":179.93,"gpr":0.75,"tpe":0.2108,"ee":0.1379},{"task_id":"task_264_enterprise_industry_analysis_medium_medium079","category":"enterprise_industry_analysis","score":0.0,"tokens":268287,"cost_usd":0.031187,"elapsed_time":97.24,"gpr":0.5,"tpe":0.4226,"ee":0.2222},{"task_id":"task_265_enterprise_industry_analysis_medium_medium080","category":"enterprise_industry_analysis","score":0.0,"tokens":287028,"cost_usd":0.040827,"elapsed_time":114.02,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_266_enterprise_industry_analysis_medium_medium081","category":"enterprise_industry_analysis","score":1.0,"tokens":284147,"cost_usd":0.033287,"elapsed_time":113.42,"gpr":1.0,"tpe":0.2565,"ee":0.1818},{"task_id":"task_267_enterprise_industry_analysis_medium_medium082","category":"enterprise_industry_analysis","score":0.0,"tokens":213740,"cost_usd":0.025574,"elapsed_time":100.32,"gpr":1.0,"tpe":0.2824,"ee":0.25},{"task_id":"task_268_enterprise_industry_analysis_medium_medium083","category":"enterprise_industry_analysis","score":0.0,"tokens":451318,"cost_usd":0.047802,"elapsed_time":131.29,"gpr":0.5,"tpe":0.2545,"ee":0.1538},{"task_id":"task_269_enterprise_industry_analysis_medium_medium084","category":"enterprise_industry_analysis","score":1.0,"tokens":376524,"cost_usd":0.054704,"elapsed_time":110.42,"gpr":1.0,"tpe":0.4133,"ee":0.2353},{"task_id":"task_270_enterprise_industry_analysis_medium_medium085","category":"enterprise_industry_analysis","score":0.0,"tokens":863290,"cost_usd":0.135897,"elapsed_time":173.47,"gpr":0.75,"tpe":0.81,"ee":0.1481},{"task_id":"task_271_enterprise_industry_analysis_medium_medium086","category":"enterprise_industry_analysis","score":1.0,"tokens":526856,"cost_usd":0.069108,"elapsed_time":418.87,"gpr":1.0,"tpe":0.4474,"ee":0.1379},{"task_id":"task_272_enterprise_industry_analysis_medium_medium087","category":"enterprise_industry_analysis","score":0.0,"tokens":178046,"cost_usd":0.022629,"elapsed_time":91.86,"gpr":0.75,"tpe":1.0,"ee":0.2857},{"task_id":"task_273_enterprise_industry_analysis_medium_medium088","category":"enterprise_industry_analysis","score":1.0,"tokens":1264726,"cost_usd":0.120612,"elapsed_time":147.61,"gpr":1.0,"tpe":0.1823,"ee":0.1905},{"task_id":"task_274_enterprise_industry_analysis_medium_medium089","category":"enterprise_industry_analysis","score":1.0,"tokens":2466405,"cost_usd":0.21926,"elapsed_time":527.16,"gpr":1.0,"tpe":0.1532,"ee":0.0727},{"task_id":"task_275_enterprise_industry_analysis_medium_medium090","category":"enterprise_industry_analysis","score":1.0,"tokens":974367,"cost_usd":0.169852,"elapsed_time":144.08,"gpr":1.0,"tpe":0.4101,"ee":0.3077},{"task_id":"task_276_enterprise_industry_analysis_medium_medium091","category":"enterprise_industry_analysis","score":1.0,"tokens":912085,"cost_usd":0.097045,"elapsed_time":116.37,"gpr":1.0,"tpe":0.4671,"ee":0.2105},{"task_id":"task_277_enterprise_industry_analysis_medium_medium092","category":"enterprise_industry_analysis","score":1.0,"tokens":446857,"cost_usd":0.055219,"elapsed_time":126.42,"gpr":1.0,"tpe":0.2415,"ee":0.2222},{"task_id":"task_278_enterprise_industry_analysis_medium_medium093","category":"enterprise_industry_analysis","score":1.0,"tokens":704943,"cost_usd":0.063794,"elapsed_time":199.08,"gpr":1.0,"tpe":0.0298,"ee":0.0909},{"task_id":"task_279_enterprise_industry_analysis_medium_medium094","category":"enterprise_industry_analysis","score":1.0,"tokens":2473387,"cost_usd":0.245495,"elapsed_time":221.02,"gpr":1.0,"tpe":0.2236,"ee":0.1538},{"task_id":"task_280_enterprise_industry_analysis_medium_medium095","category":"enterprise_industry_analysis","score":0.0,"tokens":788084,"cost_usd":0.087057,"elapsed_time":365.56,"gpr":0.75,"tpe":0.2409,"ee":0.1111},{"task_id":"task_281_enterprise_industry_analysis_medium_medium096","category":"enterprise_industry_analysis","score":1.0,"tokens":346094,"cost_usd":0.043556,"elapsed_time":140.28,"gpr":1.0,"tpe":0.2601,"ee":0.2},{"task_id":"task_282_enterprise_industry_analysis_medium_medium097","category":"enterprise_industry_analysis","score":1.0,"tokens":316410,"cost_usd":0.049362,"elapsed_time":92.51,"gpr":1.0,"tpe":0.3777,"ee":0.2857},{"task_id":"task_283_enterprise_industry_analysis_medium_medium098","category":"enterprise_industry_analysis","score":0.0,"tokens":1308792,"cost_usd":0.207001,"elapsed_time":255.26,"gpr":1.0,"tpe":0.1047,"ee":0.1},{"task_id":"task_284_enterprise_industry_analysis_medium_medium099","category":"enterprise_industry_analysis","score":0.0,"tokens":318671,"cost_usd":0.041321,"elapsed_time":133.72,"gpr":0.25,"tpe":0.2824,"ee":0.2},{"task_id":"task_285_enterprise_industry_analysis_medium_medium100","category":"enterprise_industry_analysis","score":0.0,"tokens":265437,"cost_usd":0.032701,"elapsed_time":98.97,"gpr":0.75,"tpe":0.2824,"ee":0.2105},{"task_id":"task_286_enterprise_industry_analysis_medium_medium101","category":"enterprise_industry_analysis","score":0.0,"tokens":439313,"cost_usd":0.043701,"elapsed_time":169.28,"gpr":0.5,"tpe":0.3982,"ee":0.1481},{"task_id":"task_287_enterprise_industry_analysis_medium_medium102","category":"enterprise_industry_analysis","score":0.0,"tokens":659090,"cost_usd":0.111364,"elapsed_time":116.17,"gpr":0.25,"tpe":0.5905,"ee":0.3333},{"task_id":"task_288_enterprise_industry_analysis_medium_medium103","category":"enterprise_industry_analysis","score":0.0,"tokens":189754,"cost_usd":0.022098,"elapsed_time":76.24,"gpr":0.25,"tpe":0.3487,"ee":0.2667},{"task_id":"task_289_enterprise_industry_analysis_medium_medium104","category":"enterprise_industry_analysis","score":0.0,"tokens":452978,"cost_usd":0.045513,"elapsed_time":180.52,"gpr":0.0,"tpe":0.0,"ee":0.125},{"task_id":"task_290_enterprise_industry_analysis_medium_medium105","category":"enterprise_industry_analysis","score":0.0,"tokens":2429983,"cost_usd":0.227091,"elapsed_time":290.12,"gpr":0.75,"tpe":0.0664,"ee":0.1143},{"task_id":"task_291_enterprise_industry_analysis_medium_medium106","category":"enterprise_industry_analysis","score":0.0,"tokens":376626,"cost_usd":0.064093,"elapsed_time":131.07,"gpr":0.25,"tpe":0.4783,"ee":0.3077},{"task_id":"task_292_enterprise_industry_analysis_medium_medium107","category":"enterprise_industry_analysis","score":0.0,"tokens":781842,"cost_usd":0.075639,"elapsed_time":182.34,"gpr":0.25,"tpe":0.0646,"ee":0.1111},{"task_id":"task_293_enterprise_industry_analysis_medium_medium108","category":"enterprise_industry_analysis","score":0.0,"tokens":377935,"cost_usd":0.038927,"elapsed_time":113.71,"gpr":0.25,"tpe":0.4783,"ee":0.16},{"task_id":"task_294_enterprise_industry_analysis_medium_medium109","category":"enterprise_industry_analysis","score":1.0,"tokens":395729,"cost_usd":0.054391,"elapsed_time":231.41,"gpr":1.0,"tpe":0.0911,"ee":0.1481},{"task_id":"task_295_enterprise_industry_analysis_medium_medium110","category":"enterprise_industry_analysis","score":0.0,"tokens":3637770,"cost_usd":0.326061,"elapsed_time":333.24,"gpr":0.25,"tpe":0.729,"ee":0.1081},{"task_id":"task_296_enterprise_industry_analysis_medium_medium111","category":"enterprise_industry_analysis","score":0.0,"tokens":675080,"cost_usd":0.15637,"elapsed_time":93.59,"gpr":0.25,"tpe":0.5905,"ee":0.4},{"task_id":"task_297_enterprise_industry_policy_analysis_easy_easy067","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":86367,"cost_usd":0.012501,"elapsed_time":41.03,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_298_enterprise_industry_policy_analysis_easy_easy068","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":61206,"cost_usd":0.01009,"elapsed_time":38.77,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_299_enterprise_industry_policy_analysis_easy_easy069","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":344867,"cost_usd":0.039095,"elapsed_time":174.08,"gpr":0.0,"tpe":0.0,"ee":0.1304},{"task_id":"task_300_enterprise_industry_policy_analysis_easy_easy070","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":140715,"cost_usd":0.026113,"elapsed_time":69.5,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_301_enterprise_industry_policy_analysis_easy_easy071","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":53861,"cost_usd":0.01209,"elapsed_time":35.57,"gpr":0.0,"tpe":0.0,"ee":0.75},{"task_id":"task_302_enterprise_industry_policy_analysis_easy_easy072","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":98030,"cost_usd":0.014422,"elapsed_time":45.49,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_303_enterprise_industry_policy_analysis_easy_easy073","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":49948,"cost_usd":0.008323,"elapsed_time":40.27,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_304_enterprise_industry_policy_analysis_easy_easy074","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":82552,"cost_usd":0.015508,"elapsed_time":45.03,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_305_enterprise_industry_policy_analysis_easy_easy075","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":124686,"cost_usd":0.036132,"elapsed_time":48.04,"gpr":0.0,"tpe":0.0,"ee":0.75},{"task_id":"task_306_enterprise_industry_policy_analysis_easy_easy076","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":127993,"cost_usd":0.018463,"elapsed_time":54.34,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_307_enterprise_industry_policy_analysis_medium_medium001","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2304467,"cost_usd":0.199616,"elapsed_time":452.69,"gpr":0.8,"tpe":0.1825,"ee":0.0847},{"task_id":"task_308_enterprise_industry_policy_analysis_medium_medium002","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":187650,"cost_usd":0.030846,"elapsed_time":99.38,"gpr":1.0,"tpe":0.5114,"ee":0.4167},{"task_id":"task_309_enterprise_industry_policy_analysis_medium_medium003","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":861756,"cost_usd":0.075183,"elapsed_time":218.84,"gpr":0.4,"tpe":0.4544,"ee":0.125},{"task_id":"task_310_enterprise_industry_policy_analysis_medium_medium004","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":332004,"cost_usd":0.042571,"elapsed_time":159.99,"gpr":0.0,"tpe":0.0,"ee":0.3125},{"task_id":"task_311_enterprise_industry_policy_analysis_medium_medium005","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1855279,"cost_usd":0.221909,"elapsed_time":286.31,"gpr":0.8,"tpe":0.4477,"ee":0.1724},{"task_id":"task_312_enterprise_industry_policy_analysis_medium_medium006","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":912963,"cost_usd":0.087322,"elapsed_time":261.7,"gpr":1.0,"tpe":0.1108,"ee":0.1282},{"task_id":"task_313_enterprise_industry_policy_analysis_medium_medium007","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":582159,"cost_usd":0.065791,"elapsed_time":231.67,"gpr":1.0,"tpe":0.2156,"ee":0.1724},{"task_id":"task_314_enterprise_industry_policy_analysis_medium_medium008","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":508976,"cost_usd":0.057179,"elapsed_time":169.05,"gpr":1.0,"tpe":0.1302,"ee":0.1852},{"task_id":"task_315_enterprise_industry_policy_analysis_medium_medium009","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":301995,"cost_usd":0.03494,"elapsed_time":108.16,"gpr":0.0,"tpe":0.0,"ee":0.2941},{"task_id":"task_316_enterprise_industry_policy_analysis_medium_medium010","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1652361,"cost_usd":0.152944,"elapsed_time":326.5,"gpr":0.0,"tpe":0.0,"ee":0.122},{"task_id":"task_317_enterprise_industry_policy_analysis_medium_medium011","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":2369491,"cost_usd":0.2216,"elapsed_time":291.15,"gpr":1.0,"tpe":0.4642,"ee":0.1852},{"task_id":"task_318_enterprise_industry_policy_analysis_medium_medium012","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":603745,"cost_usd":0.073196,"elapsed_time":200.9,"gpr":0.6,"tpe":0.4253,"ee":0.1852},{"task_id":"task_319_enterprise_industry_policy_analysis_medium_medium013","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":368709,"cost_usd":0.044091,"elapsed_time":156.41,"gpr":1.0,"tpe":0.7202,"ee":0.3125},{"task_id":"task_320_enterprise_industry_policy_analysis_medium_medium014","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":968021,"cost_usd":0.11369,"elapsed_time":284.8,"gpr":1.0,"tpe":0.3498,"ee":0.1786},{"task_id":"task_321_enterprise_industry_policy_analysis_medium_medium015","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":404905,"cost_usd":0.0461,"elapsed_time":230.86,"gpr":0.2,"tpe":0.1668,"ee":0.1852},{"task_id":"task_322_enterprise_industry_policy_analysis_medium_medium016","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":978158,"cost_usd":0.115377,"elapsed_time":306.35,"gpr":1.0,"tpe":0.091,"ee":0.1724},{"task_id":"task_323_enterprise_industry_policy_analysis_medium_medium017","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":672302,"cost_usd":0.079053,"elapsed_time":279.36,"gpr":1.0,"tpe":0.4576,"ee":0.1667},{"task_id":"task_324_enterprise_industry_policy_analysis_medium_medium018","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":244952,"cost_usd":0.036806,"elapsed_time":122.93,"gpr":0.8,"tpe":0.7952,"ee":0.3846},{"task_id":"task_325_enterprise_industry_policy_analysis_medium_medium019","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":628538,"cost_usd":0.068274,"elapsed_time":223.36,"gpr":1.0,"tpe":0.2779,"ee":0.1786},{"task_id":"task_326_enterprise_industry_policy_analysis_medium_medium020","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":751553,"cost_usd":0.085579,"elapsed_time":188.7,"gpr":1.0,"tpe":0.5503,"ee":0.2273},{"task_id":"task_327_enterprise_industry_policy_analysis_medium_medium021","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":900957,"cost_usd":0.13146,"elapsed_time":232.65,"gpr":0.4,"tpe":0.3956,"ee":0.2},{"task_id":"task_328_enterprise_industry_policy_analysis_medium_medium022","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1269630,"cost_usd":0.130388,"elapsed_time":170.66,"gpr":0.0,"tpe":0.0,"ee":0.2174},{"task_id":"task_329_enterprise_industry_policy_analysis_medium_medium023","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1865847,"cost_usd":0.174224,"elapsed_time":481.58,"gpr":1.0,"tpe":0.2472,"ee":0.0943},{"task_id":"task_330_enterprise_industry_policy_analysis_medium_medium024","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2243663,"cost_usd":0.236962,"elapsed_time":539.07,"gpr":0.4,"tpe":0.0403,"ee":0.0862},{"task_id":"task_331_enterprise_industry_policy_analysis_medium_medium025","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1347458,"cost_usd":0.156167,"elapsed_time":414.89,"gpr":1.0,"tpe":0.2276,"ee":0.1282},{"task_id":"task_332_enterprise_industry_policy_analysis_medium_medium026","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":745345,"cost_usd":0.085735,"elapsed_time":192.74,"gpr":1.0,"tpe":0.2889,"ee":0.25},{"task_id":"task_333_enterprise_industry_policy_analysis_medium_medium027","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":255703,"cost_usd":0.031734,"elapsed_time":111.76,"gpr":0.4,"tpe":1.0,"ee":0.3571},{"task_id":"task_334_enterprise_industry_policy_analysis_medium_medium028","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":683427,"cost_usd":0.077312,"elapsed_time":245.58,"gpr":0.4,"tpe":0.0738,"ee":0.1351},{"task_id":"task_335_enterprise_industry_policy_analysis_medium_medium029","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":479475,"cost_usd":0.06533,"elapsed_time":233.52,"gpr":1.0,"tpe":0.2126,"ee":0.2174},{"task_id":"task_336_enterprise_industry_policy_analysis_medium_medium030","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":3104995,"cost_usd":0.393614,"elapsed_time":415.25,"gpr":0.8,"tpe":0.307,"ee":0.1389},{"task_id":"task_337_enterprise_industry_policy_analysis_medium_medium031","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":409890,"cost_usd":0.051286,"elapsed_time":153.71,"gpr":1.0,"tpe":0.3212,"ee":0.2},{"task_id":"task_338_enterprise_industry_policy_analysis_medium_medium032","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1139815,"cost_usd":0.129026,"elapsed_time":144.3,"gpr":0.75,"tpe":0.783,"ee":0.3077},{"task_id":"task_339_enterprise_industry_policy_analysis_medium_medium033","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":450348,"cost_usd":0.055039,"elapsed_time":190.28,"gpr":1.0,"tpe":0.3414,"ee":0.1905},{"task_id":"task_340_enterprise_industry_policy_analysis_medium_medium034","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":446057,"cost_usd":0.046729,"elapsed_time":135.93,"gpr":0.5,"tpe":0.2173,"ee":0.1905},{"task_id":"task_341_enterprise_industry_policy_analysis_medium_medium035","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":535734,"cost_usd":0.071435,"elapsed_time":151.6,"gpr":1.0,"tpe":0.3662,"ee":0.2353},{"task_id":"task_342_enterprise_industry_policy_analysis_medium_medium036","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1301029,"cost_usd":0.120347,"elapsed_time":202.82,"gpr":0.75,"tpe":0.1383,"ee":0.125},{"task_id":"task_343_enterprise_industry_policy_analysis_medium_medium037","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":703760,"cost_usd":0.084243,"elapsed_time":213.88,"gpr":0.75,"tpe":0.783,"ee":0.2222},{"task_id":"task_344_enterprise_industry_policy_analysis_medium_medium038","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":502882,"cost_usd":0.052492,"elapsed_time":171.74,"gpr":1.0,"tpe":0.1968,"ee":0.1538},{"task_id":"task_345_enterprise_industry_policy_analysis_medium_medium039","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1478573,"cost_usd":0.14631,"elapsed_time":222.55,"gpr":0.5,"tpe":0.6561,"ee":0.1333},{"task_id":"task_346_enterprise_industry_policy_analysis_medium_medium040","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1632240,"cost_usd":0.209345,"elapsed_time":241.79,"gpr":0.75,"tpe":0.4056,"ee":0.129},{"task_id":"task_347_enterprise_industry_policy_analysis_medium_medium041","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1139654,"cost_usd":0.190072,"elapsed_time":175.38,"gpr":0.5,"tpe":1.0,"ee":0.3333},{"task_id":"task_348_enterprise_industry_policy_analysis_medium_medium042","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1256237,"cost_usd":0.119713,"elapsed_time":241.98,"gpr":0.5,"tpe":0.0552,"ee":0.1111},{"task_id":"task_349_enterprise_industry_policy_analysis_medium_medium043","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2249041,"cost_usd":0.209824,"elapsed_time":504.53,"gpr":0.75,"tpe":0.2459,"ee":0.0833},{"task_id":"task_350_enterprise_industry_policy_analysis_medium_medium044","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":407976,"cost_usd":0.058866,"elapsed_time":182.45,"gpr":0.5,"tpe":0.81,"ee":0.2353},{"task_id":"task_351_enterprise_industry_policy_analysis_medium_medium045","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":250464,"cost_usd":0.046261,"elapsed_time":138.71,"gpr":0.0,"tpe":0.0,"ee":0.3077},{"task_id":"task_352_enterprise_industry_policy_analysis_medium_medium046","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1022979,"cost_usd":0.124476,"elapsed_time":150.42,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_353_enterprise_industry_policy_analysis_medium_medium047","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1307670,"cost_usd":0.130986,"elapsed_time":325.28,"gpr":0.5,"tpe":0.855,"ee":0.093},{"task_id":"task_354_enterprise_industry_policy_analysis_medium_medium048","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":552279,"cost_usd":0.064593,"elapsed_time":174.76,"gpr":1.0,"tpe":0.4838,"ee":0.2},{"task_id":"task_355_enterprise_industry_policy_analysis_medium_medium049","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1849720,"cost_usd":0.173439,"elapsed_time":623.77,"gpr":0.75,"tpe":0.0589,"ee":0.0702},{"task_id":"task_356_enterprise_industry_policy_analysis_medium_medium050","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":759413,"cost_usd":0.079451,"elapsed_time":289.76,"gpr":0.5,"tpe":1.0,"ee":0.125},{"task_id":"task_357_enterprise_industry_policy_analysis_medium_medium051","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1213112,"cost_usd":0.17202,"elapsed_time":127.68,"gpr":0.0,"tpe":0.0,"ee":0.2353},{"task_id":"task_358_enterprise_industry_policy_analysis_medium_medium052","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1206616,"cost_usd":0.118739,"elapsed_time":365.26,"gpr":0.0,"tpe":0.0,"ee":0.0889},{"task_id":"task_359_enterprise_industry_policy_analysis_medium_medium053","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":819377,"cost_usd":0.100784,"elapsed_time":435.64,"gpr":0.75,"tpe":0.391,"ee":0.1176},{"task_id":"task_360_enterprise_industry_policy_analysis_medium_medium054","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":538249,"cost_usd":0.116676,"elapsed_time":87.38,"gpr":0.25,"tpe":0.81,"ee":0.4},{"task_id":"task_361_enterprise_industry_policy_analysis_medium_medium055","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":536841,"cost_usd":0.060444,"elapsed_time":219.14,"gpr":0.5,"tpe":0.4916,"ee":0.1818},{"task_id":"task_362_enterprise_industry_policy_analysis_medium_medium056","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":473284,"cost_usd":0.050954,"elapsed_time":173.51,"gpr":0.5,"tpe":0.3182,"ee":0.16},{"task_id":"task_363_enterprise_industry_policy_analysis_medium_medium057","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2687163,"cost_usd":0.227488,"elapsed_time":429.31,"gpr":0.25,"tpe":0.3138,"ee":0.0952},{"task_id":"task_364_enterprise_industry_policy_analysis_medium_medium058","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":294655,"cost_usd":0.05777,"elapsed_time":88.05,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_365_enterprise_industry_policy_analysis_medium_medium059","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2197919,"cost_usd":0.286775,"elapsed_time":418.28,"gpr":0.5,"tpe":0.5905,"ee":0.1818},{"task_id":"task_366_enterprise_industry_policy_analysis_medium_medium060","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":423281,"cost_usd":0.040194,"elapsed_time":176.07,"gpr":0.25,"tpe":0.5905,"ee":0.1429},{"task_id":"task_367_enterprise_industry_policy_analysis_medium_medium061","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1172887,"cost_usd":0.112462,"elapsed_time":261.82,"gpr":0.0,"tpe":0.0,"ee":0.093},{"task_id":"task_368_enterprise_industry_policy_analysis_medium_medium062","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1264431,"cost_usd":0.120563,"elapsed_time":456.24,"gpr":0.75,"tpe":0.2437,"ee":0.0755},{"task_id":"task_369_enterprise_industry_policy_analysis_medium_medium063","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":656391,"cost_usd":0.063353,"elapsed_time":141.64,"gpr":0.0,"tpe":0.0,"ee":0.1379},{"task_id":"task_370_enterprise_industry_policy_analysis_medium_medium064","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1687894,"cost_usd":0.208069,"elapsed_time":276.7,"gpr":0.25,"tpe":0.2288,"ee":0.1429},{"task_id":"task_371_enterprise_industry_policy_analysis_medium_medium065","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":3388600,"cost_usd":0.278683,"elapsed_time":555.41,"gpr":0.0,"tpe":0.0,"ee":0.0526},{"task_id":"task_372_enterprise_industry_policy_analysis_medium_medium066","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1151898,"cost_usd":0.107626,"elapsed_time":315.15,"gpr":0.0,"tpe":0.0,"ee":0.0909},{"task_id":"task_373_hypothesis_verification_hard_hard001","category":"hypothesis_verification","score":0.0,"tokens":510123,"cost_usd":0.056595,"elapsed_time":250.51,"gpr":0.1111,"tpe":1.0,"ee":0.3333},{"task_id":"task_374_hypothesis_verification_hard_hard002","category":"hypothesis_verification","score":0.0,"tokens":222170,"cost_usd":0.032228,"elapsed_time":150.97,"gpr":0.375,"tpe":0.6561,"ee":0.3571},{"task_id":"task_375_hypothesis_verification_hard_hard003","category":"hypothesis_verification","score":0.0,"tokens":976110,"cost_usd":0.1026,"elapsed_time":366.37,"gpr":0.2222,"tpe":0.4031,"ee":0.2105},{"task_id":"task_376_hypothesis_verification_hard_hard004","category":"hypothesis_verification","score":0.0,"tokens":956307,"cost_usd":0.102801,"elapsed_time":210.14,"gpr":0.1,"tpe":0.2542,"ee":0.28},{"task_id":"task_377_hypothesis_verification_hard_hard005","category":"hypothesis_verification","score":0.0,"tokens":1952963,"cost_usd":0.202683,"elapsed_time":260.19,"gpr":0.2308,"tpe":0.2406,"ee":0.3333},{"task_id":"task_378_hypothesis_verification_hard_hard006","category":"hypothesis_verification","score":0.0,"tokens":4527222,"cost_usd":0.457654,"elapsed_time":423.86,"gpr":0.0,"tpe":0.0,"ee":0.1458},{"task_id":"task_379_hypothesis_verification_hard_hard007","category":"hypothesis_verification","score":0.0,"tokens":4547385,"cost_usd":0.439494,"elapsed_time":857.44,"gpr":0.0625,"tpe":0.0225,"ee":0.1111},{"task_id":"task_380_hypothesis_verification_hard_hard008","category":"hypothesis_verification","score":0.0,"tokens":1647773,"cost_usd":0.167818,"elapsed_time":315.3,"gpr":0.5556,"tpe":0.3054,"ee":0.1795},{"task_id":"task_381_hypothesis_verification_hard_hard009","category":"hypothesis_verification","score":0.0,"tokens":598465,"cost_usd":0.06621,"elapsed_time":222.33,"gpr":0.1111,"tpe":1.0,"ee":0.2143},{"task_id":"task_382_hypothesis_verification_hard_hard010","category":"hypothesis_verification","score":0.0,"tokens":3997570,"cost_usd":0.554158,"elapsed_time":601.0,"gpr":0.0,"tpe":0.0,"ee":0.1458},{"task_id":"task_383_hypothesis_verification_hard_hard011","category":"hypothesis_verification","score":0.0,"tokens":1390843,"cost_usd":0.131746,"elapsed_time":375.9,"gpr":0.0,"tpe":0.0,"ee":0.1707},{"task_id":"task_384_hypothesis_verification_hard_hard012","category":"hypothesis_verification","score":0.0,"tokens":1369884,"cost_usd":0.158295,"elapsed_time":435.79,"gpr":0.4545,"tpe":1.0,"ee":0.2105},{"task_id":"task_385_hypothesis_verification_hard_hard013","category":"hypothesis_verification","score":0.0,"tokens":1997768,"cost_usd":0.18126,"elapsed_time":534.03,"gpr":0.2667,"tpe":0.1501,"ee":0.1458},{"task_id":"task_386_hypothesis_verification_hard_hard014","category":"hypothesis_verification","score":0.0,"tokens":2222718,"cost_usd":0.196335,"elapsed_time":513.52,"gpr":0.125,"tpe":0.81,"ee":0.1},{"task_id":"task_387_hypothesis_verification_hard_hard015","category":"hypothesis_verification","score":0.0,"tokens":1059571,"cost_usd":0.113362,"elapsed_time":375.97,"gpr":0.0909,"tpe":0.6561,"ee":0.1944},{"task_id":"task_388_hypothesis_verification_medium_medium001","category":"hypothesis_verification","score":0.0,"tokens":3462790,"cost_usd":0.322631,"elapsed_time":554.8,"gpr":0.0,"tpe":0.0,"ee":0.1458},{"task_id":"task_389_hypothesis_verification_medium_medium002","category":"hypothesis_verification","score":0.0,"tokens":529674,"cost_usd":0.05969,"elapsed_time":234.02,"gpr":0.0,"tpe":0.0,"ee":0.2188},{"task_id":"task_390_hypothesis_verification_medium_medium003","category":"hypothesis_verification","score":0.0,"tokens":1703461,"cost_usd":0.172815,"elapsed_time":544.54,"gpr":0.0,"tpe":0.0,"ee":0.1176},{"task_id":"task_391_hypothesis_verification_medium_medium004","category":"hypothesis_verification","score":0.0,"tokens":667893,"cost_usd":0.07127,"elapsed_time":172.45,"gpr":0.2,"tpe":1.0,"ee":0.2593},{"task_id":"task_392_hypothesis_verification_medium_medium005","category":"hypothesis_verification","score":0.0,"tokens":1765596,"cost_usd":0.175234,"elapsed_time":542.37,"gpr":0.4,"tpe":0.5644,"ee":0.1458},{"task_id":"task_393_hypothesis_verification_medium_medium006","category":"hypothesis_verification","score":0.0,"tokens":789574,"cost_usd":0.089841,"elapsed_time":344.75,"gpr":0.0,"tpe":0.0,"ee":0.1818},{"task_id":"task_394_hypothesis_verification_medium_medium007","category":"hypothesis_verification","score":0.0,"tokens":1513333,"cost_usd":0.164469,"elapsed_time":478.12,"gpr":0.0,"tpe":0.0,"ee":0.1522},{"task_id":"task_395_hypothesis_verification_medium_medium008","category":"hypothesis_verification","score":0.0,"tokens":529334,"cost_usd":0.071858,"elapsed_time":274.13,"gpr":0.2,"tpe":0.2542,"ee":0.381},{"task_id":"task_396_hypothesis_verification_medium_medium009","category":"hypothesis_verification","score":0.0,"tokens":1613209,"cost_usd":0.175169,"elapsed_time":215.25,"gpr":0.25,"tpe":0.4305,"ee":0.375},{"task_id":"task_397_hypothesis_verification_medium_medium010","category":"hypothesis_verification","score":0.0,"tokens":680897,"cost_usd":0.084748,"elapsed_time":271.35,"gpr":0.5714,"tpe":0.6342,"ee":0.3333},{"task_id":"task_398_hypothesis_verification_medium_medium011","category":"hypothesis_verification","score":0.0,"tokens":4011299,"cost_usd":0.391597,"elapsed_time":756.27,"gpr":0.1667,"tpe":0.0025,"ee":0.0882},{"task_id":"task_399_hypothesis_verification_medium_medium012","category":"hypothesis_verification","score":1.0,"tokens":639180,"cost_usd":0.073386,"elapsed_time":257.86,"gpr":1.0,"tpe":0.133,"ee":0.2308},{"task_id":"task_400_hypothesis_verification_medium_medium013","category":"hypothesis_verification","score":0.0,"tokens":1548727,"cost_usd":0.170926,"elapsed_time":505.68,"gpr":0.3333,"tpe":0.0471,"ee":0.175},{"task_id":"task_401_hypothesis_verification_medium_medium014","category":"hypothesis_verification","score":1.0,"tokens":573292,"cost_usd":0.07006,"elapsed_time":151.24,"gpr":1.0,"tpe":0.4709,"ee":0.4},{"task_id":"task_402_industry_planning_hard_hard001","category":"industry_planning","score":0.0,"tokens":2002933,"cost_usd":0.358762,"elapsed_time":190.24,"gpr":0.2222,"tpe":0.6925,"ee":0.3684},{"task_id":"task_403_industry_planning_hard_hard002","category":"industry_planning","score":0.0,"tokens":6258795,"cost_usd":0.661427,"elapsed_time":724.9,"gpr":0.25,"tpe":0.7031,"ee":0.1522},{"task_id":"task_404_industry_planning_hard_hard003","category":"industry_planning","score":0.0,"tokens":1389733,"cost_usd":0.149517,"elapsed_time":406.16,"gpr":0.2727,"tpe":0.7072,"ee":0.2424},{"task_id":"task_405_industry_planning_hard_hard004","category":"industry_planning","score":1.0,"tokens":938784,"cost_usd":0.114115,"elapsed_time":366.05,"gpr":1.0,"tpe":0.1988,"ee":0.2222},{"task_id":"task_406_industry_planning_hard_hard005","category":"industry_planning","score":0.0,"tokens":2342058,"cost_usd":0.230799,"elapsed_time":515.47,"gpr":0.4444,"tpe":0.6172,"ee":0.1224},{"task_id":"task_407_industry_planning_hard_hard006","category":"industry_planning","score":0.0,"tokens":1887088,"cost_usd":0.179258,"elapsed_time":449.9,"gpr":0.25,"tpe":0.4963,"ee":0.1489},{"task_id":"task_408_industry_planning_hard_hard007","category":"industry_planning","score":0.0,"tokens":1365458,"cost_usd":0.132993,"elapsed_time":362.51,"gpr":0.1,"tpe":0.0523,"ee":0.1364},{"task_id":"task_409_industry_planning_hard_hard008","category":"industry_planning","score":0.5,"tokens":3386821,"cost_usd":0.322276,"elapsed_time":873.48,"gpr":0.125,"tpe":0.1094,"ee":0.0923},{"task_id":"task_410_industry_planning_hard_hard009","category":"industry_planning","score":0.0,"tokens":3501113,"cost_usd":0.427445,"elapsed_time":804.38,"gpr":0.4,"tpe":0.4689,"ee":0.1111},{"task_id":"task_411_industry_planning_hard_hard010","category":"industry_planning","score":0.0,"tokens":3009476,"cost_usd":0.298366,"elapsed_time":594.74,"gpr":0.2222,"tpe":0.2019,"ee":0.1277},{"task_id":"task_412_industry_planning_hard_hard011","category":"industry_planning","score":1.0,"tokens":1340707,"cost_usd":0.170682,"elapsed_time":746.03,"gpr":1.0,"tpe":0.0293,"ee":0.1395},{"task_id":"task_413_industry_planning_hard_hard012","category":"industry_planning","score":1.0,"tokens":1415164,"cost_usd":0.229693,"elapsed_time":783.25,"gpr":1.0,"tpe":0.2174,"ee":0.1875},{"task_id":"task_414_industry_planning_hard_hard013","category":"industry_planning","score":0.0,"tokens":2438804,"cost_usd":0.251585,"elapsed_time":760.78,"gpr":0.0,"tpe":0.0,"ee":0.1207},{"task_id":"task_415_industry_planning_hard_hard014","category":"industry_planning","score":0.0,"tokens":2933017,"cost_usd":0.333395,"elapsed_time":1208.59,"gpr":0.375,"tpe":0.0395,"ee":0.0735},{"task_id":"task_416_industry_planning_medium_medium001","category":"industry_planning","score":0.5,"tokens":1075502,"cost_usd":0.114501,"elapsed_time":389.37,"gpr":0.1667,"tpe":0.1216,"ee":0.2581},{"task_id":"task_417_industry_planning_medium_medium002","category":"industry_planning","score":0.0,"tokens":1858136,"cost_usd":0.197218,"elapsed_time":625.98,"gpr":0.6,"tpe":0.0638,"ee":0.1667},{"task_id":"task_418_industry_planning_medium_medium003","category":"industry_planning","score":0.0,"tokens":1799865,"cost_usd":0.182131,"elapsed_time":563.09,"gpr":0.0,"tpe":0.0,"ee":0.2162},{"task_id":"task_419_industry_planning_medium_medium004","category":"industry_planning","score":1.0,"tokens":1969679,"cost_usd":0.189883,"elapsed_time":701.93,"gpr":1.0,"tpe":0.129,"ee":0.1321},{"task_id":"task_420_industry_planning_medium_medium005","category":"industry_planning","score":0.0,"tokens":1817384,"cost_usd":0.202617,"elapsed_time":710.4,"gpr":0.3333,"tpe":0.0646,"ee":0.1892},{"task_id":"task_421_industry_planning_medium_medium006","category":"industry_planning","score":0.0,"tokens":940947,"cost_usd":0.122275,"elapsed_time":457.03,"gpr":0.0,"tpe":0.0,"ee":0.3478},{"task_id":"task_422_industry_planning_medium_medium007","category":"industry_planning","score":1.0,"tokens":3911144,"cost_usd":0.596226,"elapsed_time":823.25,"gpr":0.0,"tpe":0.0,"ee":0.1702},{"task_id":"task_423_industry_planning_medium_medium008","category":"industry_planning","score":1.0,"tokens":1638249,"cost_usd":0.165059,"elapsed_time":596.7,"gpr":1.0,"tpe":0.0278,"ee":0.1905},{"task_id":"task_424_industry_planning_medium_medium009","category":"industry_planning","score":1.0,"tokens":1091279,"cost_usd":0.121704,"elapsed_time":525.46,"gpr":1.0,"tpe":0.179,"ee":0.2222},{"task_id":"task_425_industry_planning_medium_medium010","category":"industry_planning","score":0.0,"tokens":1138645,"cost_usd":0.155725,"elapsed_time":401.01,"gpr":0.4,"tpe":0.2307,"ee":0.32},{"task_id":"task_426_industry_planning_medium_medium011","category":"industry_planning","score":1.0,"tokens":1971255,"cost_usd":0.184747,"elapsed_time":537.44,"gpr":0.5833,"tpe":0.0035,"ee":0.1176},{"task_id":"task_427_industry_planning_medium_medium012","category":"industry_planning","score":0.0,"tokens":5013797,"cost_usd":0.51248,"elapsed_time":1042.21,"gpr":0.7778,"tpe":0.0002,"ee":0.0851},{"task_id":"task_428_industry_planning_medium_medium013","category":"industry_planning","score":0.0,"tokens":1356827,"cost_usd":0.161173,"elapsed_time":637.46,"gpr":0.1667,"tpe":0.5314,"ee":0.2105},{"task_id":"task_429_industry_planning_medium_medium014","category":"industry_planning","score":0.0,"tokens":1458970,"cost_usd":0.171232,"elapsed_time":271.87,"gpr":0.3333,"tpe":1.0,"ee":0.3462},{"task_id":"task_430_international_comparison_hard_hard001","category":"international_comparison","score":0.25,"tokens":2822710,"cost_usd":0.291063,"elapsed_time":378.27,"gpr":0.2,"tpe":0.5032,"ee":0.1111},{"task_id":"task_431_international_comparison_hard_hard002","category":"international_comparison","score":0.0,"tokens":5411026,"cost_usd":0.761935,"elapsed_time":840.26,"gpr":0.0,"tpe":0.0,"ee":0.0652},{"task_id":"task_432_international_comparison_hard_hard003","category":"international_comparison","score":0.0,"tokens":6525910,"cost_usd":0.563198,"elapsed_time":553.25,"gpr":0.2222,"tpe":0.6561,"ee":0.0706},{"task_id":"task_433_international_comparison_hard_hard004","category":"international_comparison","score":0.0,"tokens":3716493,"cost_usd":0.279148,"elapsed_time":440.04,"gpr":0.1818,"tpe":0.0424,"ee":0.1224},{"task_id":"task_434_international_comparison_hard_hard005","category":"international_comparison","score":0.0,"tokens":2679487,"cost_usd":0.302463,"elapsed_time":382.37,"gpr":0.1818,"tpe":0.5314,"ee":0.1714},{"task_id":"task_435_international_comparison_hard_hard006","category":"international_comparison","score":0.25,"tokens":1700641,"cost_usd":0.281477,"elapsed_time":267.35,"gpr":0.1667,"tpe":0.6561,"ee":0.5},{"task_id":"task_436_international_comparison_hard_hard007","category":"international_comparison","score":0.0,"tokens":3331136,"cost_usd":0.572092,"elapsed_time":450.07,"gpr":0.1818,"tpe":0.1703,"ee":0.2059},{"task_id":"task_437_international_comparison_hard_hard008","category":"international_comparison","score":0.0,"tokens":1057012,"cost_usd":0.09705,"elapsed_time":366.15,"gpr":0.1667,"tpe":0.4305,"ee":0.2143},{"task_id":"task_438_international_comparison_hard_hard009","category":"international_comparison","score":0.0,"tokens":2875804,"cost_usd":0.443722,"elapsed_time":440.73,"gpr":0,"tpe":0.0,"ee":0},{"task_id":"task_439_international_comparison_hard_hard010","category":"international_comparison","score":0.0,"tokens":1525509,"cost_usd":0.226409,"elapsed_time":236.86,"gpr":0.3333,"tpe":0.1894,"ee":0.25},{"task_id":"task_440_international_comparison_hard_hard011","category":"international_comparison","score":0.25,"tokens":378397,"cost_usd":0.039405,"elapsed_time":140.16,"gpr":0.6,"tpe":0.3919,"ee":0.375},{"task_id":"task_441_international_comparison_hard_hard012","category":"international_comparison","score":0.0,"tokens":713748,"cost_usd":0.09758,"elapsed_time":306.58,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_442_international_comparison_hard_hard013","category":"international_comparison","score":0.25,"tokens":2449837,"cost_usd":0.337222,"elapsed_time":330.44,"gpr":0.6,"tpe":0.4373,"ee":0.2609},{"task_id":"task_443_international_comparison_hard_hard014","category":"international_comparison","score":0.5,"tokens":1376679,"cost_usd":0.116555,"elapsed_time":280.86,"gpr":0.5,"tpe":0.0646,"ee":0.1875},{"task_id":"task_444_international_comparison_medium_medium001","category":"international_comparison","score":0.0,"tokens":3118343,"cost_usd":0.248651,"elapsed_time":589.73,"gpr":1.0,"tpe":0.0028,"ee":0.0656},{"task_id":"task_445_international_comparison_medium_medium002","category":"international_comparison","score":0.0,"tokens":5875954,"cost_usd":0.759753,"elapsed_time":1207.35,"gpr":0.1667,"tpe":0.3138,"ee":0.0444},{"task_id":"task_446_international_comparison_medium_medium003","category":"international_comparison","score":0.5,"tokens":1200257,"cost_usd":0.102816,"elapsed_time":323.18,"gpr":0.6667,"tpe":0.276,"ee":0.1},{"task_id":"task_447_international_comparison_medium_medium004","category":"international_comparison","score":0.0,"tokens":482208,"cost_usd":0.063603,"elapsed_time":295.86,"gpr":0.6667,"tpe":0.3547,"ee":0.2353},{"task_id":"task_448_international_comparison_medium_medium005","category":"international_comparison","score":0.0,"tokens":3062043,"cost_usd":0.290097,"elapsed_time":828.18,"gpr":1.0,"tpe":0.0253,"ee":0.0678},{"task_id":"task_449_international_comparison_medium_medium006","category":"international_comparison","score":0.0,"tokens":2537736,"cost_usd":0.337845,"elapsed_time":458.39,"gpr":0.1667,"tpe":0.0523,"ee":0.0806},{"task_id":"task_450_international_comparison_medium_medium007","category":"international_comparison","score":0.33,"tokens":5123926,"cost_usd":0.507505,"elapsed_time":1003.47,"gpr":0.5,"tpe":0.0029,"ee":0.0597},{"task_id":"task_451_international_comparison_medium_medium008","category":"international_comparison","score":1.0,"tokens":406913,"cost_usd":0.045268,"elapsed_time":216.47,"gpr":1.0,"tpe":0.2891,"ee":0.2222},{"task_id":"task_452_international_comparison_medium_medium009","category":"international_comparison","score":0.0,"tokens":2721025,"cost_usd":0.206256,"elapsed_time":352.1,"gpr":0.5714,"tpe":0.0421,"ee":0.1163},{"task_id":"task_453_international_comparison_medium_medium010","category":"international_comparison","score":0.5,"tokens":1260560,"cost_usd":0.109131,"elapsed_time":344.84,"gpr":0.6,"tpe":0.3865,"ee":0.1471},{"task_id":"task_454_international_comparison_medium_medium011","category":"international_comparison","score":0.333,"tokens":1731988,"cost_usd":0.160769,"elapsed_time":465.41,"gpr":0.3333,"tpe":0.0424,"ee":0.093},{"task_id":"task_455_international_comparison_medium_medium012","category":"international_comparison","score":0.3333,"tokens":4682676,"cost_usd":0.402221,"elapsed_time":432.57,"gpr":0.5,"tpe":0.003,"ee":0.0806},{"task_id":"task_456_international_comparison_medium_medium013","category":"international_comparison","score":1.0,"tokens":1572064,"cost_usd":0.176767,"elapsed_time":175.08,"gpr":1.0,"tpe":0.2063,"ee":0.2},{"task_id":"task_457_international_comparison_medium_medium014","category":"international_comparison","score":0.0,"tokens":1233881,"cost_usd":0.111868,"elapsed_time":384.14,"gpr":0.5,"tpe":0.0911,"ee":0.125},{"task_id":"task_458_international_comparison_medium_medium015","category":"international_comparison","score":0.0,"tokens":2456529,"cost_usd":0.201737,"elapsed_time":520.6,"gpr":0.1429,"tpe":0.0182,"ee":0.0877},{"task_id":"task_459_international_comparison_medium_medium016","category":"international_comparison","score":0.5,"tokens":4164173,"cost_usd":0.616308,"elapsed_time":724.99,"gpr":0.5,"tpe":0.0071,"ee":0.0962},{"task_id":"task_460_international_comparison_medium_medium017","category":"international_comparison","score":0.0,"tokens":1751199,"cost_usd":0.156166,"elapsed_time":614.3,"gpr":0.0,"tpe":0.0,"ee":0.1111},{"task_id":"task_461_international_comparison_medium_medium018","category":"international_comparison","score":0.5,"tokens":4712000,"cost_usd":0.478105,"elapsed_time":671.65,"gpr":0.5714,"tpe":0.2845,"ee":0.0702},{"task_id":"task_462_international_comparison_medium_medium019","category":"international_comparison","score":0.5,"tokens":4858038,"cost_usd":0.452549,"elapsed_time":836.03,"gpr":0.6667,"tpe":0.1121,"ee":0.0741},{"task_id":"task_463_international_comparison_medium_medium020","category":"international_comparison","score":0.0,"tokens":2529527,"cost_usd":0.255161,"elapsed_time":507.85,"gpr":0.3333,"tpe":0.0529,"ee":0.1064},{"task_id":"task_464_international_comparison_medium_medium021","category":"international_comparison","score":0.0,"tokens":3061359,"cost_usd":0.256723,"elapsed_time":652.78,"gpr":0.4286,"tpe":0.0189,"ee":0.098},{"task_id":"task_465_international_comparison_medium_medium022","category":"international_comparison","score":0.67,"tokens":3459146,"cost_usd":0.257859,"elapsed_time":314.08,"gpr":0.4545,"tpe":0.0507,"ee":0.0962},{"task_id":"task_466_international_comparison_medium_medium023","category":"international_comparison","score":0.5,"tokens":3098466,"cost_usd":0.288932,"elapsed_time":477.41,"gpr":0.5,"tpe":0.2401,"ee":0.119},{"task_id":"task_467_international_comparison_medium_medium024","category":"international_comparison","score":0.33,"tokens":1285012,"cost_usd":0.130646,"elapsed_time":567.01,"gpr":0.3333,"tpe":0.1466,"ee":0.125},{"task_id":"task_468_international_comparison_medium_medium025","category":"international_comparison","score":0.0,"tokens":4079188,"cost_usd":0.50416,"elapsed_time":742.11,"gpr":0.5,"tpe":0.5177,"ee":0.1026},{"task_id":"task_469_risk_assessment_hard_hard001","category":"risk_assessment","score":1.0,"tokens":2009825,"cost_usd":0.218698,"elapsed_time":425.78,"gpr":1.0,"tpe":0.1126,"ee":0.1795},{"task_id":"task_470_risk_assessment_hard_hard002","category":"risk_assessment","score":1.0,"tokens":1556614,"cost_usd":0.14908,"elapsed_time":491.97,"gpr":1.0,"tpe":0.1678,"ee":0.1489},{"task_id":"task_471_risk_assessment_hard_hard003","category":"risk_assessment","score":0.3333,"tokens":3056460,"cost_usd":0.305292,"elapsed_time":443.56,"gpr":0.2727,"tpe":0.2249,"ee":0.1892},{"task_id":"task_472_risk_assessment_hard_hard004","category":"risk_assessment","score":1.0,"tokens":2141710,"cost_usd":0.217263,"elapsed_time":860.89,"gpr":1.0,"tpe":0.0541,"ee":0.1346},{"task_id":"task_473_risk_assessment_hard_hard005","category":"risk_assessment","score":1.0,"tokens":486785,"cost_usd":0.065764,"elapsed_time":217.37,"gpr":1.0,"tpe":0.3318,"ee":0.3684},{"task_id":"task_474_risk_assessment_hard_hard006","category":"risk_assessment","score":1.0,"tokens":442689,"cost_usd":0.062987,"elapsed_time":224.78,"gpr":1.0,"tpe":0.2824,"ee":0.3684},{"task_id":"task_475_risk_assessment_hard_hard007","category":"risk_assessment","score":1.0,"tokens":689715,"cost_usd":0.092075,"elapsed_time":239.68,"gpr":1.0,"tpe":0.6243,"ee":0.4667},{"task_id":"task_476_risk_assessment_hard_hard008","category":"risk_assessment","score":0.0,"tokens":1110427,"cost_usd":0.12307,"elapsed_time":323.77,"gpr":0.5,"tpe":0.2288,"ee":0.2692},{"task_id":"task_477_risk_assessment_hard_hard009","category":"risk_assessment","score":0.0,"tokens":2721333,"cost_usd":0.274445,"elapsed_time":774.24,"gpr":0.0,"tpe":0.0,"ee":0.1556},{"task_id":"task_478_risk_assessment_hard_hard010","category":"risk_assessment","score":0.0,"tokens":3017799,"cost_usd":0.380037,"elapsed_time":414.23,"gpr":0.0,"tpe":0.0,"ee":0.2424},{"task_id":"task_479_risk_assessment_hard_hard011","category":"risk_assessment","score":0.0,"tokens":3906934,"cost_usd":0.355019,"elapsed_time":455.63,"gpr":0.3,"tpe":0.1347,"ee":0.1944},{"task_id":"task_480_risk_assessment_hard_hard012","category":"risk_assessment","score":0.0,"tokens":1986860,"cost_usd":0.18467,"elapsed_time":493.02,"gpr":0.75,"tpe":0.0149,"ee":0.1429},{"task_id":"task_481_risk_assessment_hard_hard013","category":"risk_assessment","score":0.0,"tokens":2829862,"cost_usd":0.267412,"elapsed_time":651.72,"gpr":0.3333,"tpe":0.0182,"ee":0.1607},{"task_id":"task_482_risk_assessment_medium_medium001","category":"risk_assessment","score":0.5,"tokens":2371527,"cost_usd":0.219469,"elapsed_time":393.66,"gpr":0.6667,"tpe":0.7391,"ee":0.2143},{"task_id":"task_483_risk_assessment_medium_medium002","category":"risk_assessment","score":0.0,"tokens":642080,"cost_usd":0.070092,"elapsed_time":230.44,"gpr":0.25,"tpe":0.1668,"ee":0.2069},{"task_id":"task_484_risk_assessment_medium_medium003","category":"risk_assessment","score":0.0,"tokens":1767560,"cost_usd":0.19395,"elapsed_time":514.74,"gpr":0.0,"tpe":0.0,"ee":0.15},{"task_id":"task_485_risk_assessment_medium_medium004","category":"risk_assessment","score":0.0,"tokens":4094158,"cost_usd":0.490781,"elapsed_time":452.3,"gpr":0.7,"tpe":0.1481,"ee":0.1515},{"task_id":"task_486_risk_assessment_medium_medium005","category":"risk_assessment","score":1.0,"tokens":2148657,"cost_usd":0.198854,"elapsed_time":370.39,"gpr":0.875,"tpe":0.2668,"ee":0.1852},{"task_id":"task_487_risk_assessment_medium_medium006","category":"risk_assessment","score":0.0,"tokens":2689412,"cost_usd":0.307151,"elapsed_time":747.48,"gpr":0.7273,"tpe":0.6024,"ee":0.119},{"task_id":"task_488_risk_assessment_medium_medium007","category":"risk_assessment","score":0.0,"tokens":618569,"cost_usd":0.075358,"elapsed_time":239.51,"gpr":0.1667,"tpe":1.0,"ee":0.2308},{"task_id":"task_489_risk_assessment_medium_medium008","category":"risk_assessment","score":0.0,"tokens":2805789,"cost_usd":0.292936,"elapsed_time":431.93,"gpr":0.1875,"tpe":0.3874,"ee":0.1667},{"task_id":"task_490_risk_assessment_medium_medium009","category":"risk_assessment","score":1.0,"tokens":3256992,"cost_usd":0.514576,"elapsed_time":674.45,"gpr":1.0,"tpe":0.2341,"ee":0.1429},{"task_id":"task_491_risk_assessment_medium_medium010","category":"risk_assessment","score":0.0,"tokens":2215458,"cost_usd":0.24329,"elapsed_time":595.28,"gpr":0.4,"tpe":0.684,"ee":0.1333},{"task_id":"task_492_risk_assessment_medium_medium011","category":"risk_assessment","score":0.0,"tokens":1195443,"cost_usd":0.135368,"elapsed_time":425.54,"gpr":0.0909,"tpe":0.3874,"ee":0.1667}]},{"model":"qwen3.5-plus","acc":0.377,"ee":0.2421,"gpr":0.2168,"tpe":0.1985,"efficiency":{"total_tokens":907110973,"total_cost_usd":240.01,"total_requests":14739},"category_scores":{"comprehensive_decision":{"acc":0.3953,"ee":0.1983,"gpr":0.2004,"tpe":0.1662,"count":70},"enterprise_industry_analysis":{"acc":0.4646,"ee":0.2605,"gpr":0.164,"tpe":0.1387,"count":226},"enterprise_industry_policy_analysis":{"acc":0.3421,"ee":0.2147,"gpr":0.304,"tpe":0.2004,"count":76},"hypothesis_verification":{"acc":0.2069,"ee":0.2164,"gpr":0.1905,"tpe":0.1893,"count":29},"industry_planning":{"acc":0.2321,"ee":0.2095,"gpr":0.2893,"tpe":0.206,"count":28},"international_comparison":{"acc":0.1621,"ee":0.55,"gpr":0.2652,"tpe":0.429,"count":39},"risk_assessment":{"acc":0.3333,"ee":0.1863,"gpr":0.2154,"tpe":0.2059,"count":24}},"tasks":[{"task_id":"task_001_comprehensive_decision_easy_easy001","category":"comprehensive_decision","score":0.0,"tokens":390956,"cost_usd":0.106699,"elapsed_time":122.31,"gpr":0.3333,"tpe":0.3487,"ee":0.125},{"task_id":"task_002_comprehensive_decision_easy_easy002","category":"comprehensive_decision","score":1.0,"tokens":144596,"cost_usd":0.045304,"elapsed_time":69.39,"gpr":1.0,"tpe":0.6872,"ee":0.2222},{"task_id":"task_003_comprehensive_decision_easy_easy003","category":"comprehensive_decision","score":0.0,"tokens":2002490,"cost_usd":0.523272,"elapsed_time":319.14,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_004_comprehensive_decision_easy_easy004","category":"comprehensive_decision","score":1.0,"tokens":303641,"cost_usd":0.090783,"elapsed_time":112.07,"gpr":1.0,"tpe":0.4509,"ee":0.125},{"task_id":"task_005_comprehensive_decision_easy_easy005","category":"comprehensive_decision","score":0.0,"tokens":2247207,"cost_usd":0.59614,"elapsed_time":219.33,"gpr":0.6667,"tpe":0.5388,"ee":0.1154},{"task_id":"task_006_comprehensive_decision_easy_easy006","category":"comprehensive_decision","score":0.0,"tokens":375656,"cost_usd":0.112875,"elapsed_time":190.0,"gpr":0.5,"tpe":0.3874,"ee":0.1579},{"task_id":"task_007_comprehensive_decision_hard_hard001","category":"comprehensive_decision","score":0.0,"tokens":14449295,"cost_usd":3.766275,"elapsed_time":736.18,"gpr":0.0,"tpe":0.0,"ee":0.0959},{"task_id":"task_008_comprehensive_decision_hard_hard002","category":"comprehensive_decision","score":0.0,"tokens":958926,"cost_usd":0.270972,"elapsed_time":405.35,"gpr":0.0,"tpe":0.0,"ee":0.1591},{"task_id":"task_009_comprehensive_decision_hard_hard003","category":"comprehensive_decision","score":0.0,"tokens":8333855,"cost_usd":2.182983,"elapsed_time":389.93,"gpr":0.1429,"tpe":1.0,"ee":0.3333},{"task_id":"task_010_comprehensive_decision_hard_hard004","category":"comprehensive_decision","score":0.0,"tokens":1449921,"cost_usd":0.395913,"elapsed_time":478.27,"gpr":0.25,"tpe":0.0105,"ee":0.0986},{"task_id":"task_011_comprehensive_decision_hard_hard005","category":"comprehensive_decision","score":0.0,"tokens":390834,"cost_usd":0.105188,"elapsed_time":172.22,"gpr":0.5714,"tpe":0.8575,"ee":0.2188},{"task_id":"task_012_comprehensive_decision_hard_hard006","category":"comprehensive_decision","score":0.0,"tokens":2466116,"cost_usd":0.67404,"elapsed_time":753.81,"gpr":0.5,"tpe":0.0002,"ee":0.0778},{"task_id":"task_013_comprehensive_decision_hard_hard007","category":"comprehensive_decision","score":0.0,"tokens":1802838,"cost_usd":0.48,"elapsed_time":372.9,"gpr":0.6667,"tpe":0.0368,"ee":0.1458},{"task_id":"task_014_comprehensive_decision_hard_hard008","category":"comprehensive_decision","score":0.0,"tokens":57152,"cost_usd":0.015915,"elapsed_time":61.33,"gpr":0.0,"tpe":0.0,"ee":0.7},{"task_id":"task_015_comprehensive_decision_hard_hard009","category":"comprehensive_decision","score":0.0,"tokens":23563,"cost_usd":0.006952,"elapsed_time":45.8,"gpr":0.4286,"tpe":1.0,"ee":1.3333},{"task_id":"task_016_comprehensive_decision_hard_hard010","category":"comprehensive_decision","score":0.0,"tokens":3550352,"cost_usd":0.926719,"elapsed_time":199.13,"gpr":0.0,"tpe":0.0,"ee":0.3043},{"task_id":"task_017_comprehensive_decision_hard_hard011","category":"comprehensive_decision","score":0.0,"tokens":3555563,"cost_usd":0.928674,"elapsed_time":291.27,"gpr":0.1111,"tpe":1.0,"ee":0.2414},{"task_id":"task_018_comprehensive_decision_hard_hard012","category":"comprehensive_decision","score":1.0,"tokens":861917,"cost_usd":0.237972,"elapsed_time":349.89,"gpr":1.0,"tpe":0.0239,"ee":0.1556},{"task_id":"task_019_comprehensive_decision_hard_hard013","category":"comprehensive_decision","score":0.5,"tokens":1785748,"cost_usd":0.476111,"elapsed_time":317.2,"gpr":0.0,"tpe":0.0,"ee":0.2258},{"task_id":"task_020_comprehensive_decision_hard_hard014","category":"comprehensive_decision","score":0.0,"tokens":1260408,"cost_usd":0.344861,"elapsed_time":211.87,"gpr":0.2857,"tpe":0.5314,"ee":0.2593},{"task_id":"task_021_comprehensive_decision_hard_hard015","category":"comprehensive_decision","score":0.0,"tokens":2161929,"cost_usd":0.598512,"elapsed_time":612.94,"gpr":0.2857,"tpe":0.0038,"ee":0.1143},{"task_id":"task_022_comprehensive_decision_hard_hard016","category":"comprehensive_decision","score":0.0,"tokens":21553770,"cost_usd":5.632279,"elapsed_time":875.11,"gpr":0.375,"tpe":0.0094,"ee":0.0769},{"task_id":"task_023_comprehensive_decision_hard_hard017","category":"comprehensive_decision","score":0.0,"tokens":1815150,"cost_usd":0.480002,"elapsed_time":197.83,"gpr":0.5,"tpe":0.0886,"ee":0.2069},{"task_id":"task_024_comprehensive_decision_hard_hard018","category":"comprehensive_decision","score":1.0,"tokens":1889372,"cost_usd":0.502353,"elapsed_time":283.86,"gpr":1.0,"tpe":0.1917,"ee":0.2917},{"task_id":"task_025_comprehensive_decision_hard_hard019","category":"comprehensive_decision","score":0.67,"tokens":2886248,"cost_usd":0.763471,"elapsed_time":243.85,"gpr":0.7143,"tpe":0.1501,"ee":0.25},{"task_id":"task_026_comprehensive_decision_medium_medium001","category":"comprehensive_decision","score":0.0,"tokens":877510,"cost_usd":0.264195,"elapsed_time":464.52,"gpr":0.0,"tpe":0.0,"ee":0.098},{"task_id":"task_027_comprehensive_decision_medium_medium002","category":"comprehensive_decision","score":1.0,"tokens":1159887,"cost_usd":0.315862,"elapsed_time":362.69,"gpr":1.0,"tpe":0.2783,"ee":0.0877},{"task_id":"task_028_comprehensive_decision_medium_medium003","category":"comprehensive_decision","score":0.0,"tokens":16896968,"cost_usd":4.400116,"elapsed_time":497.79,"gpr":0.1429,"tpe":0.1094,"ee":0.0769},{"task_id":"task_029_comprehensive_decision_medium_medium004","category":"comprehensive_decision","score":0.5,"tokens":1096276,"cost_usd":0.317013,"elapsed_time":347.37,"gpr":0.875,"tpe":0.0858,"ee":0.122},{"task_id":"task_030_comprehensive_decision_medium_medium005","category":"comprehensive_decision","score":1.0,"tokens":1672214,"cost_usd":0.441282,"elapsed_time":220.06,"gpr":1.0,"tpe":0.0407,"ee":0.1351},{"task_id":"task_031_comprehensive_decision_medium_medium006","category":"comprehensive_decision","score":1.0,"tokens":1594777,"cost_usd":0.418581,"elapsed_time":189.52,"gpr":0.0,"tpe":0.0,"ee":0.2353},{"task_id":"task_032_comprehensive_decision_medium_medium007","category":"comprehensive_decision","score":0.0,"tokens":7756989,"cost_usd":2.057575,"elapsed_time":358.69,"gpr":0.0,"tpe":0.0,"ee":0.122},{"task_id":"task_033_comprehensive_decision_medium_medium008","category":"comprehensive_decision","score":1.0,"tokens":1153736,"cost_usd":0.314633,"elapsed_time":291.89,"gpr":1.0,"tpe":0.0068,"ee":0.0909},{"task_id":"task_034_comprehensive_decision_medium_medium009","category":"comprehensive_decision","score":1.0,"tokens":772894,"cost_usd":0.212188,"elapsed_time":215.62,"gpr":1.0,"tpe":0.0872,"ee":0.1136},{"task_id":"task_035_comprehensive_decision_medium_medium010","category":"comprehensive_decision","score":0.5,"tokens":122055,"cost_usd":0.034289,"elapsed_time":56.6,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_036_comprehensive_decision_medium_medium011","category":"comprehensive_decision","score":1.0,"tokens":236798,"cost_usd":0.066281,"elapsed_time":85.09,"gpr":1.0,"tpe":0.5905,"ee":0.3846},{"task_id":"task_037_comprehensive_decision_medium_medium012","category":"comprehensive_decision","score":1.0,"tokens":922934,"cost_usd":0.257166,"elapsed_time":321.63,"gpr":1.0,"tpe":0.1589,"ee":0.1471},{"task_id":"task_038_comprehensive_decision_medium_medium013","category":"comprehensive_decision","score":1.0,"tokens":1681803,"cost_usd":0.447744,"elapsed_time":221.12,"gpr":1.0,"tpe":0.3419,"ee":0.1389},{"task_id":"task_039_comprehensive_decision_medium_medium014","category":"comprehensive_decision","score":0.5,"tokens":2928919,"cost_usd":0.788656,"elapsed_time":756.32,"gpr":0.6667,"tpe":0.0057,"ee":0.049},{"task_id":"task_040_comprehensive_decision_medium_medium015","category":"comprehensive_decision","score":0.0,"tokens":2827626,"cost_usd":0.743338,"elapsed_time":156.55,"gpr":0.25,"tpe":0.4783,"ee":0.3846},{"task_id":"task_041_comprehensive_decision_medium_medium016","category":"comprehensive_decision","score":1.0,"tokens":995394,"cost_usd":0.274479,"elapsed_time":346.81,"gpr":1.0,"tpe":0.0203,"ee":0.119},{"task_id":"task_042_comprehensive_decision_medium_medium017","category":"comprehensive_decision","score":0.0,"tokens":6651122,"cost_usd":1.735517,"elapsed_time":345.42,"gpr":0.0,"tpe":0.0,"ee":0.1064},{"task_id":"task_043_comprehensive_decision_medium_medium018","category":"comprehensive_decision","score":0.0,"tokens":2571649,"cost_usd":0.671038,"elapsed_time":166.7,"gpr":0.0,"tpe":0.0,"ee":0.2778},{"task_id":"task_044_comprehensive_decision_medium_medium019","category":"comprehensive_decision","score":0.0,"tokens":907140,"cost_usd":0.239259,"elapsed_time":161.73,"gpr":0.0,"tpe":0.0,"ee":0.1667},{"task_id":"task_045_comprehensive_decision_medium_medium020","category":"comprehensive_decision","score":1.0,"tokens":2560856,"cost_usd":0.67543,"elapsed_time":331.51,"gpr":1.0,"tpe":0.0718,"ee":0.1667},{"task_id":"task_046_comprehensive_decision_medium_medium021","category":"comprehensive_decision","score":0.0,"tokens":1520623,"cost_usd":0.404947,"elapsed_time":403.17,"gpr":0.0,"tpe":0.0,"ee":0.0909},{"task_id":"task_047_comprehensive_decision_medium_medium022","category":"comprehensive_decision","score":1.0,"tokens":21350638,"cost_usd":5.562802,"elapsed_time":854.69,"gpr":1.0,"tpe":0.0002,"ee":0.0575},{"task_id":"task_048_comprehensive_decision_medium_medium023","category":"comprehensive_decision","score":0.0,"tokens":2394345,"cost_usd":0.627319,"elapsed_time":293.59,"gpr":0.25,"tpe":0.0278,"ee":0.1053},{"task_id":"task_049_comprehensive_decision_medium_medium024","category":"comprehensive_decision","score":0.0,"tokens":126625,"cost_usd":0.034308,"elapsed_time":138.63,"gpr":0.0,"tpe":0.0,"ee":0.2381},{"task_id":"task_050_comprehensive_decision_medium_medium025","category":"comprehensive_decision","score":0.0,"tokens":1998520,"cost_usd":0.526343,"elapsed_time":334.97,"gpr":0.0,"tpe":0.0,"ee":0.0926},{"task_id":"task_051_comprehensive_decision_medium_medium026","category":"comprehensive_decision","score":1.0,"tokens":1881420,"cost_usd":0.495629,"elapsed_time":104.95,"gpr":1.0,"tpe":0.5905,"ee":0.5},{"task_id":"task_052_comprehensive_decision_medium_medium027","category":"comprehensive_decision","score":1.0,"tokens":299950,"cost_usd":0.084973,"elapsed_time":195.18,"gpr":1.0,"tpe":0.0718,"ee":0.1667},{"task_id":"task_053_comprehensive_decision_medium_medium028","category":"comprehensive_decision","score":0.0,"tokens":1652201,"cost_usd":0.450142,"elapsed_time":523.63,"gpr":0.0,"tpe":0.0,"ee":0.0658},{"task_id":"task_054_comprehensive_decision_medium_medium029","category":"comprehensive_decision","score":0.0,"tokens":659194,"cost_usd":0.174848,"elapsed_time":122.48,"gpr":0.5,"tpe":0.81,"ee":0.3333},{"task_id":"task_055_comprehensive_decision_medium_medium030","category":"comprehensive_decision","score":0.0,"tokens":1675878,"cost_usd":0.439015,"elapsed_time":258.81,"gpr":0.0,"tpe":0.0,"ee":0.1724},{"task_id":"task_056_comprehensive_decision_medium_medium031","category":"comprehensive_decision","score":1.0,"tokens":319333,"cost_usd":0.085941,"elapsed_time":84.06,"gpr":1.0,"tpe":0.5228,"ee":0.3077},{"task_id":"task_057_comprehensive_decision_medium_medium032","category":"comprehensive_decision","score":0.0,"tokens":34611,"cost_usd":0.009642,"elapsed_time":49.88,"gpr":0.0,"tpe":0.0,"ee":0.7143},{"task_id":"task_058_comprehensive_decision_medium_medium033","category":"comprehensive_decision","score":1.0,"tokens":158154,"cost_usd":0.043944,"elapsed_time":138.11,"gpr":1.0,"tpe":0.218,"ee":0.2381},{"task_id":"task_059_comprehensive_decision_medium_medium034","category":"comprehensive_decision","score":0.0,"tokens":2725930,"cost_usd":0.712795,"elapsed_time":227.6,"gpr":0.0,"tpe":0.0,"ee":0.1724},{"task_id":"task_060_comprehensive_decision_medium_medium035","category":"comprehensive_decision","score":0.0,"tokens":1579862,"cost_usd":0.414262,"elapsed_time":184.35,"gpr":0.0,"tpe":0.0,"ee":0.1923},{"task_id":"task_061_comprehensive_decision_medium_medium036","category":"comprehensive_decision","score":1.0,"tokens":1586317,"cost_usd":0.452941,"elapsed_time":509.02,"gpr":1.0,"tpe":0.001,"ee":0.0685},{"task_id":"task_062_comprehensive_decision_medium_medium037","category":"comprehensive_decision","score":1.0,"tokens":1871364,"cost_usd":0.501886,"elapsed_time":205.46,"gpr":1.0,"tpe":0.3874,"ee":0.3333},{"task_id":"task_063_comprehensive_decision_medium_medium038","category":"comprehensive_decision","score":0.0,"tokens":2268475,"cost_usd":0.599986,"elapsed_time":268.92,"gpr":0.0,"tpe":0.0,"ee":0.129},{"task_id":"task_064_comprehensive_decision_medium_medium039","category":"comprehensive_decision","score":1.0,"tokens":459835,"cost_usd":0.141801,"elapsed_time":204.3,"gpr":1.0,"tpe":0.4264,"ee":0.2353},{"task_id":"task_065_comprehensive_decision_medium_medium040","category":"comprehensive_decision","score":1.0,"tokens":513575,"cost_usd":0.149767,"elapsed_time":173.75,"gpr":1.0,"tpe":0.3811,"ee":0.1923},{"task_id":"task_066_comprehensive_decision_medium_medium041","category":"comprehensive_decision","score":1.0,"tokens":1199941,"cost_usd":0.32864,"elapsed_time":273.54,"gpr":1.0,"tpe":0.0148,"ee":0.1111},{"task_id":"task_067_comprehensive_decision_medium_medium042","category":"comprehensive_decision","score":0.0,"tokens":3130509,"cost_usd":0.827861,"elapsed_time":178.07,"gpr":0.0,"tpe":0.0,"ee":0.1389},{"task_id":"task_068_comprehensive_decision_medium_medium043","category":"comprehensive_decision","score":1.0,"tokens":223724,"cost_usd":0.075943,"elapsed_time":96.24,"gpr":1.0,"tpe":0.4305,"ee":0.3333},{"task_id":"task_069_comprehensive_decision_medium_medium044","category":"comprehensive_decision","score":0.0,"tokens":379682,"cost_usd":0.102002,"elapsed_time":57.3,"gpr":0.0,"tpe":0.0,"ee":0.7143},{"task_id":"task_070_comprehensive_decision_medium_medium045","category":"comprehensive_decision","score":0.0,"tokens":1324637,"cost_usd":0.349329,"elapsed_time":185.32,"gpr":0.0,"tpe":0.0,"ee":0.1471},{"task_id":"task_071_enterprise_industry_analysis_easy_easy001","category":"enterprise_industry_analysis","score":0.0,"tokens":4648111,"cost_usd":1.214736,"elapsed_time":367.71,"gpr":0.0,"tpe":0.0,"ee":0.0536},{"task_id":"task_072_enterprise_industry_analysis_easy_easy002","category":"enterprise_industry_analysis","score":0.0,"tokens":1251108,"cost_usd":0.337031,"elapsed_time":217.79,"gpr":0.25,"tpe":0.4783,"ee":0.0714},{"task_id":"task_073_enterprise_industry_analysis_easy_easy003","category":"enterprise_industry_analysis","score":1.0,"tokens":818650,"cost_usd":0.219392,"elapsed_time":284.71,"gpr":1.0,"tpe":0.0087,"ee":0.0625},{"task_id":"task_074_enterprise_industry_analysis_easy_easy004","category":"enterprise_industry_analysis","score":0.0,"tokens":2273366,"cost_usd":0.596427,"elapsed_time":230.69,"gpr":0.0,"tpe":0.0,"ee":0.0789},{"task_id":"task_075_enterprise_industry_analysis_easy_easy005","category":"enterprise_industry_analysis","score":0.0,"tokens":11368,"cost_usd":0.003282,"elapsed_time":40.51,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_076_enterprise_industry_analysis_easy_easy006","category":"enterprise_industry_analysis","score":1.0,"tokens":149401,"cost_usd":0.041067,"elapsed_time":85.53,"gpr":1.0,"tpe":0.3034,"ee":0.2},{"task_id":"task_077_enterprise_industry_analysis_easy_easy007","category":"enterprise_industry_analysis","score":0.0,"tokens":16585606,"cost_usd":4.326494,"elapsed_time":1207.69,"gpr":0.0,"tpe":0.0,"ee":0.0286},{"task_id":"task_078_enterprise_industry_analysis_easy_easy008","category":"enterprise_industry_analysis","score":0.0,"tokens":1710349,"cost_usd":0.449954,"elapsed_time":266.27,"gpr":0.0,"tpe":0.0,"ee":0.0638},{"task_id":"task_079_enterprise_industry_analysis_easy_easy009","category":"enterprise_industry_analysis","score":1.0,"tokens":312472,"cost_usd":0.084638,"elapsed_time":137.69,"gpr":1.0,"tpe":0.118,"ee":0.1154},{"task_id":"task_080_enterprise_industry_analysis_easy_easy010","category":"enterprise_industry_analysis","score":1.0,"tokens":336400,"cost_usd":0.091441,"elapsed_time":152.9,"gpr":1.0,"tpe":0.0634,"ee":0.0968},{"task_id":"task_081_enterprise_industry_analysis_easy_easy011","category":"enterprise_industry_analysis","score":0.0,"tokens":1500236,"cost_usd":0.393062,"elapsed_time":251.63,"gpr":0.0,"tpe":0.0,"ee":0.0909},{"task_id":"task_082_enterprise_industry_analysis_easy_easy012","category":"enterprise_industry_analysis","score":0.0,"tokens":34962,"cost_usd":0.009621,"elapsed_time":72.53,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_083_enterprise_industry_analysis_easy_easy013","category":"enterprise_industry_analysis","score":0.0,"tokens":11339,"cost_usd":0.003243,"elapsed_time":44.95,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_084_enterprise_industry_analysis_easy_easy014","category":"enterprise_industry_analysis","score":0.0,"tokens":1426749,"cost_usd":0.374249,"elapsed_time":161.47,"gpr":0.0,"tpe":0.0,"ee":0.1071},{"task_id":"task_085_enterprise_industry_analysis_easy_easy015","category":"enterprise_industry_analysis","score":1.0,"tokens":620467,"cost_usd":0.170895,"elapsed_time":316.08,"gpr":1.0,"tpe":0.0261,"ee":0.0577},{"task_id":"task_086_enterprise_industry_analysis_easy_easy016","category":"enterprise_industry_analysis","score":0.0,"tokens":2245393,"cost_usd":0.587315,"elapsed_time":275.0,"gpr":0.0,"tpe":0.0,"ee":0.0968},{"task_id":"task_087_enterprise_industry_analysis_easy_easy017","category":"enterprise_industry_analysis","score":0.0,"tokens":821927,"cost_usd":0.216831,"elapsed_time":187.1,"gpr":0.6667,"tpe":0.5529,"ee":0.0909},{"task_id":"task_088_enterprise_industry_analysis_easy_easy018","category":"enterprise_industry_analysis","score":0.0,"tokens":2805554,"cost_usd":0.735594,"elapsed_time":376.51,"gpr":1.0,"tpe":0.0148,"ee":0.0698},{"task_id":"task_089_enterprise_industry_analysis_easy_easy019","category":"enterprise_industry_analysis","score":1.0,"tokens":1105454,"cost_usd":0.290915,"elapsed_time":141.05,"gpr":1.0,"tpe":0.0886,"ee":0.1154},{"task_id":"task_090_enterprise_industry_analysis_easy_easy020","category":"enterprise_industry_analysis","score":0.0,"tokens":2456489,"cost_usd":0.642796,"elapsed_time":354.37,"gpr":0.0,"tpe":0.0,"ee":0.0652},{"task_id":"task_091_enterprise_industry_analysis_easy_easy021","category":"enterprise_industry_analysis","score":0.0,"tokens":731807,"cost_usd":0.19384,"elapsed_time":217.56,"gpr":0.3333,"tpe":0.1853,"ee":0.0882},{"task_id":"task_092_enterprise_industry_analysis_easy_easy022","category":"enterprise_industry_analysis","score":0.0,"tokens":2662547,"cost_usd":0.698394,"elapsed_time":336.85,"gpr":0.3333,"tpe":0.0798,"ee":0.0682},{"task_id":"task_093_enterprise_industry_analysis_easy_easy023","category":"enterprise_industry_analysis","score":0.0,"tokens":686135,"cost_usd":0.182875,"elapsed_time":211.7,"gpr":0.0,"tpe":0.0,"ee":0.0769},{"task_id":"task_094_enterprise_industry_analysis_easy_easy024","category":"enterprise_industry_analysis","score":0.0,"tokens":2847262,"cost_usd":0.745889,"elapsed_time":274.26,"gpr":0.0,"tpe":0.0,"ee":0.0612},{"task_id":"task_095_enterprise_industry_analysis_easy_easy025","category":"enterprise_industry_analysis","score":0.0,"tokens":959815,"cost_usd":0.253036,"elapsed_time":195.82,"gpr":0.0,"tpe":0.0,"ee":0.0909},{"task_id":"task_096_enterprise_industry_analysis_easy_easy026","category":"enterprise_industry_analysis","score":0.0,"tokens":1746206,"cost_usd":0.459682,"elapsed_time":196.34,"gpr":0.3333,"tpe":0.0225,"ee":0.0652},{"task_id":"task_097_enterprise_industry_analysis_easy_easy027","category":"enterprise_industry_analysis","score":0.0,"tokens":836506,"cost_usd":0.222624,"elapsed_time":138.72,"gpr":0.3333,"tpe":0.0886,"ee":0.1},{"task_id":"task_098_enterprise_industry_analysis_easy_easy028","category":"enterprise_industry_analysis","score":0.0,"tokens":471529,"cost_usd":0.124165,"elapsed_time":73.53,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_099_enterprise_industry_analysis_easy_easy029","category":"enterprise_industry_analysis","score":0.0,"tokens":421380,"cost_usd":0.111359,"elapsed_time":61.53,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_100_enterprise_industry_analysis_easy_easy030","category":"enterprise_industry_analysis","score":0.0,"tokens":1400270,"cost_usd":0.369243,"elapsed_time":134.72,"gpr":0.0,"tpe":0.0,"ee":0.1111},{"task_id":"task_101_enterprise_industry_analysis_easy_easy031","category":"enterprise_industry_analysis","score":0.0,"tokens":713023,"cost_usd":0.187705,"elapsed_time":89.59,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_102_enterprise_industry_analysis_easy_easy032","category":"enterprise_industry_analysis","score":0.0,"tokens":914383,"cost_usd":0.241767,"elapsed_time":146.51,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_103_enterprise_industry_analysis_easy_easy033","category":"enterprise_industry_analysis","score":0.0,"tokens":4768351,"cost_usd":1.246503,"elapsed_time":283.81,"gpr":0.3333,"tpe":0.0471,"ee":0.0938},{"task_id":"task_104_enterprise_industry_analysis_easy_easy034","category":"enterprise_industry_analysis","score":0.0,"tokens":2175365,"cost_usd":0.572956,"elapsed_time":190.31,"gpr":0.0,"tpe":0.0,"ee":0.0938},{"task_id":"task_105_enterprise_industry_analysis_easy_easy035","category":"enterprise_industry_analysis","score":1.0,"tokens":146736,"cost_usd":0.040342,"elapsed_time":58.36,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_106_enterprise_industry_analysis_easy_easy036","category":"enterprise_industry_analysis","score":1.0,"tokens":387081,"cost_usd":0.104475,"elapsed_time":122.3,"gpr":1.0,"tpe":0.1351,"ee":0.1364},{"task_id":"task_107_enterprise_industry_analysis_easy_easy037","category":"enterprise_industry_analysis","score":1.0,"tokens":167767,"cost_usd":0.0462,"elapsed_time":79.67,"gpr":1.0,"tpe":0.2824,"ee":0.2},{"task_id":"task_108_enterprise_industry_analysis_easy_easy038","category":"enterprise_industry_analysis","score":0.0,"tokens":272770,"cost_usd":0.073217,"elapsed_time":85.14,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_109_enterprise_industry_analysis_easy_easy039","category":"enterprise_industry_analysis","score":1.0,"tokens":344679,"cost_usd":0.092045,"elapsed_time":44.43,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_110_enterprise_industry_analysis_easy_easy040","category":"enterprise_industry_analysis","score":1.0,"tokens":148392,"cost_usd":0.041922,"elapsed_time":49.27,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_111_enterprise_industry_analysis_easy_easy041","category":"enterprise_industry_analysis","score":0.0,"tokens":1069952,"cost_usd":0.282905,"elapsed_time":95.22,"gpr":0.3333,"tpe":0.1216,"ee":0.1304},{"task_id":"task_112_enterprise_industry_analysis_easy_easy042","category":"enterprise_industry_analysis","score":0.0,"tokens":1091972,"cost_usd":0.288632,"elapsed_time":191.04,"gpr":0.0,"tpe":0.0,"ee":0.1071},{"task_id":"task_113_enterprise_industry_analysis_easy_easy043","category":"enterprise_industry_analysis","score":1.0,"tokens":224506,"cost_usd":0.061776,"elapsed_time":60.66,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_114_enterprise_industry_analysis_easy_easy044","category":"enterprise_industry_analysis","score":0.0,"tokens":370974,"cost_usd":0.099909,"elapsed_time":86.93,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_115_enterprise_industry_analysis_easy_easy045","category":"enterprise_industry_analysis","score":1.0,"tokens":196826,"cost_usd":0.054819,"elapsed_time":67.44,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_116_enterprise_industry_analysis_easy_easy046","category":"enterprise_industry_analysis","score":1.0,"tokens":807189,"cost_usd":0.215135,"elapsed_time":112.28,"gpr":1.0,"tpe":0.1351,"ee":0.1364},{"task_id":"task_117_enterprise_industry_analysis_easy_easy047","category":"enterprise_industry_analysis","score":0.0,"tokens":195588,"cost_usd":0.054501,"elapsed_time":104.7,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_118_enterprise_industry_analysis_easy_easy048","category":"enterprise_industry_analysis","score":0.0,"tokens":217945,"cost_usd":0.057824,"elapsed_time":42.02,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_119_enterprise_industry_analysis_easy_easy049","category":"enterprise_industry_analysis","score":0.0,"tokens":1729846,"cost_usd":0.452455,"elapsed_time":136.53,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_120_enterprise_industry_analysis_easy_easy050","category":"enterprise_industry_analysis","score":0.0,"tokens":753970,"cost_usd":0.200122,"elapsed_time":147.19,"gpr":0.0,"tpe":0.0,"ee":0.1},{"task_id":"task_121_enterprise_industry_analysis_easy_easy051","category":"enterprise_industry_analysis","score":1.0,"tokens":239509,"cost_usd":0.064663,"elapsed_time":96.35,"gpr":1.0,"tpe":0.1668,"ee":0.15},{"task_id":"task_122_enterprise_industry_analysis_easy_easy052","category":"enterprise_industry_analysis","score":1.0,"tokens":187774,"cost_usd":0.05127,"elapsed_time":84.16,"gpr":1.0,"tpe":0.2288,"ee":0.1765},{"task_id":"task_123_enterprise_industry_analysis_easy_easy053","category":"enterprise_industry_analysis","score":1.0,"tokens":207005,"cost_usd":0.056849,"elapsed_time":69.5,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_124_enterprise_industry_analysis_easy_easy054","category":"enterprise_industry_analysis","score":1.0,"tokens":217519,"cost_usd":0.059421,"elapsed_time":51.76,"gpr":1.0,"tpe":0.2824,"ee":0.2},{"task_id":"task_125_enterprise_industry_analysis_easy_easy055","category":"enterprise_industry_analysis","score":1.0,"tokens":315723,"cost_usd":0.087068,"elapsed_time":133.32,"gpr":1.0,"tpe":0.1216,"ee":0.1304},{"task_id":"task_126_enterprise_industry_analysis_easy_easy056","category":"enterprise_industry_analysis","score":0.0,"tokens":659139,"cost_usd":0.17335,"elapsed_time":74.12,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_127_enterprise_industry_analysis_easy_easy057","category":"enterprise_industry_analysis","score":1.0,"tokens":517820,"cost_usd":0.137294,"elapsed_time":74.7,"gpr":1.0,"tpe":0.3874,"ee":0.25},{"task_id":"task_128_enterprise_industry_analysis_easy_easy058","category":"enterprise_industry_analysis","score":0.0,"tokens":1061292,"cost_usd":0.278656,"elapsed_time":126.65,"gpr":0.0,"tpe":0.0,"ee":0.1304},{"task_id":"task_129_enterprise_industry_analysis_easy_easy059","category":"enterprise_industry_analysis","score":1.0,"tokens":226150,"cost_usd":0.061721,"elapsed_time":62.11,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_130_enterprise_industry_analysis_easy_easy060","category":"enterprise_industry_analysis","score":0.0,"tokens":295444,"cost_usd":0.080091,"elapsed_time":97.46,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_131_enterprise_industry_analysis_easy_easy061","category":"enterprise_industry_analysis","score":0.0,"tokens":401618,"cost_usd":0.108014,"elapsed_time":96.78,"gpr":0.3333,"tpe":0.4783,"ee":0.3},{"task_id":"task_132_enterprise_industry_analysis_easy_easy062","category":"enterprise_industry_analysis","score":1.0,"tokens":278078,"cost_usd":0.075869,"elapsed_time":92.52,"gpr":0.6667,"tpe":0.5797,"ee":0.2143},{"task_id":"task_133_enterprise_industry_analysis_easy_easy063","category":"enterprise_industry_analysis","score":1.0,"tokens":96948,"cost_usd":0.027445,"elapsed_time":43.31,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_134_enterprise_industry_analysis_easy_easy064","category":"enterprise_industry_analysis","score":1.0,"tokens":1538256,"cost_usd":0.403363,"elapsed_time":111.97,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_135_enterprise_industry_analysis_easy_easy065","category":"enterprise_industry_analysis","score":1.0,"tokens":117086,"cost_usd":0.032897,"elapsed_time":50.51,"gpr":1.0,"tpe":0.6901,"ee":0.3333},{"task_id":"task_136_enterprise_industry_analysis_easy_easy066","category":"enterprise_industry_analysis","score":1.0,"tokens":4669450,"cost_usd":1.216349,"elapsed_time":188.93,"gpr":1.0,"tpe":0.2245,"ee":0.1579},{"task_id":"task_137_enterprise_industry_analysis_easy_easy067","category":"enterprise_industry_analysis","score":1.0,"tokens":95755,"cost_usd":0.026663,"elapsed_time":35.03,"gpr":1.0,"tpe":0.7317,"ee":0.4286},{"task_id":"task_138_enterprise_industry_analysis_easy_easy068","category":"enterprise_industry_analysis","score":0.0,"tokens":663117,"cost_usd":0.174318,"elapsed_time":75.08,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_139_enterprise_industry_analysis_easy_easy069","category":"enterprise_industry_analysis","score":0.0,"tokens":1356609,"cost_usd":0.358389,"elapsed_time":159.78,"gpr":1.0,"tpe":0.1356,"ee":0.1304},{"task_id":"task_140_enterprise_industry_analysis_easy_easy070","category":"enterprise_industry_analysis","score":0.0,"tokens":137686,"cost_usd":0.039128,"elapsed_time":69.4,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_141_enterprise_industry_analysis_easy_easy071","category":"enterprise_industry_analysis","score":1.0,"tokens":168088,"cost_usd":0.045993,"elapsed_time":63.15,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_142_enterprise_industry_analysis_easy_easy072","category":"enterprise_industry_analysis","score":1.0,"tokens":257932,"cost_usd":0.070528,"elapsed_time":84.46,"gpr":1.0,"tpe":0.6052,"ee":0.2727},{"task_id":"task_143_enterprise_industry_analysis_easy_easy073","category":"enterprise_industry_analysis","score":0.0,"tokens":281092,"cost_usd":0.090027,"elapsed_time":278.66,"gpr":0.3333,"tpe":0.4305,"ee":0.2},{"task_id":"task_144_enterprise_industry_analysis_easy_easy074","category":"enterprise_industry_analysis","score":1.0,"tokens":91496,"cost_usd":0.02628,"elapsed_time":49.56,"gpr":1.0,"tpe":0.6804,"ee":0.4286},{"task_id":"task_145_enterprise_industry_analysis_easy_easy075","category":"enterprise_industry_analysis","score":1.0,"tokens":466643,"cost_usd":0.122599,"elapsed_time":66.6,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_146_enterprise_industry_analysis_easy_easy076","category":"enterprise_industry_analysis","score":1.0,"tokens":685236,"cost_usd":0.179282,"elapsed_time":49.03,"gpr":1.0,"tpe":0.87,"ee":0.6},{"task_id":"task_147_enterprise_industry_analysis_easy_easy077","category":"enterprise_industry_analysis","score":1.0,"tokens":104064,"cost_usd":0.028375,"elapsed_time":36.58,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_148_enterprise_industry_analysis_easy_easy078","category":"enterprise_industry_analysis","score":1.0,"tokens":63466,"cost_usd":0.017883,"elapsed_time":26.1,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_149_enterprise_industry_analysis_easy_easy079","category":"enterprise_industry_analysis","score":1.0,"tokens":81366,"cost_usd":0.022554,"elapsed_time":36.1,"gpr":1.0,"tpe":0.756,"ee":0.5},{"task_id":"task_150_enterprise_industry_analysis_easy_easy080","category":"enterprise_industry_analysis","score":1.0,"tokens":83955,"cost_usd":0.023032,"elapsed_time":28.53,"gpr":1.0,"tpe":0.756,"ee":0.5},{"task_id":"task_151_enterprise_industry_analysis_easy_easy081","category":"enterprise_industry_analysis","score":1.0,"tokens":79159,"cost_usd":0.021832,"elapsed_time":30.81,"gpr":1.0,"tpe":0.9033,"ee":0.6},{"task_id":"task_152_enterprise_industry_analysis_easy_easy082","category":"enterprise_industry_analysis","score":1.0,"tokens":144516,"cost_usd":0.039386,"elapsed_time":34.68,"gpr":1.0,"tpe":0.8854,"ee":0.4286},{"task_id":"task_153_enterprise_industry_analysis_easy_easy083","category":"enterprise_industry_analysis","score":1.0,"tokens":1368521,"cost_usd":0.358035,"elapsed_time":72.21,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_154_enterprise_industry_analysis_easy_easy084","category":"enterprise_industry_analysis","score":0.0,"tokens":309163,"cost_usd":0.089681,"elapsed_time":133.23,"gpr":0.6667,"tpe":0.4365,"ee":0.1875},{"task_id":"task_155_enterprise_industry_analysis_easy_easy085","category":"enterprise_industry_analysis","score":1.0,"tokens":107205,"cost_usd":0.02932,"elapsed_time":48.66,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_156_enterprise_industry_analysis_easy_easy086","category":"enterprise_industry_analysis","score":1.0,"tokens":712641,"cost_usd":0.188335,"elapsed_time":79.93,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_157_enterprise_industry_analysis_easy_easy088","category":"enterprise_industry_analysis","score":1.0,"tokens":119595,"cost_usd":0.032816,"elapsed_time":29.65,"gpr":1.0,"tpe":0.81,"ee":0.5},{"task_id":"task_158_enterprise_industry_analysis_easy_easy090","category":"enterprise_industry_analysis","score":0.0,"tokens":274790,"cost_usd":0.077063,"elapsed_time":76.88,"gpr":0.6667,"tpe":0.561,"ee":0.2143},{"task_id":"task_159_enterprise_industry_analysis_easy_easy091","category":"enterprise_industry_analysis","score":0.0,"tokens":1425814,"cost_usd":0.37307,"elapsed_time":94.83,"gpr":0.0,"tpe":0.0,"ee":0.1667},{"task_id":"task_160_enterprise_industry_analysis_easy_easy092","category":"enterprise_industry_analysis","score":0.0,"tokens":267347,"cost_usd":0.071198,"elapsed_time":41.22,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_161_enterprise_industry_analysis_easy_easy093","category":"enterprise_industry_analysis","score":0.0,"tokens":436385,"cost_usd":0.118837,"elapsed_time":108.37,"gpr":0.0,"tpe":0.0,"ee":0.1111},{"task_id":"task_162_enterprise_industry_analysis_easy_easy094","category":"enterprise_industry_analysis","score":0.0,"tokens":687706,"cost_usd":0.183471,"elapsed_time":131.48,"gpr":0.0,"tpe":0.0,"ee":0.1071},{"task_id":"task_163_enterprise_industry_analysis_easy_easy095","category":"enterprise_industry_analysis","score":1.0,"tokens":502237,"cost_usd":0.134254,"elapsed_time":103.86,"gpr":1.0,"tpe":0.2824,"ee":0.2},{"task_id":"task_164_enterprise_industry_analysis_easy_easy096","category":"enterprise_industry_analysis","score":1.0,"tokens":411518,"cost_usd":0.110701,"elapsed_time":99.04,"gpr":1.0,"tpe":0.1501,"ee":0.1429},{"task_id":"task_165_enterprise_industry_analysis_easy_easy097","category":"enterprise_industry_analysis","score":0.0,"tokens":2316562,"cost_usd":0.60575,"elapsed_time":147.98,"gpr":0.0,"tpe":0.0,"ee":0.1765},{"task_id":"task_166_enterprise_industry_analysis_easy_easy098","category":"enterprise_industry_analysis","score":1.0,"tokens":676149,"cost_usd":0.180497,"elapsed_time":96.07,"gpr":1.0,"tpe":0.1351,"ee":0.1364},{"task_id":"task_167_enterprise_industry_analysis_easy_easy099","category":"enterprise_industry_analysis","score":0.0,"tokens":242492,"cost_usd":0.06571,"elapsed_time":59.2,"gpr":0.0,"tpe":0.0,"ee":0.1875},{"task_id":"task_168_enterprise_industry_analysis_easy_easy100","category":"enterprise_industry_analysis","score":0.0,"tokens":829723,"cost_usd":0.218303,"elapsed_time":58.83,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_169_enterprise_industry_analysis_easy_easy101","category":"enterprise_industry_analysis","score":0.0,"tokens":629024,"cost_usd":0.165976,"elapsed_time":73.17,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_170_enterprise_industry_analysis_easy_easy102","category":"enterprise_industry_analysis","score":1.0,"tokens":179663,"cost_usd":0.049593,"elapsed_time":58.57,"gpr":1.0,"tpe":0.3138,"ee":0.2143},{"task_id":"task_171_enterprise_industry_analysis_easy_easy103","category":"enterprise_industry_analysis","score":1.0,"tokens":283423,"cost_usd":0.07603,"elapsed_time":64.6,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_172_enterprise_industry_analysis_easy_easy104","category":"enterprise_industry_analysis","score":0.0,"tokens":1141835,"cost_usd":0.299459,"elapsed_time":99.97,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_173_enterprise_industry_analysis_easy_easy105","category":"enterprise_industry_analysis","score":0.0,"tokens":554462,"cost_usd":0.147398,"elapsed_time":84.78,"gpr":0.0,"tpe":0.0,"ee":0.1765},{"task_id":"task_174_enterprise_industry_analysis_easy_easy106","category":"enterprise_industry_analysis","score":0.0,"tokens":396335,"cost_usd":0.105702,"elapsed_time":69.24,"gpr":0.0,"tpe":0.0,"ee":0.1765},{"task_id":"task_175_enterprise_industry_analysis_easy_easy107","category":"enterprise_industry_analysis","score":0.0,"tokens":2883370,"cost_usd":0.753492,"elapsed_time":196.96,"gpr":0.0,"tpe":0.0,"ee":0.1111},{"task_id":"task_176_enterprise_industry_analysis_easy_easy108","category":"enterprise_industry_analysis","score":1.0,"tokens":1971639,"cost_usd":0.516643,"elapsed_time":150.17,"gpr":1.0,"tpe":0.1216,"ee":0.1304},{"task_id":"task_177_enterprise_industry_analysis_easy_easy109","category":"enterprise_industry_analysis","score":0.0,"tokens":439475,"cost_usd":0.116828,"elapsed_time":62.95,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_178_enterprise_industry_analysis_easy_easy110","category":"enterprise_industry_analysis","score":1.0,"tokens":62980,"cost_usd":0.017618,"elapsed_time":27.66,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_179_enterprise_industry_analysis_easy_easy111","category":"enterprise_industry_analysis","score":0.0,"tokens":744678,"cost_usd":0.196481,"elapsed_time":132.06,"gpr":0.0,"tpe":0.0,"ee":0.1364},{"task_id":"task_180_enterprise_industry_analysis_easy_easy112","category":"enterprise_industry_analysis","score":0.0,"tokens":193287,"cost_usd":0.052812,"elapsed_time":97.42,"gpr":0.0,"tpe":0.0,"ee":0.15},{"task_id":"task_181_enterprise_industry_analysis_easy_easy113","category":"enterprise_industry_analysis","score":0.0,"tokens":161034,"cost_usd":0.043776,"elapsed_time":66.07,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_182_enterprise_industry_analysis_easy_easy114","category":"enterprise_industry_analysis","score":0.0,"tokens":512682,"cost_usd":0.135143,"elapsed_time":56.65,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_183_enterprise_industry_analysis_easy_easy115","category":"enterprise_industry_analysis","score":0.0,"tokens":673684,"cost_usd":0.179118,"elapsed_time":56.73,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_184_enterprise_industry_analysis_easy_easy116","category":"enterprise_industry_analysis","score":0.0,"tokens":657344,"cost_usd":0.172814,"elapsed_time":67.94,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_185_enterprise_industry_analysis_easy_easy117","category":"enterprise_industry_analysis","score":1.0,"tokens":172701,"cost_usd":0.046149,"elapsed_time":35.19,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_186_enterprise_industry_analysis_medium_medium001","category":"enterprise_industry_analysis","score":1.0,"tokens":626328,"cost_usd":0.168538,"elapsed_time":98.21,"gpr":1.0,"tpe":0.0798,"ee":0.1429},{"task_id":"task_187_enterprise_industry_analysis_medium_medium002","category":"enterprise_industry_analysis","score":1.0,"tokens":800139,"cost_usd":0.214619,"elapsed_time":161.13,"gpr":1.0,"tpe":0.1113,"ee":0.093},{"task_id":"task_188_enterprise_industry_analysis_medium_medium003","category":"enterprise_industry_analysis","score":0.0,"tokens":304812,"cost_usd":0.086864,"elapsed_time":81.36,"gpr":0.75,"tpe":0.5157,"ee":0.25},{"task_id":"task_189_enterprise_industry_analysis_medium_medium004","category":"enterprise_industry_analysis","score":0.0,"tokens":3336704,"cost_usd":0.878269,"elapsed_time":216.74,"gpr":0.25,"tpe":0.0148,"ee":0.0909},{"task_id":"task_190_enterprise_industry_analysis_medium_medium005","category":"enterprise_industry_analysis","score":0.0,"tokens":1256160,"cost_usd":0.329322,"elapsed_time":108.36,"gpr":0.0,"tpe":0.0,"ee":0.2353},{"task_id":"task_191_enterprise_industry_analysis_medium_medium006","category":"enterprise_industry_analysis","score":1.0,"tokens":1392969,"cost_usd":0.366324,"elapsed_time":127.55,"gpr":1.0,"tpe":0.1094,"ee":0.16},{"task_id":"task_192_enterprise_industry_analysis_medium_medium007","category":"enterprise_industry_analysis","score":0.0,"tokens":3893850,"cost_usd":1.016397,"elapsed_time":215.88,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_193_enterprise_industry_analysis_medium_medium008","category":"enterprise_industry_analysis","score":0.0,"tokens":1872083,"cost_usd":0.491354,"elapsed_time":136.71,"gpr":0.2,"tpe":0.2059,"ee":0.16},{"task_id":"task_194_enterprise_industry_analysis_medium_medium009","category":"enterprise_industry_analysis","score":1.0,"tokens":2025189,"cost_usd":0.54165,"elapsed_time":240.32,"gpr":1.0,"tpe":0.1886,"ee":0.0851},{"task_id":"task_195_enterprise_industry_analysis_medium_medium010","category":"enterprise_industry_analysis","score":0.0,"tokens":234048,"cost_usd":0.064563,"elapsed_time":148.55,"gpr":0.0,"tpe":0.0,"ee":0.2857},{"task_id":"task_196_enterprise_industry_analysis_medium_medium011","category":"enterprise_industry_analysis","score":0.0,"tokens":2816573,"cost_usd":0.735796,"elapsed_time":176.35,"gpr":0.0,"tpe":0.0,"ee":0.1667},{"task_id":"task_197_enterprise_industry_analysis_medium_medium012","category":"enterprise_industry_analysis","score":0.0,"tokens":548373,"cost_usd":0.144068,"elapsed_time":70.01,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_198_enterprise_industry_analysis_medium_medium013","category":"enterprise_industry_analysis","score":0.0,"tokens":838131,"cost_usd":0.223704,"elapsed_time":155.32,"gpr":0.0,"tpe":0.0,"ee":0.1026},{"task_id":"task_199_enterprise_industry_analysis_medium_medium014","category":"enterprise_industry_analysis","score":1.0,"tokens":552576,"cost_usd":0.151801,"elapsed_time":111.42,"gpr":1.0,"tpe":0.2916,"ee":0.2},{"task_id":"task_200_enterprise_industry_analysis_medium_medium015","category":"enterprise_industry_analysis","score":1.0,"tokens":686387,"cost_usd":0.182326,"elapsed_time":73.21,"gpr":1.0,"tpe":0.2059,"ee":0.25},{"task_id":"task_201_enterprise_industry_analysis_medium_medium016","category":"enterprise_industry_analysis","score":1.0,"tokens":465746,"cost_usd":0.127395,"elapsed_time":83.56,"gpr":1.0,"tpe":0.3868,"ee":0.2941},{"task_id":"task_202_enterprise_industry_analysis_medium_medium017","category":"enterprise_industry_analysis","score":1.0,"tokens":907020,"cost_usd":0.243313,"elapsed_time":162.95,"gpr":1.0,"tpe":0.0972,"ee":0.1724},{"task_id":"task_203_enterprise_industry_analysis_medium_medium018","category":"enterprise_industry_analysis","score":0.0,"tokens":1035969,"cost_usd":0.27193,"elapsed_time":98.69,"gpr":0.0,"tpe":0.0,"ee":0.2778},{"task_id":"task_204_enterprise_industry_analysis_medium_medium019","category":"enterprise_industry_analysis","score":1.0,"tokens":241702,"cost_usd":0.068083,"elapsed_time":107.28,"gpr":1.0,"tpe":0.2288,"ee":0.2632},{"task_id":"task_205_enterprise_industry_analysis_medium_medium020","category":"enterprise_industry_analysis","score":0.0,"tokens":269241,"cost_usd":0.076072,"elapsed_time":95.31,"gpr":0.4,"tpe":0.5049,"ee":0.3125},{"task_id":"task_206_enterprise_industry_analysis_medium_medium021","category":"enterprise_industry_analysis","score":0.0,"tokens":395591,"cost_usd":0.113646,"elapsed_time":120.11,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_207_enterprise_industry_analysis_medium_medium022","category":"enterprise_industry_analysis","score":0.0,"tokens":828610,"cost_usd":0.224656,"elapsed_time":180.5,"gpr":0.0,"tpe":0.0,"ee":0.119},{"task_id":"task_208_enterprise_industry_analysis_medium_medium023","category":"enterprise_industry_analysis","score":0.0,"tokens":2085335,"cost_usd":0.552113,"elapsed_time":167.38,"gpr":0.0,"tpe":0.0,"ee":0.1351},{"task_id":"task_209_enterprise_industry_analysis_medium_medium024","category":"enterprise_industry_analysis","score":0.0,"tokens":1605049,"cost_usd":0.429127,"elapsed_time":167.79,"gpr":0.8,"tpe":0.3204,"ee":0.1282},{"task_id":"task_210_enterprise_industry_analysis_medium_medium025","category":"enterprise_industry_analysis","score":1.0,"tokens":841296,"cost_usd":0.224331,"elapsed_time":89.8,"gpr":1.0,"tpe":0.4828,"ee":0.2941},{"task_id":"task_211_enterprise_industry_analysis_medium_medium026","category":"enterprise_industry_analysis","score":0.0,"tokens":1997888,"cost_usd":0.525526,"elapsed_time":198.68,"gpr":0.4,"tpe":0.0382,"ee":0.1282},{"task_id":"task_212_enterprise_industry_analysis_medium_medium027","category":"enterprise_industry_analysis","score":1.0,"tokens":359859,"cost_usd":0.099552,"elapsed_time":135.15,"gpr":1.0,"tpe":0.1232,"ee":0.1852},{"task_id":"task_213_enterprise_industry_analysis_medium_medium028","category":"enterprise_industry_analysis","score":1.0,"tokens":557907,"cost_usd":0.148767,"elapsed_time":102.52,"gpr":1.0,"tpe":0.318,"ee":0.2381},{"task_id":"task_214_enterprise_industry_analysis_medium_medium029","category":"enterprise_industry_analysis","score":0.0,"tokens":2608273,"cost_usd":0.68258,"elapsed_time":133.21,"gpr":0.0,"tpe":0.0,"ee":0.3571},{"task_id":"task_215_enterprise_industry_analysis_medium_medium030","category":"enterprise_industry_analysis","score":1.0,"tokens":976826,"cost_usd":0.264346,"elapsed_time":155.77,"gpr":1.0,"tpe":0.0343,"ee":0.1351},{"task_id":"task_216_enterprise_industry_analysis_medium_medium031","category":"enterprise_industry_analysis","score":1.0,"tokens":702474,"cost_usd":0.195265,"elapsed_time":141.08,"gpr":1.0,"tpe":0.1771,"ee":0.1429},{"task_id":"task_217_enterprise_industry_analysis_medium_medium032","category":"enterprise_industry_analysis","score":1.0,"tokens":211901,"cost_usd":0.061074,"elapsed_time":58.59,"gpr":1.0,"tpe":0.682,"ee":0.4167},{"task_id":"task_218_enterprise_industry_analysis_medium_medium033","category":"enterprise_industry_analysis","score":1.0,"tokens":166944,"cost_usd":0.047386,"elapsed_time":49.91,"gpr":1.0,"tpe":0.7543,"ee":0.5},{"task_id":"task_219_enterprise_industry_analysis_medium_medium034","category":"enterprise_industry_analysis","score":1.0,"tokens":524025,"cost_usd":0.143069,"elapsed_time":89.56,"gpr":1.0,"tpe":0.2059,"ee":0.25},{"task_id":"task_220_enterprise_industry_analysis_medium_medium035","category":"enterprise_industry_analysis","score":1.0,"tokens":976865,"cost_usd":0.268059,"elapsed_time":142.25,"gpr":1.0,"tpe":0.3969,"ee":0.1923},{"task_id":"task_221_enterprise_industry_analysis_medium_medium036","category":"enterprise_industry_analysis","score":0.0,"tokens":881331,"cost_usd":0.246314,"elapsed_time":200.17,"gpr":0.6,"tpe":0.245,"ee":0.1316},{"task_id":"task_222_enterprise_industry_analysis_medium_medium037","category":"enterprise_industry_analysis","score":0.0,"tokens":792334,"cost_usd":0.211788,"elapsed_time":94.74,"gpr":0.2,"tpe":1.0,"ee":0.25},{"task_id":"task_223_enterprise_industry_analysis_medium_medium038","category":"enterprise_industry_analysis","score":1.0,"tokens":342140,"cost_usd":0.097501,"elapsed_time":84.28,"gpr":1.0,"tpe":0.428,"ee":0.3571},{"task_id":"task_224_enterprise_industry_analysis_medium_medium039","category":"enterprise_industry_analysis","score":1.0,"tokens":306918,"cost_usd":0.086742,"elapsed_time":106.9,"gpr":1.0,"tpe":0.3746,"ee":0.2174},{"task_id":"task_225_enterprise_industry_analysis_medium_medium040","category":"enterprise_industry_analysis","score":0.0,"tokens":1074154,"cost_usd":0.290058,"elapsed_time":159.28,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_226_enterprise_industry_analysis_medium_medium041","category":"enterprise_industry_analysis","score":1.0,"tokens":1215047,"cost_usd":0.328318,"elapsed_time":110.76,"gpr":0.2,"tpe":0.1501,"ee":0.2632},{"task_id":"task_227_enterprise_industry_analysis_medium_medium042","category":"enterprise_industry_analysis","score":0.0,"tokens":1144257,"cost_usd":0.307725,"elapsed_time":177.44,"gpr":0.8,"tpe":0.6246,"ee":0.1282},{"task_id":"task_228_enterprise_industry_analysis_medium_medium043","category":"enterprise_industry_analysis","score":0.0,"tokens":468224,"cost_usd":0.126668,"elapsed_time":84.45,"gpr":1.0,"tpe":0.7805,"ee":0.3571},{"task_id":"task_229_enterprise_industry_analysis_medium_medium044","category":"enterprise_industry_analysis","score":0.0,"tokens":176628,"cost_usd":0.052256,"elapsed_time":58.59,"gpr":0.2,"tpe":1.0,"ee":0.5556},{"task_id":"task_230_enterprise_industry_analysis_medium_medium045","category":"enterprise_industry_analysis","score":0.0,"tokens":1465301,"cost_usd":0.404173,"elapsed_time":297.49,"gpr":0.4,"tpe":0.95,"ee":0.125},{"task_id":"task_231_enterprise_industry_analysis_medium_medium046","category":"enterprise_industry_analysis","score":1.0,"tokens":474829,"cost_usd":0.133492,"elapsed_time":117.1,"gpr":1.0,"tpe":0.2059,"ee":0.25},{"task_id":"task_232_enterprise_industry_analysis_medium_medium047","category":"enterprise_industry_analysis","score":1.0,"tokens":7702689,"cost_usd":2.015066,"elapsed_time":374.93,"gpr":1.0,"tpe":0.0068,"ee":0.0943},{"task_id":"task_233_enterprise_industry_analysis_medium_medium048","category":"enterprise_industry_analysis","score":1.0,"tokens":681828,"cost_usd":0.186245,"elapsed_time":124.73,"gpr":1.0,"tpe":0.4459,"ee":0.2273},{"task_id":"task_234_enterprise_industry_analysis_medium_medium049","category":"enterprise_industry_analysis","score":1.0,"tokens":404826,"cost_usd":0.111321,"elapsed_time":98.6,"gpr":1.0,"tpe":0.6646,"ee":0.2778},{"task_id":"task_235_enterprise_industry_analysis_medium_medium050","category":"enterprise_industry_analysis","score":1.0,"tokens":401210,"cost_usd":0.116173,"elapsed_time":156.39,"gpr":1.0,"tpe":0.1501,"ee":0.1852},{"task_id":"task_236_enterprise_industry_analysis_medium_medium051","category":"enterprise_industry_analysis","score":0.0,"tokens":868364,"cost_usd":0.238318,"elapsed_time":198.57,"gpr":0.6,"tpe":0.5113,"ee":0.2083},{"task_id":"task_237_enterprise_industry_analysis_medium_medium052","category":"enterprise_industry_analysis","score":1.0,"tokens":390440,"cost_usd":0.107894,"elapsed_time":202.09,"gpr":1.0,"tpe":0.126,"ee":0.1613},{"task_id":"task_238_enterprise_industry_analysis_medium_medium053","category":"enterprise_industry_analysis","score":0.0,"tokens":342495,"cost_usd":0.095624,"elapsed_time":118.65,"gpr":0.4,"tpe":0.6561,"ee":0.2632},{"task_id":"task_239_enterprise_industry_analysis_medium_medium054","category":"enterprise_industry_analysis","score":1.0,"tokens":975179,"cost_usd":0.261373,"elapsed_time":154.1,"gpr":1.0,"tpe":0.1539,"ee":0.1429},{"task_id":"task_240_enterprise_industry_analysis_medium_medium055","category":"enterprise_industry_analysis","score":1.0,"tokens":746268,"cost_usd":0.200437,"elapsed_time":84.63,"gpr":1.0,"tpe":0.5591,"ee":0.3333},{"task_id":"task_241_enterprise_industry_analysis_medium_medium056","category":"enterprise_industry_analysis","score":1.0,"tokens":578379,"cost_usd":0.158238,"elapsed_time":103.89,"gpr":1.0,"tpe":0.2458,"ee":0.2632},{"task_id":"task_242_enterprise_industry_analysis_medium_medium057","category":"enterprise_industry_analysis","score":1.0,"tokens":4097474,"cost_usd":1.07073,"elapsed_time":217.94,"gpr":1.0,"tpe":0.134,"ee":0.1786},{"task_id":"task_243_enterprise_industry_analysis_medium_medium058","category":"enterprise_industry_analysis","score":0.0,"tokens":1261771,"cost_usd":0.34251,"elapsed_time":107.95,"gpr":0.4,"tpe":0.2442,"ee":0.25},{"task_id":"task_244_enterprise_industry_analysis_medium_medium059","category":"enterprise_industry_analysis","score":1.0,"tokens":684228,"cost_usd":0.183704,"elapsed_time":110.64,"gpr":1.0,"tpe":0.2502,"ee":0.2632},{"task_id":"task_245_enterprise_industry_analysis_medium_medium060","category":"enterprise_industry_analysis","score":0.0,"tokens":2170675,"cost_usd":0.569264,"elapsed_time":169.5,"gpr":0.0,"tpe":0.0,"ee":0.1923},{"task_id":"task_246_enterprise_industry_analysis_medium_medium061","category":"enterprise_industry_analysis","score":0.0,"tokens":1269538,"cost_usd":0.333997,"elapsed_time":103.13,"gpr":0.2,"tpe":0.4783,"ee":0.2778},{"task_id":"task_247_enterprise_industry_analysis_medium_medium062","category":"enterprise_industry_analysis","score":0.0,"tokens":1016501,"cost_usd":0.270052,"elapsed_time":155.51,"gpr":0.0,"tpe":0.0,"ee":0.1786},{"task_id":"task_248_enterprise_industry_analysis_medium_medium063","category":"enterprise_industry_analysis","score":1.0,"tokens":1105724,"cost_usd":0.298612,"elapsed_time":180.28,"gpr":1.0,"tpe":0.2051,"ee":0.1333},{"task_id":"task_249_enterprise_industry_analysis_medium_medium064","category":"enterprise_industry_analysis","score":1.0,"tokens":908474,"cost_usd":0.241806,"elapsed_time":119.52,"gpr":1.0,"tpe":0.383,"ee":0.1667},{"task_id":"task_250_enterprise_industry_analysis_medium_medium065","category":"enterprise_industry_analysis","score":1.0,"tokens":172175,"cost_usd":0.047697,"elapsed_time":57.22,"gpr":1.0,"tpe":0.6197,"ee":0.3333},{"task_id":"task_251_enterprise_industry_analysis_medium_medium066","category":"enterprise_industry_analysis","score":1.0,"tokens":477969,"cost_usd":0.127831,"elapsed_time":61.58,"gpr":1.0,"tpe":0.5905,"ee":0.4444},{"task_id":"task_252_enterprise_industry_analysis_medium_medium067","category":"enterprise_industry_analysis","score":0.0,"tokens":1022889,"cost_usd":0.270912,"elapsed_time":116.34,"gpr":0.0,"tpe":0.0,"ee":0.1667},{"task_id":"task_253_enterprise_industry_analysis_medium_medium068","category":"enterprise_industry_analysis","score":1.0,"tokens":1191872,"cost_usd":0.317624,"elapsed_time":110.56,"gpr":1.0,"tpe":0.2256,"ee":0.2},{"task_id":"task_254_enterprise_industry_analysis_medium_medium069","category":"enterprise_industry_analysis","score":1.0,"tokens":1630736,"cost_usd":0.430845,"elapsed_time":187.68,"gpr":1.0,"tpe":0.1351,"ee":0.1739},{"task_id":"task_255_enterprise_industry_analysis_medium_medium070","category":"enterprise_industry_analysis","score":0.0,"tokens":220921,"cost_usd":0.062894,"elapsed_time":62.02,"gpr":0.3333,"tpe":0.5905,"ee":0.3077},{"task_id":"task_256_enterprise_industry_analysis_medium_medium071","category":"enterprise_industry_analysis","score":0.0,"tokens":1330630,"cost_usd":0.352182,"elapsed_time":182.59,"gpr":0.0,"tpe":0.0,"ee":0.1053},{"task_id":"task_257_enterprise_industry_analysis_medium_medium072","category":"enterprise_industry_analysis","score":0.0,"tokens":1331234,"cost_usd":0.355418,"elapsed_time":212.65,"gpr":0.25,"tpe":0.1094,"ee":0.0889},{"task_id":"task_258_enterprise_industry_analysis_medium_medium073","category":"enterprise_industry_analysis","score":0.0,"tokens":1512477,"cost_usd":0.397136,"elapsed_time":94.32,"gpr":0.25,"tpe":1.0,"ee":0.2222},{"task_id":"task_259_enterprise_industry_analysis_medium_medium074","category":"enterprise_industry_analysis","score":1.0,"tokens":1659768,"cost_usd":0.437206,"elapsed_time":121.83,"gpr":1.0,"tpe":0.1905,"ee":0.1538},{"task_id":"task_260_enterprise_industry_analysis_medium_medium075","category":"enterprise_industry_analysis","score":1.0,"tokens":1109193,"cost_usd":0.304566,"elapsed_time":278.87,"gpr":1.0,"tpe":0.0723,"ee":0.1053},{"task_id":"task_261_enterprise_industry_analysis_medium_medium076","category":"enterprise_industry_analysis","score":0.0,"tokens":1487961,"cost_usd":0.389973,"elapsed_time":104.6,"gpr":0.5,"tpe":0.3879,"ee":0.1905},{"task_id":"task_262_enterprise_industry_analysis_medium_medium077","category":"enterprise_industry_analysis","score":0.0,"tokens":534772,"cost_usd":0.145182,"elapsed_time":125.4,"gpr":1.0,"tpe":0.14,"ee":0.1481},{"task_id":"task_263_enterprise_industry_analysis_medium_medium078","category":"enterprise_industry_analysis","score":0.0,"tokens":2374388,"cost_usd":0.624373,"elapsed_time":249.44,"gpr":0.25,"tpe":0.2824,"ee":0.08},{"task_id":"task_264_enterprise_industry_analysis_medium_medium079","category":"enterprise_industry_analysis","score":0.0,"tokens":3841810,"cost_usd":1.005688,"elapsed_time":230.26,"gpr":0.25,"tpe":0.1668,"ee":0.1333},{"task_id":"task_265_enterprise_industry_analysis_medium_medium080","category":"enterprise_industry_analysis","score":0.0,"tokens":754162,"cost_usd":0.204914,"elapsed_time":171.91,"gpr":0.0,"tpe":0.0,"ee":0.1212},{"task_id":"task_266_enterprise_industry_analysis_medium_medium081","category":"enterprise_industry_analysis","score":1.0,"tokens":166323,"cost_usd":0.046022,"elapsed_time":61.66,"gpr":1.0,"tpe":0.7197,"ee":0.3636},{"task_id":"task_267_enterprise_industry_analysis_medium_medium082","category":"enterprise_industry_analysis","score":1.0,"tokens":1025188,"cost_usd":0.275155,"elapsed_time":266.37,"gpr":1.0,"tpe":0.0278,"ee":0.1053},{"task_id":"task_268_enterprise_industry_analysis_medium_medium083","category":"enterprise_industry_analysis","score":0.0,"tokens":2117309,"cost_usd":0.554883,"elapsed_time":138.15,"gpr":0.25,"tpe":0.2288,"ee":0.1379},{"task_id":"task_269_enterprise_industry_analysis_medium_medium084","category":"enterprise_industry_analysis","score":1.0,"tokens":586316,"cost_usd":0.160909,"elapsed_time":174.39,"gpr":1.0,"tpe":0.1923,"ee":0.1739},{"task_id":"task_270_enterprise_industry_analysis_medium_medium085","category":"enterprise_industry_analysis","score":1.0,"tokens":896262,"cost_usd":0.238226,"elapsed_time":137.03,"gpr":1.0,"tpe":0.3144,"ee":0.1905},{"task_id":"task_271_enterprise_industry_analysis_medium_medium086","category":"enterprise_industry_analysis","score":1.0,"tokens":1128912,"cost_usd":0.295983,"elapsed_time":90.78,"gpr":1.0,"tpe":0.6268,"ee":0.4},{"task_id":"task_272_enterprise_industry_analysis_medium_medium087","category":"enterprise_industry_analysis","score":0.0,"tokens":535787,"cost_usd":0.146633,"elapsed_time":158.18,"gpr":0.0,"tpe":0.0,"ee":0.2105},{"task_id":"task_273_enterprise_industry_analysis_medium_medium088","category":"enterprise_industry_analysis","score":1.0,"tokens":741620,"cost_usd":0.199341,"elapsed_time":143.0,"gpr":1.0,"tpe":0.1705,"ee":0.1739},{"task_id":"task_274_enterprise_industry_analysis_medium_medium089","category":"enterprise_industry_analysis","score":1.0,"tokens":360837,"cost_usd":0.097858,"elapsed_time":82.23,"gpr":1.0,"tpe":0.4638,"ee":0.2},{"task_id":"task_275_enterprise_industry_analysis_medium_medium090","category":"enterprise_industry_analysis","score":0.0,"tokens":5819942,"cost_usd":1.521835,"elapsed_time":274.25,"gpr":0.5,"tpe":0.0375,"ee":0.0976},{"task_id":"task_276_enterprise_industry_analysis_medium_medium091","category":"enterprise_industry_analysis","score":1.0,"tokens":302693,"cost_usd":0.081652,"elapsed_time":67.46,"gpr":1.0,"tpe":0.4929,"ee":0.3333},{"task_id":"task_277_enterprise_industry_analysis_medium_medium092","category":"enterprise_industry_analysis","score":1.0,"tokens":376600,"cost_usd":0.101669,"elapsed_time":94.71,"gpr":1.0,"tpe":0.284,"ee":0.2353},{"task_id":"task_278_enterprise_industry_analysis_medium_medium093","category":"enterprise_industry_analysis","score":0.0,"tokens":1176163,"cost_usd":0.308547,"elapsed_time":117.23,"gpr":0.0,"tpe":0.0,"ee":0.2667},{"task_id":"task_279_enterprise_industry_analysis_medium_medium094","category":"enterprise_industry_analysis","score":0.0,"tokens":813899,"cost_usd":0.215756,"elapsed_time":90.18,"gpr":0.25,"tpe":0.3487,"ee":0.2222},{"task_id":"task_280_enterprise_industry_analysis_medium_medium095","category":"enterprise_industry_analysis","score":1.0,"tokens":1577801,"cost_usd":0.418323,"elapsed_time":170.82,"gpr":1.0,"tpe":0.1153,"ee":0.1429},{"task_id":"task_281_enterprise_industry_analysis_medium_medium096","category":"enterprise_industry_analysis","score":1.0,"tokens":294014,"cost_usd":0.080788,"elapsed_time":139.72,"gpr":1.0,"tpe":0.7024,"ee":0.16},{"task_id":"task_282_enterprise_industry_analysis_medium_medium097","category":"enterprise_industry_analysis","score":1.0,"tokens":192634,"cost_usd":0.05338,"elapsed_time":102.33,"gpr":1.0,"tpe":0.38,"ee":0.2222},{"task_id":"task_283_enterprise_industry_analysis_medium_medium098","category":"enterprise_industry_analysis","score":0.0,"tokens":2224773,"cost_usd":0.58378,"elapsed_time":379.33,"gpr":0.0,"tpe":0.0,"ee":0.0645},{"task_id":"task_284_enterprise_industry_analysis_medium_medium099","category":"enterprise_industry_analysis","score":0.0,"tokens":531076,"cost_usd":0.140759,"elapsed_time":110.08,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_285_enterprise_industry_analysis_medium_medium100","category":"enterprise_industry_analysis","score":1.0,"tokens":326075,"cost_usd":0.088287,"elapsed_time":192.85,"gpr":1.0,"tpe":0.1193,"ee":0.1379},{"task_id":"task_286_enterprise_industry_analysis_medium_medium101","category":"enterprise_industry_analysis","score":1.0,"tokens":405512,"cost_usd":0.111307,"elapsed_time":215.96,"gpr":1.0,"tpe":0.2044,"ee":0.125},{"task_id":"task_287_enterprise_industry_analysis_medium_medium102","category":"enterprise_industry_analysis","score":0.0,"tokens":248982,"cost_usd":0.067381,"elapsed_time":183.23,"gpr":0.25,"tpe":0.1668,"ee":0.1429},{"task_id":"task_288_enterprise_industry_analysis_medium_medium103","category":"enterprise_industry_analysis","score":0.0,"tokens":123564,"cost_usd":0.033988,"elapsed_time":153.57,"gpr":0.0,"tpe":0.0,"ee":0.1739},{"task_id":"task_289_enterprise_industry_analysis_medium_medium104","category":"enterprise_industry_analysis","score":0.0,"tokens":34666,"cost_usd":0.009528,"elapsed_time":46.74,"gpr":0.0,"tpe":0.0,"ee":0.5714},{"task_id":"task_290_enterprise_industry_analysis_medium_medium105","category":"enterprise_industry_analysis","score":0.0,"tokens":1113116,"cost_usd":0.293159,"elapsed_time":197.77,"gpr":0.25,"tpe":0.1216,"ee":0.1379},{"task_id":"task_291_enterprise_industry_analysis_medium_medium106","category":"enterprise_industry_analysis","score":0.0,"tokens":356738,"cost_usd":0.096134,"elapsed_time":137.68,"gpr":0.25,"tpe":0.2542,"ee":0.1739},{"task_id":"task_292_enterprise_industry_analysis_medium_medium107","category":"enterprise_industry_analysis","score":0.0,"tokens":325118,"cost_usd":0.087976,"elapsed_time":142.77,"gpr":0.25,"tpe":0.2059,"ee":0.16},{"task_id":"task_293_enterprise_industry_analysis_medium_medium108","category":"enterprise_industry_analysis","score":0.0,"tokens":34606,"cost_usd":0.009515,"elapsed_time":75.58,"gpr":0.25,"tpe":0.81,"ee":0.4},{"task_id":"task_294_enterprise_industry_analysis_medium_medium109","category":"enterprise_industry_analysis","score":1.0,"tokens":188518,"cost_usd":0.052025,"elapsed_time":101.45,"gpr":1.0,"tpe":0.2174,"ee":0.2},{"task_id":"task_295_enterprise_industry_analysis_medium_medium110","category":"enterprise_industry_analysis","score":0.0,"tokens":979925,"cost_usd":0.258185,"elapsed_time":115.11,"gpr":0.25,"tpe":0.2059,"ee":0.1739},{"task_id":"task_296_enterprise_industry_analysis_medium_medium111","category":"enterprise_industry_analysis","score":0.0,"tokens":630495,"cost_usd":0.169498,"elapsed_time":140.09,"gpr":0.0,"tpe":0.0,"ee":0.1053},{"task_id":"task_297_enterprise_industry_policy_analysis_easy_easy067","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":36512,"cost_usd":0.010207,"elapsed_time":25.91,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_298_enterprise_industry_policy_analysis_easy_easy068","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":231271,"cost_usd":0.062379,"elapsed_time":41.9,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_299_enterprise_industry_policy_analysis_easy_easy069","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":157145,"cost_usd":0.045306,"elapsed_time":79.11,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_300_enterprise_industry_policy_analysis_easy_easy070","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":891496,"cost_usd":0.235542,"elapsed_time":110.31,"gpr":1.0,"tpe":0.1853,"ee":0.1579},{"task_id":"task_301_enterprise_industry_policy_analysis_easy_easy071","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":990310,"cost_usd":0.260862,"elapsed_time":129.34,"gpr":1.0,"tpe":0.1351,"ee":0.1364},{"task_id":"task_302_enterprise_industry_policy_analysis_easy_easy072","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":553904,"cost_usd":0.145621,"elapsed_time":80.42,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_303_enterprise_industry_policy_analysis_easy_easy073","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":303028,"cost_usd":0.080497,"elapsed_time":65.57,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_304_enterprise_industry_policy_analysis_easy_easy074","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":255713,"cost_usd":0.069382,"elapsed_time":79.11,"gpr":0.0,"tpe":0.0,"ee":0.15},{"task_id":"task_305_enterprise_industry_policy_analysis_easy_easy075","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":98712,"cost_usd":0.027533,"elapsed_time":46.85,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_306_enterprise_industry_policy_analysis_easy_easy076","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":274443,"cost_usd":0.073023,"elapsed_time":65.61,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_307_enterprise_industry_policy_analysis_medium_medium001","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":3427564,"cost_usd":0.897784,"elapsed_time":242.64,"gpr":1.0,"tpe":0.3103,"ee":0.1667},{"task_id":"task_308_enterprise_industry_policy_analysis_medium_medium002","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2057175,"cost_usd":0.536922,"elapsed_time":131.75,"gpr":0.8,"tpe":0.6707,"ee":0.2941},{"task_id":"task_309_enterprise_industry_policy_analysis_medium_medium003","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":5221468,"cost_usd":1.365824,"elapsed_time":222.26,"gpr":0.8,"tpe":0.336,"ee":0.1724},{"task_id":"task_310_enterprise_industry_policy_analysis_medium_medium004","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":610652,"cost_usd":0.161831,"elapsed_time":90.25,"gpr":0.0,"tpe":0.0,"ee":0.3571},{"task_id":"task_311_enterprise_industry_policy_analysis_medium_medium005","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":4122191,"cost_usd":1.080976,"elapsed_time":261.66,"gpr":1.0,"tpe":0.0949,"ee":0.1562},{"task_id":"task_312_enterprise_industry_policy_analysis_medium_medium006","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":4898937,"cost_usd":1.279399,"elapsed_time":401.05,"gpr":0.0,"tpe":0.0,"ee":0.0909},{"task_id":"task_313_enterprise_industry_policy_analysis_medium_medium007","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1134753,"cost_usd":0.306721,"elapsed_time":490.52,"gpr":1.0,"tpe":0.0791,"ee":0.0568},{"task_id":"task_314_enterprise_industry_policy_analysis_medium_medium008","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1438500,"cost_usd":0.379571,"elapsed_time":199.73,"gpr":1.0,"tpe":0.0527,"ee":0.1429},{"task_id":"task_315_enterprise_industry_policy_analysis_medium_medium009","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":6137712,"cost_usd":1.603042,"elapsed_time":630.2,"gpr":0.4,"tpe":0.0023,"ee":0.0617},{"task_id":"task_316_enterprise_industry_policy_analysis_medium_medium010","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":952114,"cost_usd":0.253548,"elapsed_time":320.2,"gpr":0.0,"tpe":0.0,"ee":0.1},{"task_id":"task_317_enterprise_industry_policy_analysis_medium_medium011","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":962020,"cost_usd":0.257078,"elapsed_time":249.35,"gpr":1.0,"tpe":0.0523,"ee":0.1515},{"task_id":"task_318_enterprise_industry_policy_analysis_medium_medium012","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1623234,"cost_usd":0.424069,"elapsed_time":153.77,"gpr":0.8,"tpe":0.4783,"ee":0.2941},{"task_id":"task_319_enterprise_industry_policy_analysis_medium_medium013","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":11293,"cost_usd":0.003118,"elapsed_time":40.85,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_320_enterprise_industry_policy_analysis_medium_medium014","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":524695,"cost_usd":0.144121,"elapsed_time":250.95,"gpr":1.0,"tpe":0.1706,"ee":0.1282},{"task_id":"task_321_enterprise_industry_policy_analysis_medium_medium015","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":968269,"cost_usd":0.259976,"elapsed_time":404.42,"gpr":0.6,"tpe":0.0268,"ee":0.0862},{"task_id":"task_322_enterprise_industry_policy_analysis_medium_medium016","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":2039350,"cost_usd":0.536584,"elapsed_time":263.23,"gpr":1.0,"tpe":0.0164,"ee":0.1136},{"task_id":"task_323_enterprise_industry_policy_analysis_medium_medium017","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":512372,"cost_usd":0.143762,"elapsed_time":253.68,"gpr":0.8,"tpe":0.0607,"ee":0.1087},{"task_id":"task_324_enterprise_industry_policy_analysis_medium_medium018","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":633656,"cost_usd":0.172028,"elapsed_time":342.38,"gpr":1.0,"tpe":0.1355,"ee":0.0847},{"task_id":"task_325_enterprise_industry_policy_analysis_medium_medium019","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":195625,"cost_usd":0.054497,"elapsed_time":160.62,"gpr":1.0,"tpe":0.3282,"ee":0.1923},{"task_id":"task_326_enterprise_industry_policy_analysis_medium_medium020","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":3867276,"cost_usd":1.011574,"elapsed_time":452.24,"gpr":0.6,"tpe":0.2168,"ee":0.0893},{"task_id":"task_327_enterprise_industry_policy_analysis_medium_medium021","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":129986,"cost_usd":0.035063,"elapsed_time":95.16,"gpr":0.0,"tpe":0.0,"ee":0.3125},{"task_id":"task_328_enterprise_industry_policy_analysis_medium_medium022","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":914649,"cost_usd":0.245425,"elapsed_time":303.5,"gpr":0.2,"tpe":0.1668,"ee":0.1042},{"task_id":"task_329_enterprise_industry_policy_analysis_medium_medium023","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":713332,"cost_usd":0.193746,"elapsed_time":227.59,"gpr":1.0,"tpe":0.1414,"ee":0.1724},{"task_id":"task_330_enterprise_industry_policy_analysis_medium_medium024","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":5271293,"cost_usd":1.379769,"elapsed_time":349.67,"gpr":0.4,"tpe":0.0416,"ee":0.1136},{"task_id":"task_331_enterprise_industry_policy_analysis_medium_medium025","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":983609,"cost_usd":0.266158,"elapsed_time":215.67,"gpr":1.0,"tpe":0.3761,"ee":0.1667},{"task_id":"task_332_enterprise_industry_policy_analysis_medium_medium026","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1873066,"cost_usd":0.492378,"elapsed_time":125.89,"gpr":1.0,"tpe":0.5007,"ee":0.3571},{"task_id":"task_333_enterprise_industry_policy_analysis_medium_medium027","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1014398,"cost_usd":0.271247,"elapsed_time":285.41,"gpr":0.0,"tpe":0.0,"ee":0.102},{"task_id":"task_334_enterprise_industry_policy_analysis_medium_medium028","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1157299,"cost_usd":0.3061,"elapsed_time":167.51,"gpr":0.4,"tpe":0.1155,"ee":0.1613},{"task_id":"task_335_enterprise_industry_policy_analysis_medium_medium029","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":2267264,"cost_usd":0.592635,"elapsed_time":200.13,"gpr":0.4,"tpe":0.2981,"ee":0.2083},{"task_id":"task_336_enterprise_industry_policy_analysis_medium_medium030","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2386509,"cost_usd":0.623893,"elapsed_time":161.56,"gpr":0.4,"tpe":0.5905,"ee":0.25},{"task_id":"task_337_enterprise_industry_policy_analysis_medium_medium031","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1783359,"cost_usd":0.46882,"elapsed_time":120.41,"gpr":1.0,"tpe":0.4762,"ee":0.2857},{"task_id":"task_338_enterprise_industry_policy_analysis_medium_medium032","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2636352,"cost_usd":0.690336,"elapsed_time":288.89,"gpr":0.0,"tpe":0.0,"ee":0.0851},{"task_id":"task_339_enterprise_industry_policy_analysis_medium_medium033","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":760752,"cost_usd":0.201853,"elapsed_time":151.69,"gpr":1.0,"tpe":0.1823,"ee":0.1905},{"task_id":"task_340_enterprise_industry_policy_analysis_medium_medium034","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1294640,"cost_usd":0.340807,"elapsed_time":88.03,"gpr":0.5,"tpe":0.729,"ee":0.5714},{"task_id":"task_341_enterprise_industry_policy_analysis_medium_medium035","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":540007,"cost_usd":0.14483,"elapsed_time":197.42,"gpr":1.0,"tpe":0.0631,"ee":0.1176},{"task_id":"task_342_enterprise_industry_policy_analysis_medium_medium036","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2015416,"cost_usd":0.529664,"elapsed_time":237.24,"gpr":0.5,"tpe":0.0906,"ee":0.1081},{"task_id":"task_343_enterprise_industry_policy_analysis_medium_medium037","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1518475,"cost_usd":0.401479,"elapsed_time":172.12,"gpr":0.75,"tpe":0.6561,"ee":0.1481},{"task_id":"task_344_enterprise_industry_policy_analysis_medium_medium038","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":741120,"cost_usd":0.201236,"elapsed_time":279.03,"gpr":0.5,"tpe":0.1959,"ee":0.0851},{"task_id":"task_345_enterprise_industry_policy_analysis_medium_medium039","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1700263,"cost_usd":0.447375,"elapsed_time":214.25,"gpr":0.0,"tpe":0.0,"ee":0.1481},{"task_id":"task_346_enterprise_industry_policy_analysis_medium_medium040","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":2281414,"cost_usd":0.600311,"elapsed_time":212.43,"gpr":1.0,"tpe":0.0301,"ee":0.1053},{"task_id":"task_347_enterprise_industry_policy_analysis_medium_medium041","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":4224124,"cost_usd":1.105712,"elapsed_time":208.97,"gpr":0.75,"tpe":0.2772,"ee":0.1538},{"task_id":"task_348_enterprise_industry_policy_analysis_medium_medium042","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":555954,"cost_usd":0.150865,"elapsed_time":175.83,"gpr":0.0,"tpe":0.0,"ee":0.1481},{"task_id":"task_349_enterprise_industry_policy_analysis_medium_medium043","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1302250,"cost_usd":0.344997,"elapsed_time":286.55,"gpr":0.0,"tpe":0.0,"ee":0.087},{"task_id":"task_350_enterprise_industry_policy_analysis_medium_medium044","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2047507,"cost_usd":0.538372,"elapsed_time":226.09,"gpr":0.0,"tpe":0.0,"ee":0.1026},{"task_id":"task_351_enterprise_industry_policy_analysis_medium_medium045","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":223759,"cost_usd":0.061202,"elapsed_time":98.16,"gpr":0.0,"tpe":0.0,"ee":0.2667},{"task_id":"task_352_enterprise_industry_policy_analysis_medium_medium046","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":322820,"cost_usd":0.088226,"elapsed_time":179.98,"gpr":0.0,"tpe":0.0,"ee":0.129},{"task_id":"task_353_enterprise_industry_policy_analysis_medium_medium047","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":4742818,"cost_usd":1.240839,"elapsed_time":287.78,"gpr":0.25,"tpe":0.1668,"ee":0.1026},{"task_id":"task_354_enterprise_industry_policy_analysis_medium_medium048","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1844895,"cost_usd":0.488402,"elapsed_time":241.76,"gpr":1.0,"tpe":0.1346,"ee":0.1053},{"task_id":"task_355_enterprise_industry_policy_analysis_medium_medium049","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":264166,"cost_usd":0.07327,"elapsed_time":147.4,"gpr":1.0,"tpe":0.4835,"ee":0.1481},{"task_id":"task_356_enterprise_industry_policy_analysis_medium_medium050","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":76304,"cost_usd":0.021064,"elapsed_time":89.58,"gpr":0.75,"tpe":1.0,"ee":0.2857},{"task_id":"task_357_enterprise_industry_policy_analysis_medium_medium051","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1014245,"cost_usd":0.266761,"elapsed_time":134.2,"gpr":0.0,"tpe":0.0,"ee":0.2353},{"task_id":"task_358_enterprise_industry_policy_analysis_medium_medium052","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":6320899,"cost_usd":1.648287,"elapsed_time":285.55,"gpr":0.25,"tpe":1.0,"ee":0.1081},{"task_id":"task_359_enterprise_industry_policy_analysis_medium_medium053","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2586462,"cost_usd":0.678248,"elapsed_time":306.13,"gpr":1.0,"tpe":0.0382,"ee":0.1143},{"task_id":"task_360_enterprise_industry_policy_analysis_medium_medium054","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":4271840,"cost_usd":1.115358,"elapsed_time":159.93,"gpr":0.5,"tpe":0.4395,"ee":0.1818},{"task_id":"task_361_enterprise_industry_policy_analysis_medium_medium055","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":418651,"cost_usd":0.116547,"elapsed_time":241.11,"gpr":1.0,"tpe":0.0262,"ee":0.093},{"task_id":"task_362_enterprise_industry_policy_analysis_medium_medium056","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1086800,"cost_usd":0.297903,"elapsed_time":342.35,"gpr":0.0,"tpe":0.0,"ee":0.0909},{"task_id":"task_363_enterprise_industry_policy_analysis_medium_medium057","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":5636084,"cost_usd":1.470525,"elapsed_time":255.44,"gpr":0.25,"tpe":0.2059,"ee":0.2105},{"task_id":"task_364_enterprise_industry_policy_analysis_medium_medium058","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":360088,"cost_usd":0.100024,"elapsed_time":83.9,"gpr":0.0,"tpe":0.0,"ee":0.2353},{"task_id":"task_365_enterprise_industry_policy_analysis_medium_medium059","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1574121,"cost_usd":0.415251,"elapsed_time":100.39,"gpr":0.0,"tpe":0.0,"ee":0.1905},{"task_id":"task_366_enterprise_industry_policy_analysis_medium_medium060","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":896647,"cost_usd":0.235202,"elapsed_time":83.98,"gpr":0.0,"tpe":0.0,"ee":0.3077},{"task_id":"task_367_enterprise_industry_policy_analysis_medium_medium061","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":497891,"cost_usd":0.132711,"elapsed_time":138.49,"gpr":0.5,"tpe":0.81,"ee":0.1818},{"task_id":"task_368_enterprise_industry_policy_analysis_medium_medium062","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2662643,"cost_usd":0.701556,"elapsed_time":274.23,"gpr":1.0,"tpe":0.3762,"ee":0.1081},{"task_id":"task_369_enterprise_industry_policy_analysis_medium_medium063","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":313408,"cost_usd":0.082787,"elapsed_time":37.8,"gpr":0.0,"tpe":0.0,"ee":1.3333},{"task_id":"task_370_enterprise_industry_policy_analysis_medium_medium064","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":985868,"cost_usd":0.260301,"elapsed_time":138.98,"gpr":0.5,"tpe":1.0,"ee":0.1538},{"task_id":"task_371_enterprise_industry_policy_analysis_medium_medium065","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":8829210,"cost_usd":2.30838,"elapsed_time":264.75,"gpr":0.0,"tpe":0.0,"ee":0.1111},{"task_id":"task_372_enterprise_industry_policy_analysis_medium_medium066","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":968926,"cost_usd":0.254259,"elapsed_time":55.77,"gpr":0.0,"tpe":0.0,"ee":0.6667},{"task_id":"task_373_hypothesis_verification_hard_hard001","category":"hypothesis_verification","score":0.0,"tokens":521217,"cost_usd":0.150321,"elapsed_time":203.5,"gpr":0.3333,"tpe":0.8261,"ee":0.25},{"task_id":"task_374_hypothesis_verification_hard_hard002","category":"hypothesis_verification","score":0.0,"tokens":507283,"cost_usd":0.144432,"elapsed_time":229.17,"gpr":0.375,"tpe":0.3487,"ee":0.1667},{"task_id":"task_375_hypothesis_verification_hard_hard003","category":"hypothesis_verification","score":0.0,"tokens":2225385,"cost_usd":0.588617,"elapsed_time":308.24,"gpr":0.1111,"tpe":0.0022,"ee":0.1194},{"task_id":"task_376_hypothesis_verification_hard_hard004","category":"hypothesis_verification","score":0.0,"tokens":2120876,"cost_usd":0.570405,"elapsed_time":378.31,"gpr":0.3,"tpe":0.0069,"ee":0.0959},{"task_id":"task_377_hypothesis_verification_hard_hard005","category":"hypothesis_verification","score":0.0,"tokens":914279,"cost_usd":0.260333,"elapsed_time":287.4,"gpr":0.0,"tpe":0.0,"ee":0.2195},{"task_id":"task_378_hypothesis_verification_hard_hard006","category":"hypothesis_verification","score":0.0,"tokens":2336057,"cost_usd":0.614741,"elapsed_time":200.06,"gpr":0.3636,"tpe":0.6561,"ee":0.1556},{"task_id":"task_379_hypothesis_verification_hard_hard007","category":"hypothesis_verification","score":0.0,"tokens":2034975,"cost_usd":0.534093,"elapsed_time":151.69,"gpr":0.0,"tpe":0.0,"ee":0.36},{"task_id":"task_380_hypothesis_verification_hard_hard008","category":"hypothesis_verification","score":0.0,"tokens":2183128,"cost_usd":0.603118,"elapsed_time":733.93,"gpr":0.5556,"tpe":0.1509,"ee":0.0722},{"task_id":"task_381_hypothesis_verification_hard_hard009","category":"hypothesis_verification","score":0.0,"tokens":953016,"cost_usd":0.263414,"elapsed_time":277.38,"gpr":0.1111,"tpe":0.1094,"ee":0.1818},{"task_id":"task_382_hypothesis_verification_hard_hard010","category":"hypothesis_verification","score":0.0,"tokens":4432077,"cost_usd":1.167412,"elapsed_time":349.23,"gpr":0.0,"tpe":0.0,"ee":0.1},{"task_id":"task_383_hypothesis_verification_hard_hard011","category":"hypothesis_verification","score":0.0,"tokens":50769746,"cost_usd":13.216805,"elapsed_time":1207.37,"gpr":0.0769,"tpe":1.0,"ee":0.0422},{"task_id":"task_384_hypothesis_verification_hard_hard012","category":"hypothesis_verification","score":0.0,"tokens":3228058,"cost_usd":0.857972,"elapsed_time":629.62,"gpr":0.2727,"tpe":0.0087,"ee":0.0748},{"task_id":"task_385_hypothesis_verification_hard_hard013","category":"hypothesis_verification","score":0.0,"tokens":1555962,"cost_usd":0.42709,"elapsed_time":370.94,"gpr":0.0,"tpe":0.0,"ee":0.1489},{"task_id":"task_386_hypothesis_verification_hard_hard014","category":"hypothesis_verification","score":0.0,"tokens":1377359,"cost_usd":0.380798,"elapsed_time":498.38,"gpr":0.25,"tpe":0.0482,"ee":0.0984},{"task_id":"task_387_hypothesis_verification_hard_hard015","category":"hypothesis_verification","score":0.0,"tokens":1120259,"cost_usd":0.310988,"elapsed_time":359.87,"gpr":0.0,"tpe":0.0,"ee":0.1842},{"task_id":"task_388_hypothesis_verification_medium_medium001","category":"hypothesis_verification","score":0.0,"tokens":1649840,"cost_usd":0.452605,"elapsed_time":604.57,"gpr":0.0,"tpe":0.0,"ee":0.0959},{"task_id":"task_389_hypothesis_verification_medium_medium002","category":"hypothesis_verification","score":0.0,"tokens":500671,"cost_usd":0.142896,"elapsed_time":181.94,"gpr":0.0,"tpe":0.0,"ee":0.2258},{"task_id":"task_390_hypothesis_verification_medium_medium003","category":"hypothesis_verification","score":0.0,"tokens":2233470,"cost_usd":0.599486,"elapsed_time":295.07,"gpr":0.0,"tpe":0.0,"ee":0.1935},{"task_id":"task_391_hypothesis_verification_medium_medium004","category":"hypothesis_verification","score":1.0,"tokens":149178,"cost_usd":0.042412,"elapsed_time":104.67,"gpr":1.0,"tpe":0.5905,"ee":0.5},{"task_id":"task_392_hypothesis_verification_medium_medium005","category":"hypothesis_verification","score":0.0,"tokens":5731878,"cost_usd":1.509312,"elapsed_time":644.27,"gpr":0.2,"tpe":0.0985,"ee":0.1045},{"task_id":"task_393_hypothesis_verification_medium_medium006","category":"hypothesis_verification","score":0.0,"tokens":588000,"cost_usd":0.165023,"elapsed_time":286.58,"gpr":0.4,"tpe":0.5314,"ee":0.1818},{"task_id":"task_394_hypothesis_verification_medium_medium007","category":"hypothesis_verification","score":0.0,"tokens":2965644,"cost_usd":0.792465,"elapsed_time":629.69,"gpr":0.0,"tpe":0.0,"ee":0.0824},{"task_id":"task_395_hypothesis_verification_medium_medium008","category":"hypothesis_verification","score":0.0,"tokens":833542,"cost_usd":0.236689,"elapsed_time":345.83,"gpr":0.2,"tpe":0.5314,"ee":0.1739},{"task_id":"task_396_hypothesis_verification_medium_medium009","category":"hypothesis_verification","score":1.0,"tokens":731442,"cost_usd":0.209356,"elapsed_time":288.84,"gpr":1.0,"tpe":0.1638,"ee":0.1818},{"task_id":"task_397_hypothesis_verification_medium_medium010","category":"hypothesis_verification","score":1.0,"tokens":934580,"cost_usd":0.261296,"elapsed_time":405.55,"gpr":1.0,"tpe":0.1577,"ee":0.1591},{"task_id":"task_398_hypothesis_verification_medium_medium011","category":"hypothesis_verification","score":1.0,"tokens":2986166,"cost_usd":0.811211,"elapsed_time":661.52,"gpr":1.0,"tpe":0.004,"ee":0.0723},{"task_id":"task_399_hypothesis_verification_medium_medium012","category":"hypothesis_verification","score":1.0,"tokens":747758,"cost_usd":0.216973,"elapsed_time":379.67,"gpr":0.0,"tpe":0.0,"ee":0.15},{"task_id":"task_400_hypothesis_verification_medium_medium013","category":"hypothesis_verification","score":0.0,"tokens":1393249,"cost_usd":0.380488,"elapsed_time":319.05,"gpr":0.8333,"tpe":0.0355,"ee":0.1273},{"task_id":"task_401_hypothesis_verification_medium_medium014","category":"hypothesis_verification","score":1.0,"tokens":1638485,"cost_usd":0.44983,"elapsed_time":254.86,"gpr":1.0,"tpe":0.0676,"ee":0.2353},{"task_id":"task_402_industry_planning_hard_hard001","category":"industry_planning","score":1.0,"tokens":1684235,"cost_usd":0.45372,"elapsed_time":166.96,"gpr":1.0,"tpe":0.2828,"ee":0.2258},{"task_id":"task_403_industry_planning_hard_hard002","category":"industry_planning","score":0.0,"tokens":1195096,"cost_usd":0.32785,"elapsed_time":381.41,"gpr":0.1667,"tpe":0.905,"ee":0.1296},{"task_id":"task_404_industry_planning_hard_hard003","category":"industry_planning","score":0.0,"tokens":1238397,"cost_usd":0.329473,"elapsed_time":186.26,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_405_industry_planning_hard_hard004","category":"industry_planning","score":0.0,"tokens":10296987,"cost_usd":2.692745,"elapsed_time":390.58,"gpr":0.4167,"tpe":0.2427,"ee":0.1224},{"task_id":"task_406_industry_planning_hard_hard005","category":"industry_planning","score":0.0,"tokens":13721363,"cost_usd":3.57794,"elapsed_time":741.18,"gpr":0.2222,"tpe":0.1835,"ee":0.0632},{"task_id":"task_407_industry_planning_hard_hard006","category":"industry_planning","score":0.0,"tokens":1555467,"cost_usd":0.412689,"elapsed_time":532.65,"gpr":0.0,"tpe":0.0,"ee":0.0897},{"task_id":"task_408_industry_planning_hard_hard007","category":"industry_planning","score":0.0,"tokens":12040132,"cost_usd":3.15462,"elapsed_time":631.74,"gpr":0.5,"tpe":0.4905,"ee":0.0822},{"task_id":"task_409_industry_planning_hard_hard008","category":"industry_planning","score":0.0,"tokens":93232,"cost_usd":0.025991,"elapsed_time":95.59,"gpr":0.375,"tpe":0.8193,"ee":0.4286},{"task_id":"task_410_industry_planning_hard_hard009","category":"industry_planning","score":0.0,"tokens":1257914,"cost_usd":0.329695,"elapsed_time":159.83,"gpr":1.0,"tpe":0.6915,"ee":0.3158},{"task_id":"task_411_industry_planning_hard_hard010","category":"industry_planning","score":0.0,"tokens":36259,"cost_usd":0.010146,"elapsed_time":71.53,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_412_industry_planning_hard_hard011","category":"industry_planning","score":0.5,"tokens":1435816,"cost_usd":0.389223,"elapsed_time":415.52,"gpr":0.5,"tpe":0.285,"ee":0.12},{"task_id":"task_413_industry_planning_hard_hard012","category":"industry_planning","score":0.0,"tokens":1354662,"cost_usd":0.362797,"elapsed_time":495.24,"gpr":0.5556,"tpe":0.4293,"ee":0.0938},{"task_id":"task_414_industry_planning_hard_hard013","category":"industry_planning","score":0.0,"tokens":32179405,"cost_usd":8.377062,"elapsed_time":1207.37,"gpr":0.0,"tpe":0.0,"ee":0.0385},{"task_id":"task_415_industry_planning_hard_hard014","category":"industry_planning","score":0.0,"tokens":1009069,"cost_usd":0.26887,"elapsed_time":357.57,"gpr":0.625,"tpe":0.3165,"ee":0.0962},{"task_id":"task_416_industry_planning_medium_medium001","category":"industry_planning","score":1.0,"tokens":5075563,"cost_usd":1.331771,"elapsed_time":266.68,"gpr":1.0,"tpe":0.169,"ee":0.2581},{"task_id":"task_417_industry_planning_medium_medium002","category":"industry_planning","score":0.0,"tokens":6092378,"cost_usd":1.589134,"elapsed_time":456.83,"gpr":0.4,"tpe":0.0182,"ee":0.1379},{"task_id":"task_418_industry_planning_medium_medium003","category":"industry_planning","score":0.0,"tokens":788768,"cost_usd":0.222201,"elapsed_time":460.21,"gpr":0.1429,"tpe":0.0203,"ee":0.1127},{"task_id":"task_419_industry_planning_medium_medium004","category":"industry_planning","score":1.0,"tokens":5561278,"cost_usd":1.454031,"elapsed_time":246.64,"gpr":1.0,"tpe":0.3482,"ee":0.3043},{"task_id":"task_420_industry_planning_medium_medium005","category":"industry_planning","score":0.0,"tokens":1108760,"cost_usd":0.304073,"elapsed_time":419.21,"gpr":0.0,"tpe":0.0,"ee":0.1228},{"task_id":"task_421_industry_planning_medium_medium006","category":"industry_planning","score":0.0,"tokens":1814780,"cost_usd":0.48271,"elapsed_time":303.62,"gpr":0.0,"tpe":0.0,"ee":0.1404},{"task_id":"task_422_industry_planning_medium_medium007","category":"industry_planning","score":1.0,"tokens":1957696,"cost_usd":0.542092,"elapsed_time":718.57,"gpr":1.0,"tpe":0.0003,"ee":0.093},{"task_id":"task_423_industry_planning_medium_medium008","category":"industry_planning","score":0.0,"tokens":1510063,"cost_usd":0.404488,"elapsed_time":366.55,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_424_industry_planning_medium_medium009","category":"industry_planning","score":0.0,"tokens":1778430,"cost_usd":0.492955,"elapsed_time":546.92,"gpr":0.1,"tpe":0.0278,"ee":0.1379},{"task_id":"task_425_industry_planning_medium_medium010","category":"industry_planning","score":1.0,"tokens":2185916,"cost_usd":0.578413,"elapsed_time":290.43,"gpr":1.0,"tpe":0.0889,"ee":0.2353},{"task_id":"task_426_industry_planning_medium_medium011","category":"industry_planning","score":0.0,"tokens":2284034,"cost_usd":0.622511,"elapsed_time":517.82,"gpr":0.5833,"tpe":0.0282,"ee":0.127},{"task_id":"task_427_industry_planning_medium_medium012","category":"industry_planning","score":0.0,"tokens":1773804,"cost_usd":0.480746,"elapsed_time":449.13,"gpr":0.4444,"tpe":0.0702,"ee":0.1509},{"task_id":"task_428_industry_planning_medium_medium013","category":"industry_planning","score":1.0,"tokens":1338289,"cost_usd":0.377876,"elapsed_time":500.33,"gpr":1.0,"tpe":0.0611,"ee":0.1404},{"task_id":"task_429_industry_planning_medium_medium014","category":"industry_planning","score":0.0,"tokens":11672197,"cost_usd":3.042454,"elapsed_time":430.03,"gpr":0.3333,"tpe":0.0042,"ee":0.1475},{"task_id":"task_430_international_comparison_hard_hard001","category":"international_comparison","score":0.0,"tokens":1658664,"cost_usd":0.445337,"elapsed_time":214.54,"gpr":0.2,"tpe":0.4783,"ee":0.3333},{"task_id":"task_431_international_comparison_hard_hard002","category":"international_comparison","score":0.0,"tokens":550925,"cost_usd":0.147433,"elapsed_time":147.73,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_432_international_comparison_hard_hard003","category":"international_comparison","score":0.0,"tokens":636424,"cost_usd":0.171319,"elapsed_time":145.35,"gpr":0.2222,"tpe":1.0,"ee":0.4286},{"task_id":"task_433_international_comparison_hard_hard004","category":"international_comparison","score":0.0,"tokens":817498,"cost_usd":0.22156,"elapsed_time":163.32,"gpr":0.0909,"tpe":0.3487,"ee":0.375},{"task_id":"task_434_international_comparison_hard_hard005","category":"international_comparison","score":0.0,"tokens":492783,"cost_usd":0.134026,"elapsed_time":118.18,"gpr":0.4545,"tpe":0.5314,"ee":0.5},{"task_id":"task_435_international_comparison_hard_hard006","category":"international_comparison","score":0.0,"tokens":1305099,"cost_usd":0.353773,"elapsed_time":165.73,"gpr":0.0,"tpe":0.0,"ee":0.3684},{"task_id":"task_436_international_comparison_hard_hard007","category":"international_comparison","score":0.0,"tokens":1151573,"cost_usd":0.310291,"elapsed_time":230.57,"gpr":0.0909,"tpe":0.81,"ee":0.3889},{"task_id":"task_437_international_comparison_hard_hard008","category":"international_comparison","score":0.25,"tokens":722532,"cost_usd":0.193287,"elapsed_time":137.67,"gpr":0.1667,"tpe":0.9,"ee":0.8571},{"task_id":"task_438_international_comparison_hard_hard009","category":"international_comparison","score":0.0,"tokens":1006874,"cost_usd":0.269706,"elapsed_time":274.55,"gpr":0.3333,"tpe":0.729,"ee":0.7},{"task_id":"task_439_international_comparison_hard_hard010","category":"international_comparison","score":0.0,"tokens":2223510,"cost_usd":0.585806,"elapsed_time":164.67,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_440_international_comparison_hard_hard011","category":"international_comparison","score":1.0,"tokens":938468,"cost_usd":0.251161,"elapsed_time":112.57,"gpr":1.0,"tpe":0.5433,"ee":0.5},{"task_id":"task_441_international_comparison_hard_hard012","category":"international_comparison","score":0.0,"tokens":509550,"cost_usd":0.140139,"elapsed_time":182.16,"gpr":0.1429,"tpe":0.5314,"ee":0.5},{"task_id":"task_442_international_comparison_hard_hard013","category":"international_comparison","score":0.25,"tokens":789094,"cost_usd":0.216393,"elapsed_time":217.4,"gpr":0.6,"tpe":0.5216,"ee":0.3333},{"task_id":"task_443_international_comparison_hard_hard014","category":"international_comparison","score":1.0,"tokens":1237408,"cost_usd":0.32954,"elapsed_time":119.59,"gpr":0.8333,"tpe":0.6561,"ee":0.6},{"task_id":"task_444_international_comparison_medium_medium001","category":"international_comparison","score":0.5,"tokens":1026712,"cost_usd":0.274466,"elapsed_time":174.25,"gpr":0.4,"tpe":0.2288,"ee":0.2222},{"task_id":"task_445_international_comparison_medium_medium002","category":"international_comparison","score":0.0,"tokens":608164,"cost_usd":0.165681,"elapsed_time":168.64,"gpr":0.0,"tpe":0.0,"ee":0.2857},{"task_id":"task_446_international_comparison_medium_medium003","category":"international_comparison","score":0.5,"tokens":822282,"cost_usd":0.221404,"elapsed_time":169.87,"gpr":0.1667,"tpe":1.0,"ee":0.2667},{"task_id":"task_447_international_comparison_medium_medium004","category":"international_comparison","score":0.0,"tokens":803433,"cost_usd":0.219937,"elapsed_time":137.68,"gpr":0.6667,"tpe":0.2684,"ee":0.2222},{"task_id":"task_448_international_comparison_medium_medium005","category":"international_comparison","score":0.33,"tokens":1005372,"cost_usd":0.268997,"elapsed_time":98.4,"gpr":0.5,"tpe":0.4783,"ee":0.3636},{"task_id":"task_449_international_comparison_medium_medium006","category":"international_comparison","score":0.0,"tokens":1265452,"cost_usd":0.340428,"elapsed_time":238.17,"gpr":0.0,"tpe":0.0,"ee":0.2273},{"task_id":"task_450_international_comparison_medium_medium007","category":"international_comparison","score":0.0,"tokens":547104,"cost_usd":0.151478,"elapsed_time":186.93,"gpr":0.5,"tpe":0.3487,"ee":0.2857},{"task_id":"task_451_international_comparison_medium_medium008","category":"international_comparison","score":0.0,"tokens":409646,"cost_usd":0.113716,"elapsed_time":150.95,"gpr":0.5,"tpe":0.5314,"ee":0.4},{"task_id":"task_452_international_comparison_medium_medium009","category":"international_comparison","score":0.0,"tokens":809980,"cost_usd":0.222728,"elapsed_time":189.18,"gpr":0.2857,"tpe":1.0,"ee":0.3571},{"task_id":"task_453_international_comparison_medium_medium010","category":"international_comparison","score":0.5,"tokens":871533,"cost_usd":0.238005,"elapsed_time":185.55,"gpr":0.6,"tpe":0.2824,"ee":0.2941},{"task_id":"task_454_international_comparison_medium_medium011","category":"international_comparison","score":0.33,"tokens":1712634,"cost_usd":0.459236,"elapsed_time":356.36,"gpr":0.3333,"tpe":0.1677,"ee":0.1429},{"task_id":"task_455_international_comparison_medium_medium012","category":"international_comparison","score":0.33,"tokens":1588096,"cost_usd":0.423759,"elapsed_time":208.36,"gpr":0.5,"tpe":0.1094,"ee":0.1923},{"task_id":"task_456_international_comparison_medium_medium013","category":"international_comparison","score":0.333,"tokens":2202026,"cost_usd":0.587144,"elapsed_time":347.18,"gpr":0.1667,"tpe":0.0646,"ee":0.1333},{"task_id":"task_457_international_comparison_medium_medium014","category":"international_comparison","score":0.0,"tokens":808446,"cost_usd":0.221207,"elapsed_time":176.02,"gpr":0.375,"tpe":0.2824,"ee":0.2941},{"task_id":"task_458_international_comparison_medium_medium015","category":"international_comparison","score":0.0,"tokens":851247,"cost_usd":0.235856,"elapsed_time":235.21,"gpr":0.1429,"tpe":1.0,"ee":0.25},{"task_id":"task_459_international_comparison_medium_medium016","category":"international_comparison","score":0.0,"tokens":1004130,"cost_usd":0.277257,"elapsed_time":199.51,"gpr":0.5,"tpe":0.2824,"ee":0.2941},{"task_id":"task_460_international_comparison_medium_medium017","category":"international_comparison","score":0.0,"tokens":923952,"cost_usd":0.246539,"elapsed_time":170.66,"gpr":0.0,"tpe":0.0,"ee":0.3125},{"task_id":"task_461_international_comparison_medium_medium018","category":"international_comparison","score":0.5,"tokens":607614,"cost_usd":0.167909,"elapsed_time":173.94,"gpr":0.4286,"tpe":0.3487,"ee":0.2857},{"task_id":"task_462_international_comparison_medium_medium019","category":"international_comparison","score":0.5,"tokens":663515,"cost_usd":0.183327,"elapsed_time":213.82,"gpr":0.5,"tpe":0.3487,"ee":0.2857},{"task_id":"task_463_international_comparison_medium_medium020","category":"international_comparison","score":0.0,"tokens":1139480,"cost_usd":0.302653,"elapsed_time":138.78,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_464_international_comparison_medium_medium021","category":"international_comparison","score":0.0,"tokens":1277399,"cost_usd":0.347917,"elapsed_time":173.8,"gpr":0.0,"tpe":0.0,"ee":0.2381},{"task_id":"task_465_international_comparison_medium_medium022","category":"international_comparison","score":0.0,"tokens":1934235,"cost_usd":0.508076,"elapsed_time":141.49,"gpr":0.3636,"tpe":0.3804,"ee":0.25},{"task_id":"task_466_international_comparison_medium_medium023","category":"international_comparison","score":0.0,"tokens":1349772,"cost_usd":0.360067,"elapsed_time":211.06,"gpr":0.1667,"tpe":1.0,"ee":0.2778},{"task_id":"task_467_international_comparison_medium_medium024","category":"international_comparison","score":0.0,"tokens":964495,"cost_usd":0.258942,"elapsed_time":191.29,"gpr":0.0833,"tpe":0.9,"ee":0.25},{"task_id":"task_468_international_comparison_medium_medium025","category":"international_comparison","score":0.0,"tokens":843670,"cost_usd":0.231468,"elapsed_time":160.94,"gpr":0.3333,"tpe":1.0,"ee":0.3077},{"task_id":"task_469_risk_assessment_hard_hard001","category":"risk_assessment","score":0.0,"tokens":925168,"cost_usd":0.246788,"elapsed_time":116.13,"gpr":0.0,"tpe":0.0,"ee":0.3684},{"task_id":"task_470_risk_assessment_hard_hard002","category":"risk_assessment","score":1.0,"tokens":4951344,"cost_usd":1.309009,"elapsed_time":189.96,"gpr":1.0,"tpe":0.3201,"ee":0.28},{"task_id":"task_471_risk_assessment_hard_hard003","category":"risk_assessment","score":0.0,"tokens":3142830,"cost_usd":0.841913,"elapsed_time":243.01,"gpr":0.6364,"tpe":0.7596,"ee":0.1944},{"task_id":"task_472_risk_assessment_hard_hard004","category":"risk_assessment","score":1.0,"tokens":1943622,"cost_usd":0.525817,"elapsed_time":268.59,"gpr":1.0,"tpe":0.0714,"ee":0.175},{"task_id":"task_473_risk_assessment_hard_hard005","category":"risk_assessment","score":0.0,"tokens":1113490,"cost_usd":0.320026,"elapsed_time":301.44,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_474_risk_assessment_hard_hard006","category":"risk_assessment","score":0.0,"tokens":4974616,"cost_usd":1.311547,"elapsed_time":222.06,"gpr":0.6,"tpe":0.1216,"ee":0.2593},{"task_id":"task_475_risk_assessment_hard_hard007","category":"risk_assessment","score":1.0,"tokens":529076,"cost_usd":0.153391,"elapsed_time":203.51,"gpr":1.0,"tpe":0.2527,"ee":0.3043},{"task_id":"task_476_risk_assessment_hard_hard008","category":"risk_assessment","score":1.0,"tokens":1204564,"cost_usd":0.331632,"elapsed_time":209.05,"gpr":1.0,"tpe":0.3679,"ee":0.2258},{"task_id":"task_477_risk_assessment_hard_hard009","category":"risk_assessment","score":0.0,"tokens":2055094,"cost_usd":0.569309,"elapsed_time":689.25,"gpr":0.0,"tpe":0.0,"ee":0.0722},{"task_id":"task_478_risk_assessment_hard_hard010","category":"risk_assessment","score":0.0,"tokens":1021975,"cost_usd":0.273185,"elapsed_time":127.64,"gpr":0.0,"tpe":0.0,"ee":0.381},{"task_id":"task_479_risk_assessment_hard_hard011","category":"risk_assessment","score":0.0,"tokens":33288080,"cost_usd":8.665943,"elapsed_time":1207.19,"gpr":0.1,"tpe":0.2824,"ee":0.0352},{"task_id":"task_480_risk_assessment_hard_hard012","category":"risk_assessment","score":1.0,"tokens":1001295,"cost_usd":0.299949,"elapsed_time":413.53,"gpr":1.0,"tpe":0.0225,"ee":0.1346},{"task_id":"task_481_risk_assessment_hard_hard013","category":"risk_assessment","score":0.0,"tokens":4414619,"cost_usd":1.159016,"elapsed_time":190.33,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_482_risk_assessment_medium_medium001","category":"risk_assessment","score":0.5,"tokens":1453020,"cost_usd":0.388023,"elapsed_time":278.19,"gpr":0.0,"tpe":0.0,"ee":0.1714},{"task_id":"task_483_risk_assessment_medium_medium002","category":"risk_assessment","score":0.0,"tokens":2834194,"cost_usd":0.747496,"elapsed_time":342.46,"gpr":0.25,"tpe":0.0471,"ee":0.1154},{"task_id":"task_484_risk_assessment_medium_medium003","category":"risk_assessment","score":0.0,"tokens":2645339,"cost_usd":0.72537,"elapsed_time":512.2,"gpr":0.4,"tpe":1.0,"ee":0.0811},{"task_id":"task_485_risk_assessment_medium_medium004","category":"risk_assessment","score":0.0,"tokens":4896726,"cost_usd":1.296833,"elapsed_time":395.23,"gpr":0.5,"tpe":0.1685,"ee":0.0943},{"task_id":"task_486_risk_assessment_medium_medium005","category":"risk_assessment","score":0.0,"tokens":2206001,"cost_usd":0.600239,"elapsed_time":368.5,"gpr":0.375,"tpe":0.3138,"ee":0.1042},{"task_id":"task_487_risk_assessment_medium_medium006","category":"risk_assessment","score":1.0,"tokens":1472403,"cost_usd":0.402282,"elapsed_time":256.96,"gpr":1.0,"tpe":0.1018,"ee":0.0909},{"task_id":"task_488_risk_assessment_medium_medium007","category":"risk_assessment","score":0.0,"tokens":1502292,"cost_usd":0.412565,"elapsed_time":555.65,"gpr":0.0,"tpe":0.0,"ee":0.0632},{"task_id":"task_489_risk_assessment_medium_medium008","category":"risk_assessment","score":0.0,"tokens":1292446,"cost_usd":0.343377,"elapsed_time":362.13,"gpr":0.1875,"tpe":0.0097,"ee":0.0938},{"task_id":"task_490_risk_assessment_medium_medium009","category":"risk_assessment","score":0.5,"tokens":1196251,"cost_usd":0.329853,"elapsed_time":187.22,"gpr":0.25,"tpe":0.3882,"ee":0.2727},{"task_id":"task_491_risk_assessment_medium_medium010","category":"risk_assessment","score":1.0,"tokens":19695458,"cost_usd":5.134269,"elapsed_time":353.13,"gpr":1.0,"tpe":0.0834,"ee":0.0938},{"task_id":"task_492_risk_assessment_medium_medium011","category":"risk_assessment","score":0.0,"tokens":1384225,"cost_usd":0.374082,"elapsed_time":145.58,"gpr":0.3636,"tpe":0.4091,"ee":0.3}]},{"model":"deepseek-v3.2","acc":0.3415,"ee":0.1562,"gpr":0.2469,"tpe":0.1876,"efficiency":{"total_tokens":699964736,"total_cost_usd":172.17,"total_requests":17033},"category_scores":{"comprehensive_decision":{"acc":0.2062,"ee":0.1528,"gpr":0.2472,"tpe":0.1071,"count":70},"enterprise_industry_analysis":{"acc":0.4558,"ee":0.1479,"gpr":0.2267,"tpe":0.1226,"count":226},"enterprise_industry_policy_analysis":{"acc":0.3684,"ee":0.1779,"gpr":0.3747,"tpe":0.3309,"count":76},"hypothesis_verification":{"acc":0.1379,"ee":0.1166,"gpr":0.2056,"tpe":0.2805,"count":29},"industry_planning":{"acc":0.1071,"ee":0.1596,"gpr":0.2343,"tpe":0.1329,"count":28},"international_comparison":{"acc":0.1943,"ee":0.3422,"gpr":0.2002,"tpe":0.2864,"count":39},"risk_assessment":{"acc":0.3333,"ee":0.1656,"gpr":0.2127,"tpe":0.2606,"count":24}},"tasks":[{"task_id":"task_001_comprehensive_decision_easy_easy001","category":"comprehensive_decision","score":0.0,"tokens":601890,"cost_usd":0.119892,"elapsed_time":893.6,"gpr":0.0,"tpe":0.0,"ee":0.1034},{"task_id":"task_002_comprehensive_decision_easy_easy002","category":"comprehensive_decision","score":1.0,"tokens":325812,"cost_usd":0.068343,"elapsed_time":371.94,"gpr":1.0,"tpe":0.2159,"ee":0.1},{"task_id":"task_003_comprehensive_decision_easy_easy003","category":"comprehensive_decision","score":0.0,"tokens":537056,"cost_usd":0.117591,"elapsed_time":191.7,"gpr":0.0,"tpe":0.0,"ee":0.1579},{"task_id":"task_004_comprehensive_decision_easy_easy004","category":"comprehensive_decision","score":0.0,"tokens":169248,"cost_usd":0.03367,"elapsed_time":251.0,"gpr":0.8333,"tpe":0.3874,"ee":0.1538},{"task_id":"task_005_comprehensive_decision_easy_easy005","category":"comprehensive_decision","score":0.0,"tokens":658355,"cost_usd":0.125927,"elapsed_time":487.18,"gpr":0.6667,"tpe":0.2542,"ee":0.1},{"task_id":"task_006_comprehensive_decision_easy_easy006","category":"comprehensive_decision","score":0.0,"tokens":2320773,"cost_usd":0.535068,"elapsed_time":597.11,"gpr":0.5,"tpe":0.0646,"ee":0.0882},{"task_id":"task_007_comprehensive_decision_hard_hard001","category":"comprehensive_decision","score":0.0,"tokens":809000,"cost_usd":0.175807,"elapsed_time":1087.83,"gpr":0.0,"tpe":0.0,"ee":0.2059},{"task_id":"task_008_comprehensive_decision_hard_hard002","category":"comprehensive_decision","score":0.0,"tokens":3123837,"cost_usd":0.753973,"elapsed_time":1198.13,"gpr":0.0,"tpe":0.0,"ee":0.1228},{"task_id":"task_009_comprehensive_decision_hard_hard003","category":"comprehensive_decision","score":0.0,"tokens":3319370,"cost_usd":0.786401,"elapsed_time":882.34,"gpr":0.4286,"tpe":0.1682,"ee":0.1111},{"task_id":"task_010_comprehensive_decision_hard_hard004","category":"comprehensive_decision","score":0.0,"tokens":2162757,"cost_usd":0.506872,"elapsed_time":1132.94,"gpr":0.375,"tpe":0.1617,"ee":0.1273},{"task_id":"task_011_comprehensive_decision_hard_hard005","category":"comprehensive_decision","score":0.0,"tokens":2146257,"cost_usd":0.496697,"elapsed_time":1207.14,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_012_comprehensive_decision_hard_hard006","category":"comprehensive_decision","score":0.0,"tokens":478372,"cost_usd":0.10683,"elapsed_time":310.03,"gpr":0.0,"tpe":0.0,"ee":0.3043},{"task_id":"task_013_comprehensive_decision_hard_hard007","category":"comprehensive_decision","score":0.0,"tokens":1398028,"cost_usd":0.323518,"elapsed_time":1207.26,"gpr":0.1667,"tpe":1.0,"ee":0.125},{"task_id":"task_014_comprehensive_decision_hard_hard008","category":"comprehensive_decision","score":0.0,"tokens":1358605,"cost_usd":0.331577,"elapsed_time":1207.15,"gpr":0.0,"tpe":0.0,"ee":0.1296},{"task_id":"task_015_comprehensive_decision_hard_hard009","category":"comprehensive_decision","score":0.0,"tokens":792352,"cost_usd":0.186482,"elapsed_time":720.52,"gpr":0.1429,"tpe":0.3138,"ee":0.2857},{"task_id":"task_016_comprehensive_decision_hard_hard010","category":"comprehensive_decision","score":0.0,"tokens":3215218,"cost_usd":0.772657,"elapsed_time":1207.21,"gpr":0.0,"tpe":0.0,"ee":0.1373},{"task_id":"task_017_comprehensive_decision_hard_hard011","category":"comprehensive_decision","score":0.0,"tokens":3500242,"cost_usd":0.783308,"elapsed_time":1052.95,"gpr":0.1111,"tpe":1.0,"ee":0.0946},{"task_id":"task_018_comprehensive_decision_hard_hard012","category":"comprehensive_decision","score":0.0,"tokens":1430092,"cost_usd":0.331383,"elapsed_time":1202.58,"gpr":0.8333,"tpe":0.1028,"ee":0.1228},{"task_id":"task_019_comprehensive_decision_hard_hard013","category":"comprehensive_decision","score":0.0,"tokens":645514,"cost_usd":0.156205,"elapsed_time":889.16,"gpr":0.0,"tpe":0.0,"ee":0.2333},{"task_id":"task_020_comprehensive_decision_hard_hard014","category":"comprehensive_decision","score":0.0,"tokens":3796800,"cost_usd":0.847827,"elapsed_time":1205.37,"gpr":0.2857,"tpe":0.1183,"ee":0.1014},{"task_id":"task_021_comprehensive_decision_hard_hard015","category":"comprehensive_decision","score":0.0,"tokens":876273,"cost_usd":0.197659,"elapsed_time":1207.12,"gpr":0.2857,"tpe":0.0424,"ee":0.186},{"task_id":"task_022_comprehensive_decision_hard_hard016","category":"comprehensive_decision","score":0.0,"tokens":45993,"cost_usd":0.011595,"elapsed_time":30.26,"gpr":0.0,"tpe":0.0,"ee":1.5},{"task_id":"task_023_comprehensive_decision_hard_hard017","category":"comprehensive_decision","score":0.0,"tokens":1610113,"cost_usd":0.371254,"elapsed_time":424.21,"gpr":0.5,"tpe":0.1116,"ee":0.1765},{"task_id":"task_024_comprehensive_decision_hard_hard018","category":"comprehensive_decision","score":0.0,"tokens":1596628,"cost_usd":0.400876,"elapsed_time":407.27,"gpr":0.8571,"tpe":0.3442,"ee":0.3182},{"task_id":"task_025_comprehensive_decision_hard_hard019","category":"comprehensive_decision","score":0.67,"tokens":664640,"cost_usd":0.159037,"elapsed_time":661.28,"gpr":1.0,"tpe":0.0985,"ee":0.2069},{"task_id":"task_026_comprehensive_decision_medium_medium001","category":"comprehensive_decision","score":0.0,"tokens":1381528,"cost_usd":0.326405,"elapsed_time":1207.1,"gpr":0.5,"tpe":0.0087,"ee":0.0909},{"task_id":"task_027_comprehensive_decision_medium_medium002","category":"comprehensive_decision","score":1.0,"tokens":1340503,"cost_usd":0.318639,"elapsed_time":546.35,"gpr":1.0,"tpe":0.3157,"ee":0.1429},{"task_id":"task_028_comprehensive_decision_medium_medium003","category":"comprehensive_decision","score":0.5,"tokens":3744916,"cost_usd":0.942403,"elapsed_time":951.37,"gpr":0.4286,"tpe":0.0314,"ee":0.0847},{"task_id":"task_029_comprehensive_decision_medium_medium004","category":"comprehensive_decision","score":0.5,"tokens":1861385,"cost_usd":0.465708,"elapsed_time":952.55,"gpr":0.625,"tpe":0.0836,"ee":0.0746},{"task_id":"task_030_comprehensive_decision_medium_medium005","category":"comprehensive_decision","score":0.0,"tokens":222174,"cost_usd":0.044633,"elapsed_time":223.54,"gpr":0.5,"tpe":0.4783,"ee":0.3333},{"task_id":"task_031_comprehensive_decision_medium_medium006","category":"comprehensive_decision","score":1.0,"tokens":387168,"cost_usd":0.081034,"elapsed_time":210.02,"gpr":1.0,"tpe":0.2981,"ee":0.2353},{"task_id":"task_032_comprehensive_decision_medium_medium007","category":"comprehensive_decision","score":0.0,"tokens":725868,"cost_usd":0.183068,"elapsed_time":181.73,"gpr":0.0,"tpe":0.0,"ee":0.5556},{"task_id":"task_033_comprehensive_decision_medium_medium008","category":"comprehensive_decision","score":1.0,"tokens":528388,"cost_usd":0.125514,"elapsed_time":492.52,"gpr":1.0,"tpe":0.3927,"ee":0.1667},{"task_id":"task_034_comprehensive_decision_medium_medium009","category":"comprehensive_decision","score":1.0,"tokens":1322608,"cost_usd":0.323696,"elapsed_time":787.27,"gpr":1.0,"tpe":0.0346,"ee":0.1064},{"task_id":"task_035_comprehensive_decision_medium_medium010","category":"comprehensive_decision","score":0.0,"tokens":818309,"cost_usd":0.198544,"elapsed_time":629.25,"gpr":0.0,"tpe":0.0,"ee":0.1042},{"task_id":"task_036_comprehensive_decision_medium_medium011","category":"comprehensive_decision","score":0.0,"tokens":1613331,"cost_usd":0.387145,"elapsed_time":1161.81,"gpr":0.0,"tpe":0.0,"ee":0.0806},{"task_id":"task_037_comprehensive_decision_medium_medium012","category":"comprehensive_decision","score":0.0,"tokens":2483070,"cost_usd":0.605532,"elapsed_time":1057.57,"gpr":0.0,"tpe":0.0,"ee":0.0641},{"task_id":"task_038_comprehensive_decision_medium_medium013","category":"comprehensive_decision","score":0.0,"tokens":726526,"cost_usd":0.183412,"elapsed_time":663.56,"gpr":0.6,"tpe":0.0347,"ee":0.125},{"task_id":"task_039_comprehensive_decision_medium_medium014","category":"comprehensive_decision","score":0.5,"tokens":1139062,"cost_usd":0.263893,"elapsed_time":864.04,"gpr":0.8333,"tpe":0.2151,"ee":0.1087},{"task_id":"task_040_comprehensive_decision_medium_medium015","category":"comprehensive_decision","score":1.0,"tokens":1520167,"cost_usd":0.383899,"elapsed_time":669.07,"gpr":1.0,"tpe":0.1094,"ee":0.1667},{"task_id":"task_041_comprehensive_decision_medium_medium016","category":"comprehensive_decision","score":0.0,"tokens":2149303,"cost_usd":0.522614,"elapsed_time":1207.12,"gpr":0.0,"tpe":0.0,"ee":0.0725},{"task_id":"task_042_comprehensive_decision_medium_medium017","category":"comprehensive_decision","score":0.0,"tokens":168063,"cost_usd":0.042574,"elapsed_time":170.04,"gpr":0.0,"tpe":0.0,"ee":0.4167},{"task_id":"task_043_comprehensive_decision_medium_medium018","category":"comprehensive_decision","score":0.0,"tokens":1157349,"cost_usd":0.257248,"elapsed_time":334.74,"gpr":0.0,"tpe":0.0,"ee":0.1852},{"task_id":"task_044_comprehensive_decision_medium_medium019","category":"comprehensive_decision","score":0.0,"tokens":142832,"cost_usd":0.032335,"elapsed_time":74.37,"gpr":0.0,"tpe":0.0,"ee":0.5556},{"task_id":"task_045_comprehensive_decision_medium_medium020","category":"comprehensive_decision","score":0.0,"tokens":1737120,"cost_usd":0.420697,"elapsed_time":995.75,"gpr":0.0,"tpe":0.0,"ee":0.0746},{"task_id":"task_046_comprehensive_decision_medium_medium021","category":"comprehensive_decision","score":0.0,"tokens":2250567,"cost_usd":0.548654,"elapsed_time":988.95,"gpr":0.0,"tpe":0.0,"ee":0.0909},{"task_id":"task_047_comprehensive_decision_medium_medium022","category":"comprehensive_decision","score":1.0,"tokens":2350813,"cost_usd":0.590028,"elapsed_time":578.23,"gpr":1.0,"tpe":0.0225,"ee":0.122},{"task_id":"task_048_comprehensive_decision_medium_medium023","category":"comprehensive_decision","score":0.0,"tokens":2175004,"cost_usd":0.549228,"elapsed_time":465.91,"gpr":0.5,"tpe":0.0758,"ee":0.1},{"task_id":"task_049_comprehensive_decision_medium_medium024","category":"comprehensive_decision","score":0.0,"tokens":647828,"cost_usd":0.161189,"elapsed_time":288.63,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_050_comprehensive_decision_medium_medium025","category":"comprehensive_decision","score":0.0,"tokens":2560874,"cost_usd":0.644518,"elapsed_time":405.47,"gpr":0.0,"tpe":0.0,"ee":0.1724},{"task_id":"task_051_comprehensive_decision_medium_medium026","category":"comprehensive_decision","score":1.0,"tokens":755237,"cost_usd":0.182743,"elapsed_time":253.3,"gpr":1.0,"tpe":0.1501,"ee":0.2174},{"task_id":"task_052_comprehensive_decision_medium_medium027","category":"comprehensive_decision","score":0.0,"tokens":400459,"cost_usd":0.089347,"elapsed_time":361.4,"gpr":0.0,"tpe":0.0,"ee":0.2941},{"task_id":"task_053_comprehensive_decision_medium_medium028","category":"comprehensive_decision","score":0.765,"tokens":2723602,"cost_usd":0.687623,"elapsed_time":717.5,"gpr":0.0,"tpe":0.0,"ee":0.1389},{"task_id":"task_054_comprehensive_decision_medium_medium029","category":"comprehensive_decision","score":0.0,"tokens":2338963,"cost_usd":0.552608,"elapsed_time":604.79,"gpr":0.75,"tpe":0.2869,"ee":0.1667},{"task_id":"task_055_comprehensive_decision_medium_medium030","category":"comprehensive_decision","score":0.0,"tokens":2672234,"cost_usd":0.662944,"elapsed_time":659.98,"gpr":0.4286,"tpe":0.025,"ee":0.1},{"task_id":"task_056_comprehensive_decision_medium_medium031","category":"comprehensive_decision","score":1.0,"tokens":298302,"cost_usd":0.075752,"elapsed_time":272.75,"gpr":1.0,"tpe":0.2535,"ee":0.2105},{"task_id":"task_057_comprehensive_decision_medium_medium032","category":"comprehensive_decision","score":0.0,"tokens":646819,"cost_usd":0.15763,"elapsed_time":623.2,"gpr":0.5714,"tpe":0.1426,"ee":0.1316},{"task_id":"task_058_comprehensive_decision_medium_medium033","category":"comprehensive_decision","score":0.5,"tokens":1334076,"cost_usd":0.332776,"elapsed_time":764.86,"gpr":0.25,"tpe":0.0424,"ee":0.0926},{"task_id":"task_059_comprehensive_decision_medium_medium034","category":"comprehensive_decision","score":0.0,"tokens":1357231,"cost_usd":0.329463,"elapsed_time":701.36,"gpr":0.0,"tpe":0.0,"ee":0.122},{"task_id":"task_060_comprehensive_decision_medium_medium035","category":"comprehensive_decision","score":0.0,"tokens":6484383,"cost_usd":1.614913,"elapsed_time":1154.33,"gpr":0.4444,"tpe":0.002,"ee":0.0658},{"task_id":"task_061_comprehensive_decision_medium_medium036","category":"comprehensive_decision","score":0.0,"tokens":3252595,"cost_usd":0.805198,"elapsed_time":520.86,"gpr":0.3333,"tpe":0.2542,"ee":0.1163},{"task_id":"task_062_comprehensive_decision_medium_medium037","category":"comprehensive_decision","score":0.0,"tokens":2084728,"cost_usd":0.519515,"elapsed_time":341.66,"gpr":0.3333,"tpe":0.1668,"ee":0.2174},{"task_id":"task_063_comprehensive_decision_medium_medium038","category":"comprehensive_decision","score":0.0,"tokens":694145,"cost_usd":0.173638,"elapsed_time":322.53,"gpr":0.0,"tpe":0.0,"ee":0.1379},{"task_id":"task_064_comprehensive_decision_medium_medium039","category":"comprehensive_decision","score":1.0,"tokens":2272754,"cost_usd":0.567325,"elapsed_time":933.07,"gpr":1.0,"tpe":0.0571,"ee":0.0615},{"task_id":"task_065_comprehensive_decision_medium_medium040","category":"comprehensive_decision","score":1.0,"tokens":1485476,"cost_usd":0.374309,"elapsed_time":490.85,"gpr":1.0,"tpe":0.0694,"ee":0.1515},{"task_id":"task_066_comprehensive_decision_medium_medium041","category":"comprehensive_decision","score":0.0,"tokens":529085,"cost_usd":0.125179,"elapsed_time":538.41,"gpr":0.0,"tpe":0.0,"ee":0.1562},{"task_id":"task_067_comprehensive_decision_medium_medium042","category":"comprehensive_decision","score":0.0,"tokens":4966923,"cost_usd":1.218989,"elapsed_time":1207.14,"gpr":0.25,"tpe":0.2824,"ee":0.0575},{"task_id":"task_068_comprehensive_decision_medium_medium043","category":"comprehensive_decision","score":0.0,"tokens":325705,"cost_usd":0.075056,"elapsed_time":285.69,"gpr":0.0,"tpe":0.0,"ee":0.1818},{"task_id":"task_069_comprehensive_decision_medium_medium044","category":"comprehensive_decision","score":0.0,"tokens":3217726,"cost_usd":0.751625,"elapsed_time":770.12,"gpr":0.25,"tpe":0.0203,"ee":0.0877},{"task_id":"task_070_comprehensive_decision_medium_medium045","category":"comprehensive_decision","score":0.0,"tokens":2440370,"cost_usd":0.613034,"elapsed_time":459.32,"gpr":0.0,"tpe":0.0,"ee":0.1613},{"task_id":"task_071_enterprise_industry_analysis_easy_easy001","category":"enterprise_industry_analysis","score":0.0,"tokens":2298083,"cost_usd":0.563027,"elapsed_time":609.99,"gpr":0.0,"tpe":0.0,"ee":0.0667},{"task_id":"task_072_enterprise_industry_analysis_easy_easy002","category":"enterprise_industry_analysis","score":0.0,"tokens":3510877,"cost_usd":0.88445,"elapsed_time":727.75,"gpr":0.25,"tpe":0.0523,"ee":0.0667},{"task_id":"task_073_enterprise_industry_analysis_easy_easy003","category":"enterprise_industry_analysis","score":0.0,"tokens":1045142,"cost_usd":0.263708,"elapsed_time":268.03,"gpr":0.0,"tpe":0.0,"ee":0.1875},{"task_id":"task_074_enterprise_industry_analysis_easy_easy004","category":"enterprise_industry_analysis","score":0.0,"tokens":3809766,"cost_usd":0.941377,"elapsed_time":708.34,"gpr":0.0,"tpe":0.0,"ee":0.0652},{"task_id":"task_075_enterprise_industry_analysis_easy_easy005","category":"enterprise_industry_analysis","score":1.0,"tokens":908058,"cost_usd":0.227267,"elapsed_time":373.92,"gpr":1.0,"tpe":0.0523,"ee":0.0968},{"task_id":"task_076_enterprise_industry_analysis_easy_easy006","category":"enterprise_industry_analysis","score":1.0,"tokens":1125483,"cost_usd":0.27224,"elapsed_time":330.24,"gpr":1.0,"tpe":0.0801,"ee":0.1071},{"task_id":"task_077_enterprise_industry_analysis_easy_easy007","category":"enterprise_industry_analysis","score":0.0,"tokens":733232,"cost_usd":0.183593,"elapsed_time":171.22,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_078_enterprise_industry_analysis_easy_easy008","category":"enterprise_industry_analysis","score":0.0,"tokens":2882706,"cost_usd":0.727014,"elapsed_time":506.13,"gpr":0.0,"tpe":0.0,"ee":0.0857},{"task_id":"task_079_enterprise_industry_analysis_easy_easy009","category":"enterprise_industry_analysis","score":1.0,"tokens":1092392,"cost_usd":0.273792,"elapsed_time":333.49,"gpr":1.0,"tpe":0.0927,"ee":0.1},{"task_id":"task_080_enterprise_industry_analysis_easy_easy010","category":"enterprise_industry_analysis","score":1.0,"tokens":3232939,"cost_usd":0.81607,"elapsed_time":814.66,"gpr":1.0,"tpe":0.0133,"ee":0.0682},{"task_id":"task_081_enterprise_industry_analysis_easy_easy011","category":"enterprise_industry_analysis","score":0.0,"tokens":2745397,"cost_usd":0.681049,"elapsed_time":666.92,"gpr":0.3333,"tpe":0.003,"ee":0.0517},{"task_id":"task_082_enterprise_industry_analysis_easy_easy012","category":"enterprise_industry_analysis","score":1.0,"tokens":2175160,"cost_usd":0.549673,"elapsed_time":539.12,"gpr":1.0,"tpe":0.0343,"ee":0.0857},{"task_id":"task_083_enterprise_industry_analysis_easy_easy013","category":"enterprise_industry_analysis","score":0.0,"tokens":275066,"cost_usd":0.06328,"elapsed_time":230.76,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_084_enterprise_industry_analysis_easy_easy014","category":"enterprise_industry_analysis","score":1.0,"tokens":484896,"cost_usd":0.119057,"elapsed_time":218.84,"gpr":1.0,"tpe":0.1216,"ee":0.1304},{"task_id":"task_085_enterprise_industry_analysis_easy_easy015","category":"enterprise_industry_analysis","score":0.0,"tokens":3353206,"cost_usd":0.817973,"elapsed_time":860.82,"gpr":0.3333,"tpe":0.0064,"ee":0.0566},{"task_id":"task_086_enterprise_industry_analysis_easy_easy016","category":"enterprise_industry_analysis","score":1.0,"tokens":1657421,"cost_usd":0.412124,"elapsed_time":384.08,"gpr":1.0,"tpe":0.0886,"ee":0.1154},{"task_id":"task_087_enterprise_industry_analysis_easy_easy017","category":"enterprise_industry_analysis","score":0.0,"tokens":1223754,"cost_usd":0.283911,"elapsed_time":307.02,"gpr":0.6667,"tpe":0.9,"ee":0.1429},{"task_id":"task_088_enterprise_industry_analysis_easy_easy018","category":"enterprise_industry_analysis","score":0.0,"tokens":170609,"cost_usd":0.03688,"elapsed_time":173.92,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_089_enterprise_industry_analysis_easy_easy019","category":"enterprise_industry_analysis","score":0.0,"tokens":1545366,"cost_usd":0.378048,"elapsed_time":668.9,"gpr":0.0,"tpe":0.0,"ee":0.0811},{"task_id":"task_090_enterprise_industry_analysis_easy_easy020","category":"enterprise_industry_analysis","score":0.0,"tokens":1562566,"cost_usd":0.385667,"elapsed_time":511.45,"gpr":0.0,"tpe":0.0,"ee":0.075},{"task_id":"task_091_enterprise_industry_analysis_easy_easy021","category":"enterprise_industry_analysis","score":0.0,"tokens":1698813,"cost_usd":0.426019,"elapsed_time":614.92,"gpr":0.3333,"tpe":0.1094,"ee":0.0938},{"task_id":"task_092_enterprise_industry_analysis_easy_easy022","category":"enterprise_industry_analysis","score":1.0,"tokens":2250738,"cost_usd":0.567931,"elapsed_time":617.67,"gpr":1.0,"tpe":0.0279,"ee":0.0789},{"task_id":"task_093_enterprise_industry_analysis_easy_easy023","category":"enterprise_industry_analysis","score":0.0,"tokens":3523389,"cost_usd":0.888911,"elapsed_time":691.29,"gpr":0.0,"tpe":0.0,"ee":0.0667},{"task_id":"task_094_enterprise_industry_analysis_easy_easy024","category":"enterprise_industry_analysis","score":1.0,"tokens":912142,"cost_usd":0.230587,"elapsed_time":292.78,"gpr":1.0,"tpe":0.1579,"ee":0.1304},{"task_id":"task_095_enterprise_industry_analysis_easy_easy025","category":"enterprise_industry_analysis","score":1.0,"tokens":643952,"cost_usd":0.163075,"elapsed_time":338.06,"gpr":1.0,"tpe":0.1094,"ee":0.125},{"task_id":"task_096_enterprise_industry_analysis_easy_easy026","category":"enterprise_industry_analysis","score":0.0,"tokens":3920198,"cost_usd":0.970119,"elapsed_time":817.67,"gpr":0.0,"tpe":0.0,"ee":0.0526},{"task_id":"task_097_enterprise_industry_analysis_easy_easy027","category":"enterprise_industry_analysis","score":0.0,"tokens":1298402,"cost_usd":0.32132,"elapsed_time":320.13,"gpr":0.6667,"tpe":0.6561,"ee":0.0882},{"task_id":"task_098_enterprise_industry_analysis_easy_easy028","category":"enterprise_industry_analysis","score":0.0,"tokens":5305374,"cost_usd":1.338893,"elapsed_time":960.36,"gpr":0.0,"tpe":0.0,"ee":0.0536},{"task_id":"task_099_enterprise_industry_analysis_easy_easy029","category":"enterprise_industry_analysis","score":1.0,"tokens":2625778,"cost_usd":0.655692,"elapsed_time":668.88,"gpr":1.0,"tpe":0.0106,"ee":0.0612},{"task_id":"task_100_enterprise_industry_analysis_easy_easy030","category":"enterprise_industry_analysis","score":0.0,"tokens":316007,"cost_usd":0.073816,"elapsed_time":162.88,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_101_enterprise_industry_analysis_easy_easy031","category":"enterprise_industry_analysis","score":0.0,"tokens":1481902,"cost_usd":0.371114,"elapsed_time":536.15,"gpr":0.3333,"tpe":0.0424,"ee":0.0652},{"task_id":"task_102_enterprise_industry_analysis_easy_easy032","category":"enterprise_industry_analysis","score":0.0,"tokens":2514407,"cost_usd":0.619172,"elapsed_time":802.0,"gpr":0.3333,"tpe":0.0343,"ee":0.0577},{"task_id":"task_103_enterprise_industry_analysis_easy_easy033","category":"enterprise_industry_analysis","score":0.0,"tokens":2703103,"cost_usd":0.679784,"elapsed_time":1063.38,"gpr":0.0,"tpe":0.0,"ee":0.0441},{"task_id":"task_104_enterprise_industry_analysis_easy_easy034","category":"enterprise_industry_analysis","score":1.0,"tokens":1492464,"cost_usd":0.35914,"elapsed_time":569.52,"gpr":1.0,"tpe":0.0148,"ee":0.0698},{"task_id":"task_105_enterprise_industry_analysis_easy_easy035","category":"enterprise_industry_analysis","score":1.0,"tokens":380896,"cost_usd":0.093795,"elapsed_time":177.31,"gpr":1.0,"tpe":0.1668,"ee":0.15},{"task_id":"task_106_enterprise_industry_analysis_easy_easy036","category":"enterprise_industry_analysis","score":0.0,"tokens":898323,"cost_usd":0.227478,"elapsed_time":465.29,"gpr":0.0,"tpe":0.0,"ee":0.12},{"task_id":"task_107_enterprise_industry_analysis_easy_easy037","category":"enterprise_industry_analysis","score":0.0,"tokens":1014201,"cost_usd":0.253586,"elapsed_time":305.09,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_108_enterprise_industry_analysis_easy_easy038","category":"enterprise_industry_analysis","score":0.0,"tokens":1480541,"cost_usd":0.371997,"elapsed_time":514.63,"gpr":0.0,"tpe":0.0,"ee":0.0968},{"task_id":"task_109_enterprise_industry_analysis_easy_easy039","category":"enterprise_industry_analysis","score":1.0,"tokens":1272688,"cost_usd":0.321433,"elapsed_time":420.05,"gpr":1.0,"tpe":0.0718,"ee":0.1071},{"task_id":"task_110_enterprise_industry_analysis_easy_easy040","category":"enterprise_industry_analysis","score":1.0,"tokens":1444712,"cost_usd":0.364956,"elapsed_time":423.35,"gpr":1.0,"tpe":0.0886,"ee":0.1154},{"task_id":"task_111_enterprise_industry_analysis_easy_easy041","category":"enterprise_industry_analysis","score":0.0,"tokens":1459137,"cost_usd":0.366414,"elapsed_time":610.55,"gpr":0.0,"tpe":0.0,"ee":0.0789},{"task_id":"task_112_enterprise_industry_analysis_easy_easy042","category":"enterprise_industry_analysis","score":0.0,"tokens":1805321,"cost_usd":0.452054,"elapsed_time":578.1,"gpr":0.0,"tpe":0.0,"ee":0.0682},{"task_id":"task_113_enterprise_industry_analysis_easy_easy043","category":"enterprise_industry_analysis","score":1.0,"tokens":316735,"cost_usd":0.0767,"elapsed_time":269.75,"gpr":1.0,"tpe":0.1668,"ee":0.15},{"task_id":"task_114_enterprise_industry_analysis_easy_easy044","category":"enterprise_industry_analysis","score":0.0,"tokens":655277,"cost_usd":0.164214,"elapsed_time":280.2,"gpr":0.0,"tpe":0.0,"ee":0.1364},{"task_id":"task_115_enterprise_industry_analysis_easy_easy045","category":"enterprise_industry_analysis","score":1.0,"tokens":1150417,"cost_usd":0.290717,"elapsed_time":525.47,"gpr":0.0,"tpe":0.0,"ee":0.0968},{"task_id":"task_116_enterprise_industry_analysis_easy_easy046","category":"enterprise_industry_analysis","score":0.0,"tokens":1270814,"cost_usd":0.318585,"elapsed_time":415.01,"gpr":0.0,"tpe":0.0,"ee":0.125},{"task_id":"task_117_enterprise_industry_analysis_easy_easy047","category":"enterprise_industry_analysis","score":0.0,"tokens":583125,"cost_usd":0.14302,"elapsed_time":447.79,"gpr":0.0,"tpe":0.0,"ee":0.1071},{"task_id":"task_118_enterprise_industry_analysis_easy_easy048","category":"enterprise_industry_analysis","score":0.0,"tokens":870036,"cost_usd":0.205806,"elapsed_time":557.46,"gpr":0.0,"tpe":0.0,"ee":0.0968},{"task_id":"task_119_enterprise_industry_analysis_easy_easy049","category":"enterprise_industry_analysis","score":1.0,"tokens":808485,"cost_usd":0.200755,"elapsed_time":402.95,"gpr":1.0,"tpe":0.0582,"ee":0.1},{"task_id":"task_120_enterprise_industry_analysis_easy_easy050","category":"enterprise_industry_analysis","score":1.0,"tokens":463529,"cost_usd":0.111052,"elapsed_time":339.8,"gpr":1.0,"tpe":0.0886,"ee":0.1154},{"task_id":"task_121_enterprise_industry_analysis_easy_easy051","category":"enterprise_industry_analysis","score":0.0,"tokens":171483,"cost_usd":0.034177,"elapsed_time":195.14,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_122_enterprise_industry_analysis_easy_easy052","category":"enterprise_industry_analysis","score":1.0,"tokens":1420867,"cost_usd":0.34444,"elapsed_time":412.73,"gpr":1.0,"tpe":0.0343,"ee":0.0857},{"task_id":"task_123_enterprise_industry_analysis_easy_easy053","category":"enterprise_industry_analysis","score":0.0,"tokens":1457891,"cost_usd":0.364338,"elapsed_time":420.61,"gpr":0.0,"tpe":0.0,"ee":0.0909},{"task_id":"task_124_enterprise_industry_analysis_easy_easy054","category":"enterprise_industry_analysis","score":1.0,"tokens":1101915,"cost_usd":0.276767,"elapsed_time":305.95,"gpr":1.0,"tpe":0.2288,"ee":0.1765},{"task_id":"task_125_enterprise_industry_analysis_easy_easy055","category":"enterprise_industry_analysis","score":0.0,"tokens":784534,"cost_usd":0.196029,"elapsed_time":344.74,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_126_enterprise_industry_analysis_easy_easy056","category":"enterprise_industry_analysis","score":0.0,"tokens":1475148,"cost_usd":0.360507,"elapsed_time":346.3,"gpr":0.0,"tpe":0.0,"ee":0.0968},{"task_id":"task_127_enterprise_industry_analysis_easy_easy057","category":"enterprise_industry_analysis","score":1.0,"tokens":716176,"cost_usd":0.181084,"elapsed_time":299.91,"gpr":1.0,"tpe":0.2288,"ee":0.1765},{"task_id":"task_128_enterprise_industry_analysis_easy_easy058","category":"enterprise_industry_analysis","score":1.0,"tokens":630512,"cost_usd":0.158068,"elapsed_time":207.25,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_129_enterprise_industry_analysis_easy_easy059","category":"enterprise_industry_analysis","score":1.0,"tokens":1046358,"cost_usd":0.262688,"elapsed_time":304.77,"gpr":1.0,"tpe":0.3138,"ee":0.2143},{"task_id":"task_130_enterprise_industry_analysis_easy_easy060","category":"enterprise_industry_analysis","score":1.0,"tokens":475070,"cost_usd":0.116467,"elapsed_time":177.65,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_131_enterprise_industry_analysis_easy_easy061","category":"enterprise_industry_analysis","score":0.0,"tokens":1024755,"cost_usd":0.2562,"elapsed_time":220.23,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_132_enterprise_industry_analysis_easy_easy062","category":"enterprise_industry_analysis","score":0.0,"tokens":438230,"cost_usd":0.107098,"elapsed_time":314.94,"gpr":0.6667,"tpe":0.3628,"ee":0.1111},{"task_id":"task_133_enterprise_industry_analysis_easy_easy063","category":"enterprise_industry_analysis","score":1.0,"tokens":376845,"cost_usd":0.090727,"elapsed_time":214.57,"gpr":1.0,"tpe":0.0886,"ee":0.1154},{"task_id":"task_134_enterprise_industry_analysis_easy_easy064","category":"enterprise_industry_analysis","score":1.0,"tokens":815260,"cost_usd":0.205584,"elapsed_time":356.92,"gpr":1.0,"tpe":0.0582,"ee":0.1},{"task_id":"task_135_enterprise_industry_analysis_easy_easy065","category":"enterprise_industry_analysis","score":0.0,"tokens":593573,"cost_usd":0.147018,"elapsed_time":420.0,"gpr":0.6667,"tpe":0.4089,"ee":0.0968},{"task_id":"task_136_enterprise_industry_analysis_easy_easy066","category":"enterprise_industry_analysis","score":0.0,"tokens":519865,"cost_usd":0.131151,"elapsed_time":121.33,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_137_enterprise_industry_analysis_easy_easy067","category":"enterprise_industry_analysis","score":1.0,"tokens":288120,"cost_usd":0.06364,"elapsed_time":293.78,"gpr":1.0,"tpe":0.2394,"ee":0.1429},{"task_id":"task_138_enterprise_industry_analysis_easy_easy068","category":"enterprise_industry_analysis","score":1.0,"tokens":472956,"cost_usd":0.119598,"elapsed_time":166.9,"gpr":1.0,"tpe":0.2542,"ee":0.1875},{"task_id":"task_139_enterprise_industry_analysis_easy_easy069","category":"enterprise_industry_analysis","score":1.0,"tokens":1744034,"cost_usd":0.44049,"elapsed_time":440.89,"gpr":1.0,"tpe":0.2002,"ee":0.1},{"task_id":"task_140_enterprise_industry_analysis_easy_easy070","category":"enterprise_industry_analysis","score":0.0,"tokens":424166,"cost_usd":0.096573,"elapsed_time":389.8,"gpr":0.0,"tpe":0.0,"ee":0.1111},{"task_id":"task_141_enterprise_industry_analysis_easy_easy071","category":"enterprise_industry_analysis","score":1.0,"tokens":298713,"cost_usd":0.072067,"elapsed_time":230.01,"gpr":1.0,"tpe":0.1668,"ee":0.15},{"task_id":"task_142_enterprise_industry_analysis_easy_easy072","category":"enterprise_industry_analysis","score":1.0,"tokens":235922,"cost_usd":0.055755,"elapsed_time":197.3,"gpr":1.0,"tpe":0.3969,"ee":0.1765},{"task_id":"task_143_enterprise_industry_analysis_easy_easy073","category":"enterprise_industry_analysis","score":0.0,"tokens":2954964,"cost_usd":0.738472,"elapsed_time":986.34,"gpr":0.6667,"tpe":0.2105,"ee":0.0429},{"task_id":"task_144_enterprise_industry_analysis_easy_easy074","category":"enterprise_industry_analysis","score":1.0,"tokens":1347376,"cost_usd":0.318474,"elapsed_time":736.42,"gpr":1.0,"tpe":0.0042,"ee":0.0545},{"task_id":"task_145_enterprise_industry_analysis_easy_easy075","category":"enterprise_industry_analysis","score":1.0,"tokens":202731,"cost_usd":0.047591,"elapsed_time":181.06,"gpr":1.0,"tpe":0.7291,"ee":0.25},{"task_id":"task_146_enterprise_industry_analysis_easy_easy076","category":"enterprise_industry_analysis","score":1.0,"tokens":393776,"cost_usd":0.096088,"elapsed_time":386.51,"gpr":1.0,"tpe":0.4566,"ee":0.125},{"task_id":"task_147_enterprise_industry_analysis_easy_easy077","category":"enterprise_industry_analysis","score":0.0,"tokens":397631,"cost_usd":0.101076,"elapsed_time":342.36,"gpr":1.0,"tpe":0.3875,"ee":0.15},{"task_id":"task_148_enterprise_industry_analysis_easy_easy078","category":"enterprise_industry_analysis","score":1.0,"tokens":188200,"cost_usd":0.043997,"elapsed_time":226.26,"gpr":1.0,"tpe":0.5908,"ee":0.25},{"task_id":"task_149_enterprise_industry_analysis_easy_easy079","category":"enterprise_industry_analysis","score":1.0,"tokens":187493,"cost_usd":0.041351,"elapsed_time":168.26,"gpr":1.0,"tpe":0.6052,"ee":0.2727},{"task_id":"task_150_enterprise_industry_analysis_easy_easy080","category":"enterprise_industry_analysis","score":1.0,"tokens":182589,"cost_usd":0.044922,"elapsed_time":238.42,"gpr":1.0,"tpe":0.4589,"ee":0.2143},{"task_id":"task_151_enterprise_industry_analysis_easy_easy081","category":"enterprise_industry_analysis","score":0.0,"tokens":395861,"cost_usd":0.088959,"elapsed_time":278.75,"gpr":0.6667,"tpe":0.729,"ee":0.1034},{"task_id":"task_152_enterprise_industry_analysis_easy_easy082","category":"enterprise_industry_analysis","score":0.0,"tokens":371110,"cost_usd":0.094079,"elapsed_time":242.66,"gpr":1.0,"tpe":0.4725,"ee":0.1875},{"task_id":"task_153_enterprise_industry_analysis_easy_easy083","category":"enterprise_industry_analysis","score":1.0,"tokens":2527213,"cost_usd":0.635101,"elapsed_time":467.49,"gpr":1.0,"tpe":0.1845,"ee":0.0909},{"task_id":"task_154_enterprise_industry_analysis_easy_easy084","category":"enterprise_industry_analysis","score":0.0,"tokens":46903,"cost_usd":0.008106,"elapsed_time":151.95,"gpr":0.3333,"tpe":1.0,"ee":0.75},{"task_id":"task_155_enterprise_industry_analysis_easy_easy085","category":"enterprise_industry_analysis","score":1.0,"tokens":1551124,"cost_usd":0.391294,"elapsed_time":271.86,"gpr":1.0,"tpe":0.186,"ee":0.15},{"task_id":"task_156_enterprise_industry_analysis_easy_easy086","category":"enterprise_industry_analysis","score":1.0,"tokens":893015,"cost_usd":0.224071,"elapsed_time":193.65,"gpr":1.0,"tpe":0.7608,"ee":0.2},{"task_id":"task_157_enterprise_industry_analysis_easy_easy088","category":"enterprise_industry_analysis","score":1.0,"tokens":532022,"cost_usd":0.134547,"elapsed_time":179.27,"gpr":1.0,"tpe":0.2288,"ee":0.125},{"task_id":"task_158_enterprise_industry_analysis_easy_easy090","category":"enterprise_industry_analysis","score":0.0,"tokens":2302849,"cost_usd":0.579198,"elapsed_time":497.98,"gpr":0.6667,"tpe":0.0193,"ee":0.06},{"task_id":"task_159_enterprise_industry_analysis_easy_easy091","category":"enterprise_industry_analysis","score":0.0,"tokens":152235,"cost_usd":0.032391,"elapsed_time":145.39,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_160_enterprise_industry_analysis_easy_easy092","category":"enterprise_industry_analysis","score":0.0,"tokens":215322,"cost_usd":0.048335,"elapsed_time":136.9,"gpr":0.0,"tpe":0.0,"ee":0.1875},{"task_id":"task_161_enterprise_industry_analysis_easy_easy093","category":"enterprise_industry_analysis","score":1.0,"tokens":520886,"cost_usd":0.111921,"elapsed_time":581.22,"gpr":1.0,"tpe":0.0471,"ee":0.0938},{"task_id":"task_162_enterprise_industry_analysis_easy_easy094","category":"enterprise_industry_analysis","score":0.0,"tokens":1145091,"cost_usd":0.289016,"elapsed_time":277.5,"gpr":0.0,"tpe":0.0,"ee":0.12},{"task_id":"task_163_enterprise_industry_analysis_easy_easy095","category":"enterprise_industry_analysis","score":0.0,"tokens":1235433,"cost_usd":0.310524,"elapsed_time":267.66,"gpr":0.0,"tpe":0.0,"ee":0.125},{"task_id":"task_164_enterprise_industry_analysis_easy_easy096","category":"enterprise_industry_analysis","score":0.0,"tokens":414645,"cost_usd":0.105116,"elapsed_time":289.01,"gpr":0.0,"tpe":0.0,"ee":0.125},{"task_id":"task_165_enterprise_industry_analysis_easy_easy097","category":"enterprise_industry_analysis","score":0.0,"tokens":696325,"cost_usd":0.173796,"elapsed_time":288.16,"gpr":0.0,"tpe":0.0,"ee":0.1154},{"task_id":"task_166_enterprise_industry_analysis_easy_easy098","category":"enterprise_industry_analysis","score":1.0,"tokens":312578,"cost_usd":0.07786,"elapsed_time":160.67,"gpr":1.0,"tpe":0.2288,"ee":0.1765},{"task_id":"task_167_enterprise_industry_analysis_easy_easy099","category":"enterprise_industry_analysis","score":1.0,"tokens":1299367,"cost_usd":0.328091,"elapsed_time":307.54,"gpr":1.0,"tpe":0.1216,"ee":0.1304},{"task_id":"task_168_enterprise_industry_analysis_easy_easy100","category":"enterprise_industry_analysis","score":0.0,"tokens":470180,"cost_usd":0.11757,"elapsed_time":188.66,"gpr":0.3333,"tpe":0.2288,"ee":0.1667},{"task_id":"task_169_enterprise_industry_analysis_easy_easy101","category":"enterprise_industry_analysis","score":1.0,"tokens":618396,"cost_usd":0.15644,"elapsed_time":209.91,"gpr":1.0,"tpe":0.1853,"ee":0.1579},{"task_id":"task_170_enterprise_industry_analysis_easy_easy102","category":"enterprise_industry_analysis","score":0.0,"tokens":843769,"cost_usd":0.213279,"elapsed_time":260.52,"gpr":0.0,"tpe":0.0,"ee":0.125},{"task_id":"task_171_enterprise_industry_analysis_easy_easy103","category":"enterprise_industry_analysis","score":1.0,"tokens":446691,"cost_usd":0.107248,"elapsed_time":121.93,"gpr":1.0,"tpe":0.3874,"ee":0.25},{"task_id":"task_172_enterprise_industry_analysis_easy_easy104","category":"enterprise_industry_analysis","score":1.0,"tokens":184544,"cost_usd":0.046853,"elapsed_time":118.56,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_173_enterprise_industry_analysis_easy_easy105","category":"enterprise_industry_analysis","score":0.0,"tokens":782791,"cost_usd":0.197917,"elapsed_time":266.03,"gpr":0.0,"tpe":0.0,"ee":0.125},{"task_id":"task_174_enterprise_industry_analysis_easy_easy106","category":"enterprise_industry_analysis","score":0.0,"tokens":412980,"cost_usd":0.104772,"elapsed_time":226.03,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_175_enterprise_industry_analysis_easy_easy107","category":"enterprise_industry_analysis","score":0.0,"tokens":983709,"cost_usd":0.239823,"elapsed_time":409.31,"gpr":0.0,"tpe":0.0,"ee":0.0769},{"task_id":"task_176_enterprise_industry_analysis_easy_easy108","category":"enterprise_industry_analysis","score":1.0,"tokens":485213,"cost_usd":0.122792,"elapsed_time":233.49,"gpr":1.0,"tpe":0.1216,"ee":0.1304},{"task_id":"task_177_enterprise_industry_analysis_easy_easy109","category":"enterprise_industry_analysis","score":0.0,"tokens":1598781,"cost_usd":0.402257,"elapsed_time":427.18,"gpr":0.0,"tpe":0.0,"ee":0.125},{"task_id":"task_178_enterprise_industry_analysis_easy_easy110","category":"enterprise_industry_analysis","score":0.0,"tokens":499750,"cost_usd":0.126442,"elapsed_time":220.92,"gpr":0.0,"tpe":0.0,"ee":0.2143},{"task_id":"task_179_enterprise_industry_analysis_easy_easy111","category":"enterprise_industry_analysis","score":0.0,"tokens":898160,"cost_usd":0.225837,"elapsed_time":371.02,"gpr":0.0,"tpe":0.0,"ee":0.0811},{"task_id":"task_180_enterprise_industry_analysis_easy_easy112","category":"enterprise_industry_analysis","score":0.0,"tokens":444309,"cost_usd":0.108757,"elapsed_time":237.97,"gpr":0.0,"tpe":0.0,"ee":0.1364},{"task_id":"task_181_enterprise_industry_analysis_easy_easy113","category":"enterprise_industry_analysis","score":1.0,"tokens":867039,"cost_usd":0.20577,"elapsed_time":240.58,"gpr":1.0,"tpe":0.1668,"ee":0.15},{"task_id":"task_182_enterprise_industry_analysis_easy_easy114","category":"enterprise_industry_analysis","score":0.0,"tokens":1351058,"cost_usd":0.340997,"elapsed_time":297.09,"gpr":0.0,"tpe":0.0,"ee":0.1579},{"task_id":"task_183_enterprise_industry_analysis_easy_easy115","category":"enterprise_industry_analysis","score":1.0,"tokens":1619227,"cost_usd":0.390599,"elapsed_time":362.56,"gpr":1.0,"tpe":0.0886,"ee":0.1154},{"task_id":"task_184_enterprise_industry_analysis_easy_easy116","category":"enterprise_industry_analysis","score":1.0,"tokens":594966,"cost_usd":0.131829,"elapsed_time":208.3,"gpr":1.0,"tpe":0.0985,"ee":0.12},{"task_id":"task_185_enterprise_industry_analysis_easy_easy117","category":"enterprise_industry_analysis","score":0.0,"tokens":751545,"cost_usd":0.189896,"elapsed_time":221.76,"gpr":0.0,"tpe":0.0,"ee":0.1765},{"task_id":"task_186_enterprise_industry_analysis_medium_medium001","category":"enterprise_industry_analysis","score":1.0,"tokens":1738823,"cost_usd":0.433619,"elapsed_time":441.93,"gpr":1.0,"tpe":0.0738,"ee":0.1379},{"task_id":"task_187_enterprise_industry_analysis_medium_medium002","category":"enterprise_industry_analysis","score":0.0,"tokens":6079727,"cost_usd":1.516425,"elapsed_time":1207.14,"gpr":0.25,"tpe":0.0182,"ee":0.0488},{"task_id":"task_188_enterprise_industry_analysis_medium_medium003","category":"enterprise_industry_analysis","score":0.0,"tokens":4743611,"cost_usd":1.197103,"elapsed_time":885.98,"gpr":0.75,"tpe":0.0097,"ee":0.0714},{"task_id":"task_189_enterprise_industry_analysis_medium_medium004","category":"enterprise_industry_analysis","score":0.0,"tokens":1356660,"cost_usd":0.339162,"elapsed_time":638.37,"gpr":0.5,"tpe":0.0097,"ee":0.069},{"task_id":"task_190_enterprise_industry_analysis_medium_medium005","category":"enterprise_industry_analysis","score":0.0,"tokens":3948909,"cost_usd":0.976076,"elapsed_time":770.5,"gpr":0.0,"tpe":0.0,"ee":0.0727},{"task_id":"task_191_enterprise_industry_analysis_medium_medium006","category":"enterprise_industry_analysis","score":0.0,"tokens":1481161,"cost_usd":0.361437,"elapsed_time":479.85,"gpr":0.0,"tpe":0.0,"ee":0.0727},{"task_id":"task_192_enterprise_industry_analysis_medium_medium007","category":"enterprise_industry_analysis","score":0.0,"tokens":1440888,"cost_usd":0.360589,"elapsed_time":529.16,"gpr":0.2,"tpe":0.0424,"ee":0.0952},{"task_id":"task_193_enterprise_industry_analysis_medium_medium008","category":"enterprise_industry_analysis","score":1.0,"tokens":1698373,"cost_usd":0.412556,"elapsed_time":396.92,"gpr":1.0,"tpe":0.0535,"ee":0.125},{"task_id":"task_194_enterprise_industry_analysis_medium_medium009","category":"enterprise_industry_analysis","score":1.0,"tokens":939756,"cost_usd":0.238455,"elapsed_time":506.49,"gpr":1.0,"tpe":0.0651,"ee":0.087},{"task_id":"task_195_enterprise_industry_analysis_medium_medium010","category":"enterprise_industry_analysis","score":0.0,"tokens":2126853,"cost_usd":0.510764,"elapsed_time":443.12,"gpr":0.4,"tpe":0.1216,"ee":0.16},{"task_id":"task_196_enterprise_industry_analysis_medium_medium011","category":"enterprise_industry_analysis","score":0.0,"tokens":1463674,"cost_usd":0.365874,"elapsed_time":394.33,"gpr":0.0,"tpe":0.0,"ee":0.1212},{"task_id":"task_197_enterprise_industry_analysis_medium_medium012","category":"enterprise_industry_analysis","score":0.0,"tokens":1463655,"cost_usd":0.369566,"elapsed_time":436.0,"gpr":0.0,"tpe":0.0,"ee":0.1026},{"task_id":"task_198_enterprise_industry_analysis_medium_medium013","category":"enterprise_industry_analysis","score":0.0,"tokens":1585271,"cost_usd":0.390586,"elapsed_time":634.56,"gpr":0.0,"tpe":0.0,"ee":0.0615},{"task_id":"task_199_enterprise_industry_analysis_medium_medium014","category":"enterprise_industry_analysis","score":1.0,"tokens":1004624,"cost_usd":0.242446,"elapsed_time":374.72,"gpr":1.0,"tpe":0.0635,"ee":0.1316},{"task_id":"task_200_enterprise_industry_analysis_medium_medium015","category":"enterprise_industry_analysis","score":1.0,"tokens":590979,"cost_usd":0.148499,"elapsed_time":302.94,"gpr":1.0,"tpe":0.0471,"ee":0.1471},{"task_id":"task_201_enterprise_industry_analysis_medium_medium016","category":"enterprise_industry_analysis","score":1.0,"tokens":1320918,"cost_usd":0.333942,"elapsed_time":415.26,"gpr":1.0,"tpe":0.1919,"ee":0.1667},{"task_id":"task_202_enterprise_industry_analysis_medium_medium017","category":"enterprise_industry_analysis","score":1.0,"tokens":683218,"cost_usd":0.173493,"elapsed_time":365.61,"gpr":1.0,"tpe":0.0424,"ee":0.1429},{"task_id":"task_203_enterprise_industry_analysis_medium_medium018","category":"enterprise_industry_analysis","score":0.0,"tokens":2153255,"cost_usd":0.532202,"elapsed_time":561.82,"gpr":0.4,"tpe":0.0513,"ee":0.1136},{"task_id":"task_204_enterprise_industry_analysis_medium_medium019","category":"enterprise_industry_analysis","score":1.0,"tokens":1023868,"cost_usd":0.259404,"elapsed_time":433.35,"gpr":1.0,"tpe":0.1238,"ee":0.1471},{"task_id":"task_205_enterprise_industry_analysis_medium_medium020","category":"enterprise_industry_analysis","score":1.0,"tokens":827088,"cost_usd":0.200212,"elapsed_time":473.08,"gpr":1.0,"tpe":0.0788,"ee":0.1064},{"task_id":"task_206_enterprise_industry_analysis_medium_medium021","category":"enterprise_industry_analysis","score":0.0,"tokens":1513486,"cost_usd":0.371507,"elapsed_time":528.08,"gpr":0.2,"tpe":0.0718,"ee":0.1515},{"task_id":"task_207_enterprise_industry_analysis_medium_medium022","category":"enterprise_industry_analysis","score":0.0,"tokens":1006784,"cost_usd":0.255199,"elapsed_time":434.88,"gpr":0.6,"tpe":0.4131,"ee":0.1389},{"task_id":"task_208_enterprise_industry_analysis_medium_medium023","category":"enterprise_industry_analysis","score":1.0,"tokens":404501,"cost_usd":0.098959,"elapsed_time":257.33,"gpr":1.0,"tpe":0.2712,"ee":0.2174},{"task_id":"task_209_enterprise_industry_analysis_medium_medium024","category":"enterprise_industry_analysis","score":1.0,"tokens":3722896,"cost_usd":0.896186,"elapsed_time":723.72,"gpr":1.0,"tpe":0.0097,"ee":0.102},{"task_id":"task_210_enterprise_industry_analysis_medium_medium025","category":"enterprise_industry_analysis","score":0.0,"tokens":2345795,"cost_usd":0.590558,"elapsed_time":487.45,"gpr":1.0,"tpe":0.8021,"ee":0.1389},{"task_id":"task_211_enterprise_industry_analysis_medium_medium026","category":"enterprise_industry_analysis","score":1.0,"tokens":1893856,"cost_usd":0.463605,"elapsed_time":574.71,"gpr":1.0,"tpe":0.0332,"ee":0.098},{"task_id":"task_212_enterprise_industry_analysis_medium_medium027","category":"enterprise_industry_analysis","score":1.0,"tokens":1131865,"cost_usd":0.286162,"elapsed_time":332.85,"gpr":1.0,"tpe":0.354,"ee":0.1667},{"task_id":"task_213_enterprise_industry_analysis_medium_medium028","category":"enterprise_industry_analysis","score":0.0,"tokens":732045,"cost_usd":0.183302,"elapsed_time":183.7,"gpr":0.0,"tpe":0.0,"ee":0.3846},{"task_id":"task_214_enterprise_industry_analysis_medium_medium029","category":"enterprise_industry_analysis","score":0.0,"tokens":1474166,"cost_usd":0.372978,"elapsed_time":489.58,"gpr":0.4,"tpe":1.0,"ee":0.1282},{"task_id":"task_215_enterprise_industry_analysis_medium_medium030","category":"enterprise_industry_analysis","score":1.0,"tokens":1396611,"cost_usd":0.34734,"elapsed_time":646.01,"gpr":1.0,"tpe":0.098,"ee":0.125},{"task_id":"task_216_enterprise_industry_analysis_medium_medium031","category":"enterprise_industry_analysis","score":1.0,"tokens":517126,"cost_usd":0.130009,"elapsed_time":316.51,"gpr":1.0,"tpe":0.3772,"ee":0.1852},{"task_id":"task_217_enterprise_industry_analysis_medium_medium032","category":"enterprise_industry_analysis","score":0.0,"tokens":655906,"cost_usd":0.166574,"elapsed_time":361.78,"gpr":0.6,"tpe":0.5979,"ee":0.1613},{"task_id":"task_218_enterprise_industry_analysis_medium_medium033","category":"enterprise_industry_analysis","score":1.0,"tokens":404259,"cost_usd":0.097308,"elapsed_time":276.94,"gpr":1.0,"tpe":0.4078,"ee":0.2381},{"task_id":"task_219_enterprise_industry_analysis_medium_medium034","category":"enterprise_industry_analysis","score":1.0,"tokens":282620,"cost_usd":0.071688,"elapsed_time":174.2,"gpr":1.0,"tpe":0.2242,"ee":0.25},{"task_id":"task_220_enterprise_industry_analysis_medium_medium035","category":"enterprise_industry_analysis","score":0.0,"tokens":1116348,"cost_usd":0.257105,"elapsed_time":495.05,"gpr":0.4,"tpe":0.0482,"ee":0.1163},{"task_id":"task_221_enterprise_industry_analysis_medium_medium036","category":"enterprise_industry_analysis","score":0.0,"tokens":2831761,"cost_usd":0.714006,"elapsed_time":750.9,"gpr":0.4,"tpe":0.0164,"ee":0.1},{"task_id":"task_222_enterprise_industry_analysis_medium_medium037","category":"enterprise_industry_analysis","score":1.0,"tokens":953913,"cost_usd":0.23297,"elapsed_time":434.53,"gpr":1.0,"tpe":0.2806,"ee":0.1136},{"task_id":"task_223_enterprise_industry_analysis_medium_medium038","category":"enterprise_industry_analysis","score":0.0,"tokens":2420014,"cost_usd":0.605307,"elapsed_time":540.25,"gpr":0.4,"tpe":0.0424,"ee":0.122},{"task_id":"task_224_enterprise_industry_analysis_medium_medium039","category":"enterprise_industry_analysis","score":1.0,"tokens":1131082,"cost_usd":0.281616,"elapsed_time":383.05,"gpr":1.0,"tpe":0.1691,"ee":0.1471},{"task_id":"task_225_enterprise_industry_analysis_medium_medium040","category":"enterprise_industry_analysis","score":1.0,"tokens":611979,"cost_usd":0.151786,"elapsed_time":383.44,"gpr":1.0,"tpe":0.1599,"ee":0.1667},{"task_id":"task_226_enterprise_industry_analysis_medium_medium041","category":"enterprise_industry_analysis","score":1.0,"tokens":396699,"cost_usd":0.096804,"elapsed_time":266.81,"gpr":1.0,"tpe":0.2288,"ee":0.1923},{"task_id":"task_227_enterprise_industry_analysis_medium_medium042","category":"enterprise_industry_analysis","score":0.0,"tokens":597248,"cost_usd":0.141719,"elapsed_time":313.52,"gpr":0.4,"tpe":0.1223,"ee":0.1471},{"task_id":"task_228_enterprise_industry_analysis_medium_medium043","category":"enterprise_industry_analysis","score":1.0,"tokens":456156,"cost_usd":0.114802,"elapsed_time":300.12,"gpr":1.0,"tpe":0.4749,"ee":0.1852},{"task_id":"task_229_enterprise_industry_analysis_medium_medium044","category":"enterprise_industry_analysis","score":0.0,"tokens":1335056,"cost_usd":0.323903,"elapsed_time":386.41,"gpr":0.8,"tpe":0.4432,"ee":0.1724},{"task_id":"task_230_enterprise_industry_analysis_medium_medium045","category":"enterprise_industry_analysis","score":0.0,"tokens":1065992,"cost_usd":0.260485,"elapsed_time":475.38,"gpr":0.4,"tpe":0.0447,"ee":0.1064},{"task_id":"task_231_enterprise_industry_analysis_medium_medium046","category":"enterprise_industry_analysis","score":0.0,"tokens":1394739,"cost_usd":0.333925,"elapsed_time":556.8,"gpr":0.4,"tpe":0.0351,"ee":0.1042},{"task_id":"task_232_enterprise_industry_analysis_medium_medium047","category":"enterprise_industry_analysis","score":1.0,"tokens":1104795,"cost_usd":0.276679,"elapsed_time":338.66,"gpr":1.0,"tpe":0.0787,"ee":0.1667},{"task_id":"task_233_enterprise_industry_analysis_medium_medium048","category":"enterprise_industry_analysis","score":1.0,"tokens":1949338,"cost_usd":0.49335,"elapsed_time":684.11,"gpr":1.0,"tpe":0.0225,"ee":0.098},{"task_id":"task_234_enterprise_industry_analysis_medium_medium049","category":"enterprise_industry_analysis","score":1.0,"tokens":902198,"cost_usd":0.228724,"elapsed_time":502.16,"gpr":1.0,"tpe":0.0244,"ee":0.1136},{"task_id":"task_235_enterprise_industry_analysis_medium_medium050","category":"enterprise_industry_analysis","score":0.0,"tokens":1384317,"cost_usd":0.342307,"elapsed_time":613.84,"gpr":0.8,"tpe":0.03,"ee":0.0909},{"task_id":"task_236_enterprise_industry_analysis_medium_medium051","category":"enterprise_industry_analysis","score":0.0,"tokens":1280233,"cost_usd":0.324537,"elapsed_time":570.35,"gpr":0.8,"tpe":0.2249,"ee":0.1136},{"task_id":"task_237_enterprise_industry_analysis_medium_medium052","category":"enterprise_industry_analysis","score":1.0,"tokens":2988531,"cost_usd":0.754138,"elapsed_time":498.72,"gpr":1.0,"tpe":0.2028,"ee":0.1923},{"task_id":"task_238_enterprise_industry_analysis_medium_medium053","category":"enterprise_industry_analysis","score":0.0,"tokens":4506887,"cost_usd":1.138012,"elapsed_time":884.5,"gpr":0.4,"tpe":0.0067,"ee":0.0893},{"task_id":"task_239_enterprise_industry_analysis_medium_medium054","category":"enterprise_industry_analysis","score":1.0,"tokens":720797,"cost_usd":0.175644,"elapsed_time":326.95,"gpr":1.0,"tpe":0.1473,"ee":0.1786},{"task_id":"task_240_enterprise_industry_analysis_medium_medium055","category":"enterprise_industry_analysis","score":1.0,"tokens":434552,"cost_usd":0.103843,"elapsed_time":272.47,"gpr":1.0,"tpe":0.1227,"ee":0.1923},{"task_id":"task_241_enterprise_industry_analysis_medium_medium056","category":"enterprise_industry_analysis","score":1.0,"tokens":1699104,"cost_usd":0.421503,"elapsed_time":534.54,"gpr":1.0,"tpe":0.0096,"ee":0.098},{"task_id":"task_242_enterprise_industry_analysis_medium_medium057","category":"enterprise_industry_analysis","score":0.0,"tokens":2979735,"cost_usd":0.723096,"elapsed_time":530.07,"gpr":0.0,"tpe":0.0,"ee":0.1562},{"task_id":"task_243_enterprise_industry_analysis_medium_medium058","category":"enterprise_industry_analysis","score":0.0,"tokens":603184,"cost_usd":0.146055,"elapsed_time":334.25,"gpr":0.8,"tpe":0.104,"ee":0.1515},{"task_id":"task_244_enterprise_industry_analysis_medium_medium059","category":"enterprise_industry_analysis","score":1.0,"tokens":643146,"cost_usd":0.160461,"elapsed_time":319.69,"gpr":1.0,"tpe":0.148,"ee":0.1562},{"task_id":"task_245_enterprise_industry_analysis_medium_medium060","category":"enterprise_industry_analysis","score":0.0,"tokens":2206334,"cost_usd":0.554973,"elapsed_time":332.87,"gpr":0.0,"tpe":0.0,"ee":0.2381},{"task_id":"task_246_enterprise_industry_analysis_medium_medium061","category":"enterprise_industry_analysis","score":0.0,"tokens":1606748,"cost_usd":0.392426,"elapsed_time":317.49,"gpr":0.0,"tpe":0.0,"ee":0.1724},{"task_id":"task_247_enterprise_industry_analysis_medium_medium062","category":"enterprise_industry_analysis","score":0.0,"tokens":3164425,"cost_usd":0.778875,"elapsed_time":717.52,"gpr":0.0,"tpe":0.0,"ee":0.0758},{"task_id":"task_248_enterprise_industry_analysis_medium_medium063","category":"enterprise_industry_analysis","score":0.0,"tokens":560877,"cost_usd":0.141255,"elapsed_time":333.31,"gpr":0.5,"tpe":0.2351,"ee":0.1333},{"task_id":"task_249_enterprise_industry_analysis_medium_medium064","category":"enterprise_industry_analysis","score":0.0,"tokens":2702380,"cost_usd":0.648711,"elapsed_time":493.59,"gpr":0.0,"tpe":0.0,"ee":0.1026},{"task_id":"task_250_enterprise_industry_analysis_medium_medium065","category":"enterprise_industry_analysis","score":1.0,"tokens":275575,"cost_usd":0.061122,"elapsed_time":157.62,"gpr":1.0,"tpe":0.5408,"ee":0.2},{"task_id":"task_251_enterprise_industry_analysis_medium_medium066","category":"enterprise_industry_analysis","score":1.0,"tokens":382632,"cost_usd":0.092812,"elapsed_time":289.23,"gpr":1.0,"tpe":0.2342,"ee":0.1739},{"task_id":"task_252_enterprise_industry_analysis_medium_medium067","category":"enterprise_industry_analysis","score":0.0,"tokens":496228,"cost_usd":0.124519,"elapsed_time":223.61,"gpr":0.75,"tpe":0.4524,"ee":0.1667},{"task_id":"task_253_enterprise_industry_analysis_medium_medium068","category":"enterprise_industry_analysis","score":0.0,"tokens":1297600,"cost_usd":0.315408,"elapsed_time":324.08,"gpr":0.0,"tpe":0.0,"ee":0.1176},{"task_id":"task_254_enterprise_industry_analysis_medium_medium069","category":"enterprise_industry_analysis","score":0.0,"tokens":618150,"cost_usd":0.149661,"elapsed_time":223.94,"gpr":0.0,"tpe":0.0,"ee":0.1739},{"task_id":"task_255_enterprise_industry_analysis_medium_medium070","category":"enterprise_industry_analysis","score":0.0,"tokens":301026,"cost_usd":0.067501,"elapsed_time":252.21,"gpr":0.3333,"tpe":0.2542,"ee":0.2},{"task_id":"task_256_enterprise_industry_analysis_medium_medium071","category":"enterprise_industry_analysis","score":0.0,"tokens":1871704,"cost_usd":0.473431,"elapsed_time":544.87,"gpr":0.25,"tpe":0.1216,"ee":0.093},{"task_id":"task_257_enterprise_industry_analysis_medium_medium072","category":"enterprise_industry_analysis","score":0.0,"tokens":1907326,"cost_usd":0.470065,"elapsed_time":464.2,"gpr":0.25,"tpe":0.1853,"ee":0.1379},{"task_id":"task_258_enterprise_industry_analysis_medium_medium073","category":"enterprise_industry_analysis","score":1.0,"tokens":494423,"cost_usd":0.115835,"elapsed_time":348.03,"gpr":1.0,"tpe":0.1606,"ee":0.1333},{"task_id":"task_259_enterprise_industry_analysis_medium_medium074","category":"enterprise_industry_analysis","score":0.0,"tokens":2096758,"cost_usd":0.517099,"elapsed_time":738.22,"gpr":0.5,"tpe":0.284,"ee":0.0741},{"task_id":"task_260_enterprise_industry_analysis_medium_medium075","category":"enterprise_industry_analysis","score":1.0,"tokens":1218358,"cost_usd":0.291383,"elapsed_time":313.35,"gpr":1.0,"tpe":0.3145,"ee":0.1481},{"task_id":"task_261_enterprise_industry_analysis_medium_medium076","category":"enterprise_industry_analysis","score":1.0,"tokens":1511278,"cost_usd":0.375674,"elapsed_time":600.18,"gpr":1.0,"tpe":0.0235,"ee":0.0769},{"task_id":"task_262_enterprise_industry_analysis_medium_medium077","category":"enterprise_industry_analysis","score":1.0,"tokens":1588377,"cost_usd":0.401178,"elapsed_time":346.85,"gpr":1.0,"tpe":0.3724,"ee":0.1538},{"task_id":"task_263_enterprise_industry_analysis_medium_medium078","category":"enterprise_industry_analysis","score":1.0,"tokens":2373843,"cost_usd":0.553073,"elapsed_time":456.28,"gpr":1.0,"tpe":0.0646,"ee":0.1333},{"task_id":"task_264_enterprise_industry_analysis_medium_medium079","category":"enterprise_industry_analysis","score":1.0,"tokens":511429,"cost_usd":0.125987,"elapsed_time":285.78,"gpr":1.0,"tpe":0.3745,"ee":0.1538},{"task_id":"task_265_enterprise_industry_analysis_medium_medium080","category":"enterprise_industry_analysis","score":0.0,"tokens":1218895,"cost_usd":0.308394,"elapsed_time":439.69,"gpr":0.0,"tpe":0.0,"ee":0.087},{"task_id":"task_266_enterprise_industry_analysis_medium_medium081","category":"enterprise_industry_analysis","score":1.0,"tokens":414933,"cost_usd":0.102314,"elapsed_time":203.93,"gpr":1.0,"tpe":0.5891,"ee":0.2222},{"task_id":"task_267_enterprise_industry_analysis_medium_medium082","category":"enterprise_industry_analysis","score":1.0,"tokens":538963,"cost_usd":0.13658,"elapsed_time":243.08,"gpr":1.0,"tpe":0.1216,"ee":0.1667},{"task_id":"task_268_enterprise_industry_analysis_medium_medium083","category":"enterprise_industry_analysis","score":1.0,"tokens":1641297,"cost_usd":0.409385,"elapsed_time":457.26,"gpr":1.0,"tpe":0.0559,"ee":0.087},{"task_id":"task_269_enterprise_industry_analysis_medium_medium084","category":"enterprise_industry_analysis","score":0.0,"tokens":1553000,"cost_usd":0.37754,"elapsed_time":383.4,"gpr":0.5,"tpe":0.2939,"ee":0.129},{"task_id":"task_270_enterprise_industry_analysis_medium_medium085","category":"enterprise_industry_analysis","score":1.0,"tokens":261987,"cost_usd":0.064288,"elapsed_time":156.95,"gpr":1.0,"tpe":0.8372,"ee":0.2857},{"task_id":"task_271_enterprise_industry_analysis_medium_medium086","category":"enterprise_industry_analysis","score":1.0,"tokens":1362165,"cost_usd":0.342785,"elapsed_time":354.77,"gpr":1.0,"tpe":0.395,"ee":0.1379},{"task_id":"task_272_enterprise_industry_analysis_medium_medium087","category":"enterprise_industry_analysis","score":1.0,"tokens":933853,"cost_usd":0.236308,"elapsed_time":293.02,"gpr":1.0,"tpe":0.2854,"ee":0.1667},{"task_id":"task_273_enterprise_industry_analysis_medium_medium088","category":"enterprise_industry_analysis","score":1.0,"tokens":2038199,"cost_usd":0.511822,"elapsed_time":416.11,"gpr":1.0,"tpe":0.0486,"ee":0.1143},{"task_id":"task_274_enterprise_industry_analysis_medium_medium089","category":"enterprise_industry_analysis","score":1.0,"tokens":4060031,"cost_usd":1.024584,"elapsed_time":682.11,"gpr":1.0,"tpe":0.0617,"ee":0.0889},{"task_id":"task_275_enterprise_industry_analysis_medium_medium090","category":"enterprise_industry_analysis","score":0.0,"tokens":1426083,"cost_usd":0.359762,"elapsed_time":264.14,"gpr":0.25,"tpe":0.81,"ee":0.2105},{"task_id":"task_276_enterprise_industry_analysis_medium_medium091","category":"enterprise_industry_analysis","score":1.0,"tokens":3325215,"cost_usd":0.805439,"elapsed_time":638.09,"gpr":1.0,"tpe":0.0208,"ee":0.0667},{"task_id":"task_277_enterprise_industry_analysis_medium_medium092","category":"enterprise_industry_analysis","score":0.0,"tokens":1198538,"cost_usd":0.301423,"elapsed_time":293.59,"gpr":0.25,"tpe":0.1216,"ee":0.1667},{"task_id":"task_278_enterprise_industry_analysis_medium_medium093","category":"enterprise_industry_analysis","score":1.0,"tokens":867247,"cost_usd":0.212356,"elapsed_time":357.72,"gpr":1.0,"tpe":0.3206,"ee":0.129},{"task_id":"task_279_enterprise_industry_analysis_medium_medium094","category":"enterprise_industry_analysis","score":0.0,"tokens":2012913,"cost_usd":0.494613,"elapsed_time":306.45,"gpr":0.0,"tpe":0.0,"ee":0.1538},{"task_id":"task_280_enterprise_industry_analysis_medium_medium095","category":"enterprise_industry_analysis","score":0.0,"tokens":1946126,"cost_usd":0.459288,"elapsed_time":451.54,"gpr":0.5,"tpe":0.1959,"ee":0.1143},{"task_id":"task_281_enterprise_industry_analysis_medium_medium096","category":"enterprise_industry_analysis","score":1.0,"tokens":778256,"cost_usd":0.197476,"elapsed_time":388.85,"gpr":1.0,"tpe":0.0382,"ee":0.1026},{"task_id":"task_282_enterprise_industry_analysis_medium_medium097","category":"enterprise_industry_analysis","score":1.0,"tokens":576850,"cost_usd":0.146116,"elapsed_time":231.33,"gpr":1.0,"tpe":0.5108,"ee":0.1667},{"task_id":"task_283_enterprise_industry_analysis_medium_medium098","category":"enterprise_industry_analysis","score":0.0,"tokens":739052,"cost_usd":0.180255,"elapsed_time":347.68,"gpr":0.25,"tpe":0.0718,"ee":0.1111},{"task_id":"task_284_enterprise_industry_analysis_medium_medium099","category":"enterprise_industry_analysis","score":0.0,"tokens":904558,"cost_usd":0.229516,"elapsed_time":384.54,"gpr":0.25,"tpe":0.0886,"ee":0.129},{"task_id":"task_285_enterprise_industry_analysis_medium_medium100","category":"enterprise_industry_analysis","score":0.0,"tokens":1969258,"cost_usd":0.487238,"elapsed_time":537.14,"gpr":0.0,"tpe":0.0,"ee":0.0909},{"task_id":"task_286_enterprise_industry_analysis_medium_medium101","category":"enterprise_industry_analysis","score":0.0,"tokens":446223,"cost_usd":0.106295,"elapsed_time":279.24,"gpr":0.5,"tpe":0.3786,"ee":0.1481},{"task_id":"task_287_enterprise_industry_analysis_medium_medium102","category":"enterprise_industry_analysis","score":0.0,"tokens":2106429,"cost_usd":0.528369,"elapsed_time":547.37,"gpr":0.25,"tpe":0.0309,"ee":0.0851},{"task_id":"task_288_enterprise_industry_analysis_medium_medium103","category":"enterprise_industry_analysis","score":0.0,"tokens":1153205,"cost_usd":0.289878,"elapsed_time":244.96,"gpr":0.25,"tpe":0.1351,"ee":0.1667},{"task_id":"task_289_enterprise_industry_analysis_medium_medium104","category":"enterprise_industry_analysis","score":0.0,"tokens":538663,"cost_usd":0.126858,"elapsed_time":319.54,"gpr":0.0,"tpe":0.0,"ee":0.1379},{"task_id":"task_290_enterprise_industry_analysis_medium_medium105","category":"enterprise_industry_analysis","score":0.0,"tokens":2466515,"cost_usd":0.604318,"elapsed_time":676.73,"gpr":0.25,"tpe":0.0057,"ee":0.0597},{"task_id":"task_291_enterprise_industry_analysis_medium_medium106","category":"enterprise_industry_analysis","score":0.0,"tokens":312537,"cost_usd":0.075758,"elapsed_time":187.16,"gpr":0.25,"tpe":0.3487,"ee":0.2},{"task_id":"task_292_enterprise_industry_analysis_medium_medium107","category":"enterprise_industry_analysis","score":0.0,"tokens":1091334,"cost_usd":0.267045,"elapsed_time":300.18,"gpr":0.25,"tpe":0.2542,"ee":0.1026},{"task_id":"task_293_enterprise_industry_analysis_medium_medium108","category":"enterprise_industry_analysis","score":0.0,"tokens":290037,"cost_usd":0.071235,"elapsed_time":160.0,"gpr":0.5,"tpe":0.6743,"ee":0.2},{"task_id":"task_294_enterprise_industry_analysis_medium_medium109","category":"enterprise_industry_analysis","score":1.0,"tokens":217969,"cost_usd":0.054046,"elapsed_time":132.66,"gpr":1.0,"tpe":0.4931,"ee":0.3077},{"task_id":"task_295_enterprise_industry_analysis_medium_medium110","category":"enterprise_industry_analysis","score":0.0,"tokens":3118690,"cost_usd":0.756768,"elapsed_time":576.96,"gpr":0.25,"tpe":0.0309,"ee":0.0889},{"task_id":"task_296_enterprise_industry_analysis_medium_medium111","category":"enterprise_industry_analysis","score":0.0,"tokens":351739,"cost_usd":0.078733,"elapsed_time":166.85,"gpr":0.25,"tpe":0.2288,"ee":0.1905},{"task_id":"task_297_enterprise_industry_policy_analysis_easy_easy067","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":400554,"cost_usd":0.08878,"elapsed_time":122.95,"gpr":1.0,"tpe":0.3138,"ee":0.2143},{"task_id":"task_298_enterprise_industry_policy_analysis_easy_easy068","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":243020,"cost_usd":0.061883,"elapsed_time":160.36,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_299_enterprise_industry_policy_analysis_easy_easy069","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":251135,"cost_usd":0.057379,"elapsed_time":138.0,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_300_enterprise_industry_policy_analysis_easy_easy070","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":434100,"cost_usd":0.101705,"elapsed_time":199.31,"gpr":1.0,"tpe":0.1351,"ee":0.1364},{"task_id":"task_301_enterprise_industry_policy_analysis_easy_easy071","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":747228,"cost_usd":0.186689,"elapsed_time":238.1,"gpr":1.0,"tpe":0.1668,"ee":0.15},{"task_id":"task_302_enterprise_industry_policy_analysis_easy_easy072","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":72279,"cost_usd":0.016928,"elapsed_time":52.26,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_303_enterprise_industry_policy_analysis_easy_easy073","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":118629,"cost_usd":0.030044,"elapsed_time":55.82,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_304_enterprise_industry_policy_analysis_easy_easy074","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":753733,"cost_usd":0.187948,"elapsed_time":173.79,"gpr":0.2,"tpe":0.3138,"ee":0.1875},{"task_id":"task_305_enterprise_industry_policy_analysis_easy_easy075","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":321981,"cost_usd":0.081559,"elapsed_time":134.89,"gpr":1.0,"tpe":0.3874,"ee":0.25},{"task_id":"task_306_enterprise_industry_policy_analysis_easy_easy076","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":382658,"cost_usd":0.096457,"elapsed_time":135.43,"gpr":0.3333,"tpe":0.3874,"ee":0.25},{"task_id":"task_307_enterprise_industry_policy_analysis_medium_medium001","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1601624,"cost_usd":0.394649,"elapsed_time":505.05,"gpr":1.0,"tpe":0.0165,"ee":0.1064},{"task_id":"task_308_enterprise_industry_policy_analysis_medium_medium002","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":746515,"cost_usd":0.185668,"elapsed_time":369.94,"gpr":1.0,"tpe":0.0994,"ee":0.1389},{"task_id":"task_309_enterprise_industry_policy_analysis_medium_medium003","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":867764,"cost_usd":0.218252,"elapsed_time":266.21,"gpr":0.8,"tpe":0.4605,"ee":0.2174},{"task_id":"task_310_enterprise_industry_policy_analysis_medium_medium004","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":517644,"cost_usd":0.127146,"elapsed_time":201.25,"gpr":1.0,"tpe":0.1501,"ee":0.2174},{"task_id":"task_311_enterprise_industry_policy_analysis_medium_medium005","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":3111905,"cost_usd":0.763652,"elapsed_time":638.6,"gpr":0.2,"tpe":0.0471,"ee":0.098},{"task_id":"task_312_enterprise_industry_policy_analysis_medium_medium006","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":2013898,"cost_usd":0.490137,"elapsed_time":497.69,"gpr":1.0,"tpe":0.0467,"ee":0.125},{"task_id":"task_313_enterprise_industry_policy_analysis_medium_medium007","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":3554136,"cost_usd":0.859984,"elapsed_time":637.73,"gpr":1.0,"tpe":0.0154,"ee":0.1042},{"task_id":"task_314_enterprise_industry_policy_analysis_medium_medium008","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":972048,"cost_usd":0.242582,"elapsed_time":441.04,"gpr":0.6,"tpe":0.0785,"ee":0.1389},{"task_id":"task_315_enterprise_industry_policy_analysis_medium_medium009","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1888136,"cost_usd":0.476534,"elapsed_time":654.06,"gpr":1.0,"tpe":0.0063,"ee":0.0806},{"task_id":"task_316_enterprise_industry_policy_analysis_medium_medium010","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2582496,"cost_usd":0.635658,"elapsed_time":521.99,"gpr":0.2,"tpe":0.0343,"ee":0.1316},{"task_id":"task_317_enterprise_industry_policy_analysis_medium_medium011","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":511290,"cost_usd":0.129991,"elapsed_time":358.41,"gpr":1.0,"tpe":0.4296,"ee":0.1724},{"task_id":"task_318_enterprise_industry_policy_analysis_medium_medium012","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":563435,"cost_usd":0.139924,"elapsed_time":330.08,"gpr":0.6,"tpe":0.4068,"ee":0.1471},{"task_id":"task_319_enterprise_industry_policy_analysis_medium_medium013","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":808301,"cost_usd":0.20446,"elapsed_time":272.92,"gpr":1.0,"tpe":0.4392,"ee":0.1923},{"task_id":"task_320_enterprise_industry_policy_analysis_medium_medium014","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":507137,"cost_usd":0.124882,"elapsed_time":339.82,"gpr":1.0,"tpe":0.4942,"ee":0.1852},{"task_id":"task_321_enterprise_industry_policy_analysis_medium_medium015","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1305868,"cost_usd":0.322938,"elapsed_time":568.55,"gpr":0.0,"tpe":0.0,"ee":0.102},{"task_id":"task_322_enterprise_industry_policy_analysis_medium_medium016","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":2731087,"cost_usd":0.680244,"elapsed_time":572.42,"gpr":1.0,"tpe":0.0976,"ee":0.1163},{"task_id":"task_323_enterprise_industry_policy_analysis_medium_medium017","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":605157,"cost_usd":0.145751,"elapsed_time":330.19,"gpr":0.2,"tpe":0.2824,"ee":0.2273},{"task_id":"task_324_enterprise_industry_policy_analysis_medium_medium018","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":727992,"cost_usd":0.182067,"elapsed_time":310.75,"gpr":0.6,"tpe":0.4522,"ee":0.1724},{"task_id":"task_325_enterprise_industry_policy_analysis_medium_medium019","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1074794,"cost_usd":0.272137,"elapsed_time":377.11,"gpr":1.0,"tpe":0.3019,"ee":0.1667},{"task_id":"task_326_enterprise_industry_policy_analysis_medium_medium020","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":122037,"cost_usd":0.031075,"elapsed_time":98.7,"gpr":0.8,"tpe":1.0,"ee":0.5556},{"task_id":"task_327_enterprise_industry_policy_analysis_medium_medium021","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1127220,"cost_usd":0.263761,"elapsed_time":451.96,"gpr":0.6,"tpe":0.5891,"ee":0.1316},{"task_id":"task_328_enterprise_industry_policy_analysis_medium_medium022","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":3640755,"cost_usd":0.874245,"elapsed_time":627.9,"gpr":0.4,"tpe":0.5359,"ee":0.119},{"task_id":"task_329_enterprise_industry_policy_analysis_medium_medium023","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1286733,"cost_usd":0.297018,"elapsed_time":369.39,"gpr":1.0,"tpe":0.0903,"ee":0.1515},{"task_id":"task_330_enterprise_industry_policy_analysis_medium_medium024","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":799526,"cost_usd":0.200303,"elapsed_time":365.27,"gpr":0.0,"tpe":0.0,"ee":0.1562},{"task_id":"task_331_enterprise_industry_policy_analysis_medium_medium025","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":609988,"cost_usd":0.15205,"elapsed_time":230.21,"gpr":1.0,"tpe":0.1439,"ee":0.1923},{"task_id":"task_332_enterprise_industry_policy_analysis_medium_medium026","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":938586,"cost_usd":0.23001,"elapsed_time":571.66,"gpr":1.0,"tpe":0.2311,"ee":0.1163},{"task_id":"task_333_enterprise_industry_policy_analysis_medium_medium027","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":481016,"cost_usd":0.119005,"elapsed_time":211.07,"gpr":0.4,"tpe":0.2688,"ee":0.2083},{"task_id":"task_334_enterprise_industry_policy_analysis_medium_medium028","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1680357,"cost_usd":0.398528,"elapsed_time":523.17,"gpr":0.4,"tpe":0.087,"ee":0.119},{"task_id":"task_335_enterprise_industry_policy_analysis_medium_medium029","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1067690,"cost_usd":0.258176,"elapsed_time":483.96,"gpr":1.0,"tpe":0.0872,"ee":0.119},{"task_id":"task_336_enterprise_industry_policy_analysis_medium_medium030","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":623308,"cost_usd":0.155083,"elapsed_time":438.7,"gpr":0.4,"tpe":0.4943,"ee":0.1351},{"task_id":"task_337_enterprise_industry_policy_analysis_medium_medium031","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1788987,"cost_usd":0.444876,"elapsed_time":537.04,"gpr":1.0,"tpe":0.1071,"ee":0.093},{"task_id":"task_338_enterprise_industry_policy_analysis_medium_medium032","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":772117,"cost_usd":0.185428,"elapsed_time":343.23,"gpr":0.5,"tpe":1.0,"ee":0.1081},{"task_id":"task_339_enterprise_industry_policy_analysis_medium_medium033","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":588464,"cost_usd":0.147349,"elapsed_time":371.64,"gpr":1.0,"tpe":0.1034,"ee":0.1212},{"task_id":"task_340_enterprise_industry_policy_analysis_medium_medium034","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":255850,"cost_usd":0.063807,"elapsed_time":189.64,"gpr":1.0,"tpe":0.5052,"ee":0.2353},{"task_id":"task_341_enterprise_industry_policy_analysis_medium_medium035","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":779175,"cost_usd":0.192932,"elapsed_time":182.16,"gpr":1.0,"tpe":0.4041,"ee":0.2857},{"task_id":"task_342_enterprise_industry_policy_analysis_medium_medium036","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":698561,"cost_usd":0.165475,"elapsed_time":292.36,"gpr":0.5,"tpe":0.3907,"ee":0.1176},{"task_id":"task_343_enterprise_industry_policy_analysis_medium_medium037","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1394092,"cost_usd":0.351706,"elapsed_time":557.34,"gpr":0.5,"tpe":0.4226,"ee":0.0976},{"task_id":"task_344_enterprise_industry_policy_analysis_medium_medium038","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":746834,"cost_usd":0.188667,"elapsed_time":189.79,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_345_enterprise_industry_policy_analysis_medium_medium039","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1244020,"cost_usd":0.311961,"elapsed_time":296.76,"gpr":0.75,"tpe":0.3196,"ee":0.1905},{"task_id":"task_346_enterprise_industry_policy_analysis_medium_medium040","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":686719,"cost_usd":0.172384,"elapsed_time":224.86,"gpr":1.0,"tpe":0.3568,"ee":0.2353},{"task_id":"task_347_enterprise_industry_policy_analysis_medium_medium041","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1319536,"cost_usd":0.330103,"elapsed_time":489.29,"gpr":0.5,"tpe":0.0474,"ee":0.0851},{"task_id":"task_348_enterprise_industry_policy_analysis_medium_medium042","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1259337,"cost_usd":0.309308,"elapsed_time":355.94,"gpr":0.5,"tpe":0.3936,"ee":0.0976},{"task_id":"task_349_enterprise_industry_policy_analysis_medium_medium043","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":739401,"cost_usd":0.187296,"elapsed_time":327.78,"gpr":0.5,"tpe":0.1959,"ee":0.129},{"task_id":"task_350_enterprise_industry_policy_analysis_medium_medium044","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":133634,"cost_usd":0.02991,"elapsed_time":74.7,"gpr":0.75,"tpe":0.729,"ee":0.3636},{"task_id":"task_351_enterprise_industry_policy_analysis_medium_medium045","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2265797,"cost_usd":0.570839,"elapsed_time":489.96,"gpr":0.5,"tpe":0.1223,"ee":0.1111},{"task_id":"task_352_enterprise_industry_policy_analysis_medium_medium046","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":590976,"cost_usd":0.148221,"elapsed_time":214.67,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_353_enterprise_industry_policy_analysis_medium_medium047","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":535939,"cost_usd":0.132462,"elapsed_time":325.3,"gpr":1.0,"tpe":0.1886,"ee":0.1333},{"task_id":"task_354_enterprise_industry_policy_analysis_medium_medium048","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1076471,"cost_usd":0.267087,"elapsed_time":531.6,"gpr":0.5,"tpe":0.4625,"ee":0.093},{"task_id":"task_355_enterprise_industry_policy_analysis_medium_medium049","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":396830,"cost_usd":0.09447,"elapsed_time":216.44,"gpr":0.75,"tpe":0.7155,"ee":0.16},{"task_id":"task_356_enterprise_industry_policy_analysis_medium_medium050","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":427363,"cost_usd":0.099032,"elapsed_time":223.95,"gpr":0.5,"tpe":0.7452,"ee":0.1429},{"task_id":"task_357_enterprise_industry_policy_analysis_medium_medium051","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1292901,"cost_usd":0.323724,"elapsed_time":460.04,"gpr":0.0,"tpe":0.0,"ee":0.1111},{"task_id":"task_358_enterprise_industry_policy_analysis_medium_medium052","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2324070,"cost_usd":0.568867,"elapsed_time":903.05,"gpr":0.75,"tpe":0.0034,"ee":0.0571},{"task_id":"task_359_enterprise_industry_policy_analysis_medium_medium053","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":667769,"cost_usd":0.167853,"elapsed_time":270.71,"gpr":0.75,"tpe":0.2214,"ee":0.1538},{"task_id":"task_360_enterprise_industry_policy_analysis_medium_medium054","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1871708,"cost_usd":0.469699,"elapsed_time":510.93,"gpr":0.75,"tpe":0.1297,"ee":0.1176},{"task_id":"task_361_enterprise_industry_policy_analysis_medium_medium055","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1537826,"cost_usd":0.386185,"elapsed_time":374.75,"gpr":1.0,"tpe":0.0674,"ee":0.1111},{"task_id":"task_362_enterprise_industry_policy_analysis_medium_medium056","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":226100,"cost_usd":0.057412,"elapsed_time":141.47,"gpr":0.5,"tpe":0.7657,"ee":0.25},{"task_id":"task_363_enterprise_industry_policy_analysis_medium_medium057","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":209564,"cost_usd":0.051711,"elapsed_time":109.64,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_364_enterprise_industry_policy_analysis_medium_medium058","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":786941,"cost_usd":0.194915,"elapsed_time":155.43,"gpr":0.25,"tpe":1.0,"ee":0.2353},{"task_id":"task_365_enterprise_industry_policy_analysis_medium_medium059","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":35606,"cost_usd":0.006689,"elapsed_time":37.2,"gpr":0.0,"tpe":0.0,"ee":1.3333},{"task_id":"task_366_enterprise_industry_policy_analysis_medium_medium060","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":436474,"cost_usd":0.105687,"elapsed_time":309.72,"gpr":0.25,"tpe":1.0,"ee":0.1667},{"task_id":"task_367_enterprise_industry_policy_analysis_medium_medium061","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":485600,"cost_usd":0.121157,"elapsed_time":110.34,"gpr":0.75,"tpe":0.7805,"ee":0.3636},{"task_id":"task_368_enterprise_industry_policy_analysis_medium_medium062","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":526471,"cost_usd":0.116618,"elapsed_time":339.47,"gpr":0.0,"tpe":0.0,"ee":0.1212},{"task_id":"task_369_enterprise_industry_policy_analysis_medium_medium063","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":106860,"cost_usd":0.024747,"elapsed_time":68.34,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_370_enterprise_industry_policy_analysis_medium_medium064","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":893143,"cost_usd":0.222448,"elapsed_time":331.73,"gpr":0.5,"tpe":1.0,"ee":0.1176},{"task_id":"task_371_enterprise_industry_policy_analysis_medium_medium065","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1617176,"cost_usd":0.39402,"elapsed_time":538.13,"gpr":0.0,"tpe":0.0,"ee":0.0606},{"task_id":"task_372_enterprise_industry_policy_analysis_medium_medium066","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2085382,"cost_usd":0.494971,"elapsed_time":418.39,"gpr":0.0,"tpe":0.0,"ee":0.1026},{"task_id":"task_373_hypothesis_verification_hard_hard001","category":"hypothesis_verification","score":0.0,"tokens":1670001,"cost_usd":0.417644,"elapsed_time":733.52,"gpr":0.2222,"tpe":0.0343,"ee":0.1373},{"task_id":"task_374_hypothesis_verification_hard_hard002","category":"hypothesis_verification","score":0.0,"tokens":1145946,"cost_usd":0.290502,"elapsed_time":530.06,"gpr":0.375,"tpe":0.1351,"ee":0.1613},{"task_id":"task_375_hypothesis_verification_hard_hard003","category":"hypothesis_verification","score":0.0,"tokens":3525934,"cost_usd":0.888785,"elapsed_time":729.03,"gpr":0.2222,"tpe":0.1543,"ee":0.1951},{"task_id":"task_376_hypothesis_verification_hard_hard004","category":"hypothesis_verification","score":0.0,"tokens":3663885,"cost_usd":0.925299,"elapsed_time":1103.74,"gpr":0.2,"tpe":1.0,"ee":0.125},{"task_id":"task_377_hypothesis_verification_hard_hard005","category":"hypothesis_verification","score":0.0,"tokens":2852277,"cost_usd":0.70323,"elapsed_time":625.56,"gpr":0.5385,"tpe":0.8986,"ee":0.1343},{"task_id":"task_378_hypothesis_verification_hard_hard006","category":"hypothesis_verification","score":0.0,"tokens":1709953,"cost_usd":0.404519,"elapsed_time":670.96,"gpr":0.3636,"tpe":0.2123,"ee":0.14},{"task_id":"task_379_hypothesis_verification_hard_hard007","category":"hypothesis_verification","score":0.0,"tokens":3869350,"cost_usd":0.975415,"elapsed_time":782.65,"gpr":0.1875,"tpe":0.083,"ee":0.2045},{"task_id":"task_380_hypothesis_verification_hard_hard008","category":"hypothesis_verification","score":0.0,"tokens":1939729,"cost_usd":0.478088,"elapsed_time":711.43,"gpr":0.2222,"tpe":1.0,"ee":0.1667},{"task_id":"task_381_hypothesis_verification_hard_hard009","category":"hypothesis_verification","score":0.0,"tokens":2589006,"cost_usd":0.645435,"elapsed_time":928.98,"gpr":0.1111,"tpe":0.1501,"ee":0.1},{"task_id":"task_382_hypothesis_verification_hard_hard010","category":"hypothesis_verification","score":0.0,"tokens":7241108,"cost_usd":1.772756,"elapsed_time":1207.28,"gpr":0.5333,"tpe":0.975,"ee":0.1094},{"task_id":"task_383_hypothesis_verification_hard_hard011","category":"hypothesis_verification","score":0.0,"tokens":1223761,"cost_usd":0.308748,"elapsed_time":227.41,"gpr":0.0,"tpe":0.0,"ee":0.4375},{"task_id":"task_384_hypothesis_verification_hard_hard012","category":"hypothesis_verification","score":0.0,"tokens":3157014,"cost_usd":0.732567,"elapsed_time":1021.48,"gpr":0.4545,"tpe":0.3839,"ee":0.0952},{"task_id":"task_385_hypothesis_verification_hard_hard013","category":"hypothesis_verification","score":0.0,"tokens":1464114,"cost_usd":0.352239,"elapsed_time":369.17,"gpr":0.0,"tpe":0.0,"ee":0.175},{"task_id":"task_386_hypothesis_verification_hard_hard014","category":"hypothesis_verification","score":0.0,"tokens":2163970,"cost_usd":0.532826,"elapsed_time":891.59,"gpr":0.0,"tpe":0.0,"ee":0.1034},{"task_id":"task_387_hypothesis_verification_hard_hard015","category":"hypothesis_verification","score":0.0,"tokens":1802863,"cost_usd":0.456434,"elapsed_time":669.26,"gpr":0.0909,"tpe":0.3874,"ee":0.175},{"task_id":"task_388_hypothesis_verification_medium_medium001","category":"hypothesis_verification","score":0.0,"tokens":1709792,"cost_usd":0.404236,"elapsed_time":691.62,"gpr":0.0,"tpe":0.0,"ee":0.14},{"task_id":"task_389_hypothesis_verification_medium_medium002","category":"hypothesis_verification","score":0.0,"tokens":4383373,"cost_usd":1.105104,"elapsed_time":936.11,"gpr":0.0,"tpe":0.0,"ee":0.1207},{"task_id":"task_390_hypothesis_verification_medium_medium003","category":"hypothesis_verification","score":0.0,"tokens":1478159,"cost_usd":0.36743,"elapsed_time":768.2,"gpr":0.3333,"tpe":0.1256,"ee":0.0952},{"task_id":"task_391_hypothesis_verification_medium_medium004","category":"hypothesis_verification","score":0.0,"tokens":4198995,"cost_usd":1.059386,"elapsed_time":1002.19,"gpr":0.2,"tpe":0.81,"ee":0.1061},{"task_id":"task_392_hypothesis_verification_medium_medium005","category":"hypothesis_verification","score":1.0,"tokens":1438144,"cost_usd":0.338293,"elapsed_time":677.49,"gpr":1.0,"tpe":0.0228,"ee":0.1167},{"task_id":"task_393_hypothesis_verification_medium_medium006","category":"hypothesis_verification","score":0.0,"tokens":5025557,"cost_usd":1.265247,"elapsed_time":997.78,"gpr":0.0,"tpe":0.0,"ee":0.0968},{"task_id":"task_394_hypothesis_verification_medium_medium007","category":"hypothesis_verification","score":0.0,"tokens":1128892,"cost_usd":0.282277,"elapsed_time":548.16,"gpr":0.0,"tpe":0.0,"ee":0.1489},{"task_id":"task_395_hypothesis_verification_medium_medium008","category":"hypothesis_verification","score":0.0,"tokens":3105170,"cost_usd":0.784952,"elapsed_time":823.56,"gpr":0.2,"tpe":0.2288,"ee":0.1905},{"task_id":"task_396_hypothesis_verification_medium_medium009","category":"hypothesis_verification","score":1.0,"tokens":1532920,"cost_usd":0.370195,"elapsed_time":615.79,"gpr":1.0,"tpe":0.0193,"ee":0.1364},{"task_id":"task_397_hypothesis_verification_medium_medium010","category":"hypothesis_verification","score":0.0,"tokens":1560766,"cost_usd":0.390992,"elapsed_time":757.16,"gpr":0.2857,"tpe":0.2893,"ee":0.14},{"task_id":"task_398_hypothesis_verification_medium_medium011","category":"hypothesis_verification","score":1.0,"tokens":1551815,"cost_usd":0.380308,"elapsed_time":683.21,"gpr":1.0,"tpe":0.0071,"ee":0.1132},{"task_id":"task_399_hypothesis_verification_medium_medium012","category":"hypothesis_verification","score":1.0,"tokens":1867698,"cost_usd":0.460193,"elapsed_time":1005.15,"gpr":1.0,"tpe":0.0052,"ee":0.1},{"task_id":"task_400_hypothesis_verification_medium_medium013","category":"hypothesis_verification","score":0.0,"tokens":2600754,"cost_usd":0.656321,"elapsed_time":896.91,"gpr":0.0,"tpe":0.0,"ee":0.1228},{"task_id":"task_401_hypothesis_verification_medium_medium014","category":"hypothesis_verification","score":0.0,"tokens":1172704,"cost_usd":0.273727,"elapsed_time":564.39,"gpr":0.6,"tpe":0.1439,"ee":0.1951},{"task_id":"task_402_industry_planning_hard_hard001","category":"industry_planning","score":0.0,"tokens":5844762,"cost_usd":1.468534,"elapsed_time":1207.13,"gpr":0.3333,"tpe":0.0237,"ee":0.0886},{"task_id":"task_403_industry_planning_hard_hard002","category":"industry_planning","score":0.0,"tokens":2610737,"cost_usd":0.621541,"elapsed_time":913.88,"gpr":0.25,"tpe":0.1849,"ee":0.125},{"task_id":"task_404_industry_planning_hard_hard003","category":"industry_planning","score":0.0,"tokens":1555316,"cost_usd":0.385808,"elapsed_time":858.98,"gpr":0.2727,"tpe":0.2862,"ee":0.1538},{"task_id":"task_405_industry_planning_hard_hard004","category":"industry_planning","score":0.0,"tokens":4575984,"cost_usd":1.098822,"elapsed_time":962.86,"gpr":0.5,"tpe":0.0548,"ee":0.1034},{"task_id":"task_406_industry_planning_hard_hard005","category":"industry_planning","score":0.0,"tokens":2858944,"cost_usd":0.697814,"elapsed_time":668.51,"gpr":0.2222,"tpe":0.828,"ee":0.1132},{"task_id":"task_407_industry_planning_hard_hard006","category":"industry_planning","score":0.0,"tokens":890132,"cost_usd":0.216672,"elapsed_time":399.05,"gpr":0.4167,"tpe":0.3531,"ee":0.175},{"task_id":"task_408_industry_planning_hard_hard007","category":"industry_planning","score":0.0,"tokens":1740202,"cost_usd":0.40786,"elapsed_time":292.92,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_409_industry_planning_hard_hard008","category":"industry_planning","score":0.0,"tokens":1028436,"cost_usd":0.250016,"elapsed_time":438.92,"gpr":0.375,"tpe":0.7707,"ee":0.1538},{"task_id":"task_410_industry_planning_hard_hard009","category":"industry_planning","score":0.0,"tokens":1856899,"cost_usd":0.46768,"elapsed_time":849.4,"gpr":0.6,"tpe":0.1657,"ee":0.1304},{"task_id":"task_411_industry_planning_hard_hard010","category":"industry_planning","score":0.0,"tokens":2175272,"cost_usd":0.537651,"elapsed_time":654.96,"gpr":0.4444,"tpe":0.2765,"ee":0.0984},{"task_id":"task_412_industry_planning_hard_hard011","category":"industry_planning","score":0.0,"tokens":546267,"cost_usd":0.137875,"elapsed_time":130.76,"gpr":0.0,"tpe":0.0,"ee":0.75},{"task_id":"task_413_industry_planning_hard_hard012","category":"industry_planning","score":0.0,"tokens":1813357,"cost_usd":0.454802,"elapsed_time":823.69,"gpr":0.7778,"tpe":0.0202,"ee":0.1091},{"task_id":"task_414_industry_planning_hard_hard013","category":"industry_planning","score":0.0,"tokens":332299,"cost_usd":0.083871,"elapsed_time":104.57,"gpr":0.0,"tpe":0.0,"ee":0.875},{"task_id":"task_415_industry_planning_hard_hard014","category":"industry_planning","score":0.0,"tokens":5027439,"cost_usd":1.229715,"elapsed_time":1081.97,"gpr":0.0,"tpe":0.0,"ee":0.0833},{"task_id":"task_416_industry_planning_medium_medium001","category":"industry_planning","score":0.0,"tokens":3200570,"cost_usd":0.795728,"elapsed_time":950.92,"gpr":0.0,"tpe":0.0,"ee":0.1905},{"task_id":"task_417_industry_planning_medium_medium002","category":"industry_planning","score":1.0,"tokens":2693825,"cost_usd":0.682583,"elapsed_time":1129.85,"gpr":1.0,"tpe":0.0194,"ee":0.1127},{"task_id":"task_418_industry_planning_medium_medium003","category":"industry_planning","score":0.0,"tokens":1923231,"cost_usd":0.466411,"elapsed_time":637.72,"gpr":0.0,"tpe":0.0,"ee":0.1905},{"task_id":"task_419_industry_planning_medium_medium004","category":"industry_planning","score":1.0,"tokens":1331446,"cost_usd":0.324569,"elapsed_time":675.03,"gpr":1.0,"tpe":0.1014,"ee":0.1556},{"task_id":"task_420_industry_planning_medium_medium005","category":"industry_planning","score":0.0,"tokens":3747826,"cost_usd":0.930739,"elapsed_time":965.34,"gpr":0.3333,"tpe":0.0428,"ee":0.1296},{"task_id":"task_421_industry_planning_medium_medium006","category":"industry_planning","score":0.0,"tokens":949911,"cost_usd":0.234987,"elapsed_time":611.91,"gpr":0.1111,"tpe":0.0798,"ee":0.2162},{"task_id":"task_422_industry_planning_medium_medium007","category":"industry_planning","score":0.0,"tokens":1630162,"cost_usd":0.398236,"elapsed_time":832.79,"gpr":0.0,"tpe":0.0,"ee":0.1231},{"task_id":"task_423_industry_planning_medium_medium008","category":"industry_planning","score":0.0,"tokens":5500613,"cost_usd":1.326762,"elapsed_time":1126.96,"gpr":0.0,"tpe":0.0,"ee":0.0964},{"task_id":"task_424_industry_planning_medium_medium009","category":"industry_planning","score":0.0,"tokens":3298505,"cost_usd":0.834291,"elapsed_time":1005.04,"gpr":0.1,"tpe":0.0343,"ee":0.1379},{"task_id":"task_425_industry_planning_medium_medium010","category":"industry_planning","score":0.0,"tokens":2122072,"cost_usd":0.529385,"elapsed_time":920.52,"gpr":0.4,"tpe":0.0257,"ee":0.1538},{"task_id":"task_426_industry_planning_medium_medium011","category":"industry_planning","score":1.0,"tokens":1532323,"cost_usd":0.363998,"elapsed_time":803.46,"gpr":1.0,"tpe":0.1194,"ee":0.2105},{"task_id":"task_427_industry_planning_medium_medium012","category":"industry_planning","score":0.0,"tokens":992035,"cost_usd":0.252645,"elapsed_time":730.34,"gpr":0.2222,"tpe":0.0838,"ee":0.2162},{"task_id":"task_428_industry_planning_medium_medium013","category":"industry_planning","score":0.0,"tokens":983755,"cost_usd":0.225397,"elapsed_time":577.55,"gpr":0.1667,"tpe":0.0582,"ee":0.2105},{"task_id":"task_429_industry_planning_medium_medium014","category":"industry_planning","score":0.0,"tokens":1368424,"cost_usd":0.345136,"elapsed_time":596.84,"gpr":0.3333,"tpe":0.0343,"ee":0.1698},{"task_id":"task_430_international_comparison_hard_hard001","category":"international_comparison","score":0.0,"tokens":2094594,"cost_usd":0.528663,"elapsed_time":395.35,"gpr":0.2,"tpe":0.0534,"ee":0.1277},{"task_id":"task_431_international_comparison_hard_hard002","category":"international_comparison","score":0.0,"tokens":1390601,"cost_usd":0.350759,"elapsed_time":251.08,"gpr":0.0,"tpe":0.0,"ee":0.2222},{"task_id":"task_432_international_comparison_hard_hard003","category":"international_comparison","score":0.0,"tokens":929167,"cost_usd":0.201329,"elapsed_time":235.36,"gpr":0.2222,"tpe":0.4783,"ee":0.2609},{"task_id":"task_433_international_comparison_hard_hard004","category":"international_comparison","score":0.5,"tokens":864419,"cost_usd":0.208624,"elapsed_time":242.93,"gpr":0.0,"tpe":0.0,"ee":0.2609},{"task_id":"task_434_international_comparison_hard_hard005","category":"international_comparison","score":0.0,"tokens":925240,"cost_usd":0.201289,"elapsed_time":305.04,"gpr":0.0,"tpe":0.0,"ee":0.1935},{"task_id":"task_435_international_comparison_hard_hard006","category":"international_comparison","score":0.0,"tokens":96784,"cost_usd":0.024476,"elapsed_time":52.04,"gpr":0.0,"tpe":0.0,"ee":1.75},{"task_id":"task_436_international_comparison_hard_hard007","category":"international_comparison","score":0.0,"tokens":879362,"cost_usd":0.221951,"elapsed_time":211.61,"gpr":0.0909,"tpe":0.4305,"ee":0.35},{"task_id":"task_437_international_comparison_hard_hard008","category":"international_comparison","score":0.5,"tokens":789975,"cost_usd":0.199534,"elapsed_time":264.65,"gpr":0.1667,"tpe":0.0798,"ee":0.1935},{"task_id":"task_438_international_comparison_hard_hard009","category":"international_comparison","score":0.0,"tokens":358741,"cost_usd":0.082616,"elapsed_time":255.94,"gpr":0.0,"tpe":0.0,"ee":0.4375},{"task_id":"task_439_international_comparison_hard_hard010","category":"international_comparison","score":0.0,"tokens":558450,"cost_usd":0.141162,"elapsed_time":183.82,"gpr":0.3333,"tpe":0.3721,"ee":0.2308},{"task_id":"task_440_international_comparison_hard_hard011","category":"international_comparison","score":1.0,"tokens":785907,"cost_usd":0.198465,"elapsed_time":200.4,"gpr":1.0,"tpe":0.2264,"ee":0.24},{"task_id":"task_441_international_comparison_hard_hard012","category":"international_comparison","score":0.0,"tokens":1023916,"cost_usd":0.258351,"elapsed_time":203.59,"gpr":0.0,"tpe":0.0,"ee":0.2857},{"task_id":"task_442_international_comparison_hard_hard013","category":"international_comparison","score":0.0,"tokens":867112,"cost_usd":0.218903,"elapsed_time":205.12,"gpr":0.2,"tpe":1.0,"ee":0.2069},{"task_id":"task_443_international_comparison_hard_hard014","category":"international_comparison","score":0.0,"tokens":257530,"cost_usd":0.063463,"elapsed_time":99.77,"gpr":0.0,"tpe":0.0,"ee":0.4615},{"task_id":"task_444_international_comparison_medium_medium001","category":"international_comparison","score":0.5,"tokens":953312,"cost_usd":0.223802,"elapsed_time":299.2,"gpr":0.2,"tpe":0.4783,"ee":0.1212},{"task_id":"task_445_international_comparison_medium_medium002","category":"international_comparison","score":0.5,"tokens":1444208,"cost_usd":0.361181,"elapsed_time":288.13,"gpr":0.6667,"tpe":0.0985,"ee":0.1212},{"task_id":"task_446_international_comparison_medium_medium003","category":"international_comparison","score":0.5,"tokens":804529,"cost_usd":0.172501,"elapsed_time":207.12,"gpr":0.3333,"tpe":0.2713,"ee":0.1429},{"task_id":"task_447_international_comparison_medium_medium004","category":"international_comparison","score":0.0,"tokens":619715,"cost_usd":0.142395,"elapsed_time":192.61,"gpr":0.6667,"tpe":0.9525,"ee":0.2222},{"task_id":"task_448_international_comparison_medium_medium005","category":"international_comparison","score":0.0,"tokens":543497,"cost_usd":0.13098,"elapsed_time":163.21,"gpr":0.1667,"tpe":0.3874,"ee":0.2105},{"task_id":"task_449_international_comparison_medium_medium006","category":"international_comparison","score":0.33,"tokens":1310260,"cost_usd":0.29655,"elapsed_time":430.08,"gpr":0.0,"tpe":0.0,"ee":0.1064},{"task_id":"task_450_international_comparison_medium_medium007","category":"international_comparison","score":0.33,"tokens":701002,"cost_usd":0.147825,"elapsed_time":329.74,"gpr":0.5,"tpe":0.6906,"ee":0.1333},{"task_id":"task_451_international_comparison_medium_medium008","category":"international_comparison","score":0.333,"tokens":731225,"cost_usd":0.161454,"elapsed_time":220.8,"gpr":0.6667,"tpe":0.4711,"ee":0.1667},{"task_id":"task_452_international_comparison_medium_medium009","category":"international_comparison","score":0.0,"tokens":949409,"cost_usd":0.226118,"elapsed_time":226.76,"gpr":0.2857,"tpe":0.729,"ee":0.1786},{"task_id":"task_453_international_comparison_medium_medium010","category":"international_comparison","score":0.5,"tokens":715511,"cost_usd":0.162209,"elapsed_time":436.98,"gpr":0.6,"tpe":0.5472,"ee":0.2778},{"task_id":"task_454_international_comparison_medium_medium011","category":"international_comparison","score":0.333,"tokens":511558,"cost_usd":0.119561,"elapsed_time":182.71,"gpr":0.3333,"tpe":0.3874,"ee":0.2105},{"task_id":"task_455_international_comparison_medium_medium012","category":"international_comparison","score":0.0,"tokens":44313,"cost_usd":0.011193,"elapsed_time":26.51,"gpr":0.0,"tpe":0.0,"ee":1.6667},{"task_id":"task_456_international_comparison_medium_medium013","category":"international_comparison","score":0.0,"tokens":423934,"cost_usd":0.087122,"elapsed_time":186.7,"gpr":0.0,"tpe":0.0,"ee":0.2222},{"task_id":"task_457_international_comparison_medium_medium014","category":"international_comparison","score":0.0,"tokens":1994382,"cost_usd":0.503566,"elapsed_time":412.99,"gpr":0.125,"tpe":0.0133,"ee":0.1064},{"task_id":"task_458_international_comparison_medium_medium015","category":"international_comparison","score":0.0,"tokens":1579130,"cost_usd":0.399432,"elapsed_time":508.13,"gpr":0.0,"tpe":0.0,"ee":0.1163},{"task_id":"task_459_international_comparison_medium_medium016","category":"international_comparison","score":0.5,"tokens":965569,"cost_usd":0.242244,"elapsed_time":245.68,"gpr":0.3333,"tpe":0.229,"ee":0.1852},{"task_id":"task_460_international_comparison_medium_medium017","category":"international_comparison","score":0.25,"tokens":973595,"cost_usd":0.245726,"elapsed_time":185.99,"gpr":0.0,"tpe":0.0,"ee":0.1724},{"task_id":"task_461_international_comparison_medium_medium018","category":"international_comparison","score":1.0,"tokens":184076,"cost_usd":0.037956,"elapsed_time":111.61,"gpr":1.0,"tpe":0.5905,"ee":0.4444},{"task_id":"task_462_international_comparison_medium_medium019","category":"international_comparison","score":0.5,"tokens":355239,"cost_usd":0.081118,"elapsed_time":148.22,"gpr":0.5,"tpe":0.3034,"ee":0.25},{"task_id":"task_463_international_comparison_medium_medium020","category":"international_comparison","score":0.0,"tokens":119133,"cost_usd":0.02987,"elapsed_time":373.11,"gpr":0.0,"tpe":0.0,"ee":0.7143},{"task_id":"task_464_international_comparison_medium_medium021","category":"international_comparison","score":0.0,"tokens":483832,"cost_usd":0.114286,"elapsed_time":147.34,"gpr":0.1429,"tpe":1.0,"ee":0.25},{"task_id":"task_465_international_comparison_medium_medium022","category":"international_comparison","score":0.0,"tokens":379772,"cost_usd":0.094198,"elapsed_time":85.38,"gpr":0.0909,"tpe":1.0,"ee":0.7143},{"task_id":"task_466_international_comparison_medium_medium023","category":"international_comparison","score":0.0,"tokens":741173,"cost_usd":0.185367,"elapsed_time":147.3,"gpr":0.5,"tpe":0.3956,"ee":0.2778},{"task_id":"task_467_international_comparison_medium_medium024","category":"international_comparison","score":0.0,"tokens":978457,"cost_usd":0.24705,"elapsed_time":230.07,"gpr":0.0833,"tpe":0.2288,"ee":0.1471},{"task_id":"task_468_international_comparison_medium_medium025","category":"international_comparison","score":0.0,"tokens":188654,"cost_usd":0.044755,"elapsed_time":68.73,"gpr":0.0,"tpe":0.0,"ee":0.4444},{"task_id":"task_469_risk_assessment_hard_hard001","category":"risk_assessment","score":0.0,"tokens":2836478,"cost_usd":0.685865,"elapsed_time":744.57,"gpr":0.4,"tpe":0.125,"ee":0.1667},{"task_id":"task_470_risk_assessment_hard_hard002","category":"risk_assessment","score":1.0,"tokens":810825,"cost_usd":0.206346,"elapsed_time":504.76,"gpr":1.0,"tpe":0.1645,"ee":0.1842},{"task_id":"task_471_risk_assessment_hard_hard003","category":"risk_assessment","score":0.0,"tokens":1734806,"cost_usd":0.43717,"elapsed_time":590.8,"gpr":0.1818,"tpe":0.2484,"ee":0.1707},{"task_id":"task_472_risk_assessment_hard_hard004","category":"risk_assessment","score":1.0,"tokens":1687894,"cost_usd":0.427044,"elapsed_time":541.12,"gpr":1.0,"tpe":0.4532,"ee":0.2333},{"task_id":"task_473_risk_assessment_hard_hard005","category":"risk_assessment","score":0.0,"tokens":2993058,"cost_usd":0.756421,"elapsed_time":1020.3,"gpr":0.0,"tpe":0.0,"ee":0.1186},{"task_id":"task_474_risk_assessment_hard_hard006","category":"risk_assessment","score":1.0,"tokens":1927518,"cost_usd":0.462917,"elapsed_time":570.22,"gpr":1.0,"tpe":0.1486,"ee":0.1842},{"task_id":"task_475_risk_assessment_hard_hard007","category":"risk_assessment","score":0.0,"tokens":331836,"cost_usd":0.083842,"elapsed_time":109.74,"gpr":0.3333,"tpe":1.0,"ee":0.6364},{"task_id":"task_476_risk_assessment_hard_hard008","category":"risk_assessment","score":0.0,"tokens":3033528,"cost_usd":0.754615,"elapsed_time":859.37,"gpr":0.0,"tpe":0.0,"ee":0.1628},{"task_id":"task_477_risk_assessment_hard_hard009","category":"risk_assessment","score":0.0,"tokens":1448458,"cost_usd":0.364243,"elapsed_time":1089.21,"gpr":0.0,"tpe":0.0,"ee":0.1373},{"task_id":"task_478_risk_assessment_hard_hard010","category":"risk_assessment","score":0.0,"tokens":1300607,"cost_usd":0.317251,"elapsed_time":691.98,"gpr":0.0,"tpe":0.0,"ee":0.1667},{"task_id":"task_479_risk_assessment_hard_hard011","category":"risk_assessment","score":0.0,"tokens":2505913,"cost_usd":0.633694,"elapsed_time":654.92,"gpr":0.5,"tpe":0.7913,"ee":0.1842},{"task_id":"task_480_risk_assessment_hard_hard012","category":"risk_assessment","score":1.0,"tokens":1046644,"cost_usd":0.26602,"elapsed_time":607.26,"gpr":1.0,"tpe":0.0672,"ee":0.1591},{"task_id":"task_481_risk_assessment_hard_hard013","category":"risk_assessment","score":0.0,"tokens":930316,"cost_usd":0.220627,"elapsed_time":524.11,"gpr":0.3333,"tpe":0.3138,"ee":0.2045},{"task_id":"task_482_risk_assessment_medium_medium001","category":"risk_assessment","score":1.0,"tokens":2699329,"cost_usd":0.683012,"elapsed_time":815.71,"gpr":1.0,"tpe":0.2004,"ee":0.1429},{"task_id":"task_483_risk_assessment_medium_medium002","category":"risk_assessment","score":0.0,"tokens":3288480,"cost_usd":0.806017,"elapsed_time":748.86,"gpr":0.25,"tpe":0.0182,"ee":0.1132},{"task_id":"task_484_risk_assessment_medium_medium003","category":"risk_assessment","score":0.0,"tokens":2508849,"cost_usd":0.634611,"elapsed_time":712.6,"gpr":0.0,"tpe":0.0,"ee":0.2069},{"task_id":"task_485_risk_assessment_medium_medium004","category":"risk_assessment","score":0.0,"tokens":1462788,"cost_usd":0.360902,"elapsed_time":723.34,"gpr":0.0,"tpe":0.0,"ee":0.1316},{"task_id":"task_486_risk_assessment_medium_medium005","category":"risk_assessment","score":0.5,"tokens":2919431,"cost_usd":0.737046,"elapsed_time":1130.61,"gpr":0.5,"tpe":0.1898,"ee":0.0714},{"task_id":"task_487_risk_assessment_medium_medium006","category":"risk_assessment","score":1.0,"tokens":2559366,"cost_usd":0.595344,"elapsed_time":881.09,"gpr":1.0,"tpe":0.229,"ee":0.0735},{"task_id":"task_488_risk_assessment_medium_medium007","category":"risk_assessment","score":0.0,"tokens":3077658,"cost_usd":0.764032,"elapsed_time":1207.2,"gpr":0.25,"tpe":0.6009,"ee":0.0759},{"task_id":"task_489_risk_assessment_medium_medium008","category":"risk_assessment","score":0.0,"tokens":3275994,"cost_usd":0.808256,"elapsed_time":879.89,"gpr":0.25,"tpe":0.2758,"ee":0.1333},{"task_id":"task_490_risk_assessment_medium_medium009","category":"risk_assessment","score":0.0,"tokens":2146293,"cost_usd":0.51724,"elapsed_time":1135.07,"gpr":0.4167,"tpe":0.8024,"ee":0.0822},{"task_id":"task_491_risk_assessment_medium_medium010","category":"risk_assessment","score":0.5,"tokens":3786688,"cost_usd":0.940618,"elapsed_time":950.11,"gpr":0.2,"tpe":0.0646,"ee":0.1071},{"task_id":"task_492_risk_assessment_medium_medium011","category":"risk_assessment","score":1.0,"tokens":681527,"cost_usd":0.173189,"elapsed_time":458.82,"gpr":1.0,"tpe":0.0961,"ee":0.1818}]},{"model":"glm-5","acc":0.333,"ee":0.3212,"gpr":0.1951,"tpe":0.2473,"efficiency":{"total_tokens":408096337,"total_cost_usd":95.37,"total_requests":9046},"category_scores":{"comprehensive_decision":{"acc":0.2286,"ee":0.2795,"gpr":0.1357,"tpe":0.2241,"count":70},"enterprise_industry_analysis":{"acc":0.4425,"ee":0.3157,"gpr":0.1738,"tpe":0.1421,"count":226},"enterprise_industry_policy_analysis":{"acc":0.2895,"ee":0.4025,"gpr":0.2981,"tpe":0.4088,"count":76},"hypothesis_verification":{"acc":0.1034,"ee":0.3208,"gpr":0.1236,"tpe":0.2549,"count":29},"industry_planning":{"acc":0.2679,"ee":0.2422,"gpr":0.1104,"tpe":0.2405,"count":28},"international_comparison":{"acc":0.3289,"ee":0.3179,"gpr":0.3455,"tpe":0.3944,"count":39},"risk_assessment":{"acc":0.1042,"ee":0.3216,"gpr":0.1431,"tpe":0.2804,"count":24}},"tasks":[{"task_id":"task_001_comprehensive_decision_easy_easy001","category":"comprehensive_decision","score":0.0,"tokens":737976,"cost_usd":0.143589,"elapsed_time":165.91,"gpr":0.0,"tpe":0.0,"ee":0.15},{"task_id":"task_002_comprehensive_decision_easy_easy002","category":"comprehensive_decision","score":1.0,"tokens":142251,"cost_usd":0.026787,"elapsed_time":130.9,"gpr":1.0,"tpe":0.3709,"ee":0.1538},{"task_id":"task_003_comprehensive_decision_easy_easy003","category":"comprehensive_decision","score":0.0,"tokens":1863005,"cost_usd":0.328996,"elapsed_time":238.73,"gpr":0.0,"tpe":0.0,"ee":0.0909},{"task_id":"task_004_comprehensive_decision_easy_easy004","category":"comprehensive_decision","score":1.0,"tokens":173172,"cost_usd":0.041732,"elapsed_time":150.69,"gpr":1.0,"tpe":0.5153,"ee":0.1667},{"task_id":"task_005_comprehensive_decision_easy_easy005","category":"comprehensive_decision","score":0.0,"tokens":232687,"cost_usd":0.049656,"elapsed_time":154.51,"gpr":0.5,"tpe":0.6746,"ee":0.2},{"task_id":"task_006_comprehensive_decision_easy_easy006","category":"comprehensive_decision","score":0.0,"tokens":789517,"cost_usd":0.233079,"elapsed_time":134.39,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_007_comprehensive_decision_hard_hard001","category":"comprehensive_decision","score":0.0,"tokens":34134,"cost_usd":0.007035,"elapsed_time":56.72,"gpr":0.125,"tpe":1.0,"ee":1.4},{"task_id":"task_008_comprehensive_decision_hard_hard002","category":"comprehensive_decision","score":0.0,"tokens":666417,"cost_usd":0.176532,"elapsed_time":406.57,"gpr":0.0,"tpe":0.0,"ee":0.3684},{"task_id":"task_009_comprehensive_decision_hard_hard003","category":"comprehensive_decision","score":0.0,"tokens":195379,"cost_usd":0.056178,"elapsed_time":117.63,"gpr":0.0,"tpe":0.0,"ee":0.6364},{"task_id":"task_010_comprehensive_decision_hard_hard004","category":"comprehensive_decision","score":0.0,"tokens":183951,"cost_usd":0.047629,"elapsed_time":120.67,"gpr":0.5,"tpe":0.975,"ee":0.7778},{"task_id":"task_011_comprehensive_decision_hard_hard005","category":"comprehensive_decision","score":0.0,"tokens":922329,"cost_usd":0.189064,"elapsed_time":305.58,"gpr":0.4286,"tpe":0.2824,"ee":0.3182},{"task_id":"task_012_comprehensive_decision_hard_hard006","category":"comprehensive_decision","score":0.0,"tokens":3354622,"cost_usd":0.59299,"elapsed_time":931.19,"gpr":0.3333,"tpe":0.5314,"ee":0.1346},{"task_id":"task_013_comprehensive_decision_hard_hard007","category":"comprehensive_decision","score":0.0,"tokens":322064,"cost_usd":0.083018,"elapsed_time":130.09,"gpr":0.0,"tpe":0.0,"ee":0.7},{"task_id":"task_014_comprehensive_decision_hard_hard008","category":"comprehensive_decision","score":1.0,"tokens":1262790,"cost_usd":0.274777,"elapsed_time":309.62,"gpr":1.0,"tpe":0.3144,"ee":0.3043},{"task_id":"task_015_comprehensive_decision_hard_hard009","category":"comprehensive_decision","score":0.0,"tokens":1493961,"cost_usd":0.294183,"elapsed_time":517.31,"gpr":0.2857,"tpe":0.3679,"ee":0.2286},{"task_id":"task_016_comprehensive_decision_hard_hard010","category":"comprehensive_decision","score":0.0,"tokens":33980,"cost_usd":0.007147,"elapsed_time":54.9,"gpr":0.0,"tpe":0.0,"ee":1.4},{"task_id":"task_017_comprehensive_decision_hard_hard011","category":"comprehensive_decision","score":0.0,"tokens":2199481,"cost_usd":0.442447,"elapsed_time":269.59,"gpr":0.1111,"tpe":1.0,"ee":0.35},{"task_id":"task_018_comprehensive_decision_hard_hard012","category":"comprehensive_decision","score":0.0,"tokens":922526,"cost_usd":0.186771,"elapsed_time":375.53,"gpr":0.1667,"tpe":0.2824,"ee":0.2917},{"task_id":"task_019_comprehensive_decision_hard_hard013","category":"comprehensive_decision","score":0.0,"tokens":1710034,"cost_usd":0.458576,"elapsed_time":221.15,"gpr":0.0,"tpe":0.0,"ee":0.3684},{"task_id":"task_020_comprehensive_decision_hard_hard014","category":"comprehensive_decision","score":0.0,"tokens":237929,"cost_usd":0.050702,"elapsed_time":305.64,"gpr":0.5714,"tpe":0.6236,"ee":0.5},{"task_id":"task_021_comprehensive_decision_hard_hard015","category":"comprehensive_decision","score":0.0,"tokens":580069,"cost_usd":0.137151,"elapsed_time":665.11,"gpr":0.1429,"tpe":0.1668,"ee":0.2667},{"task_id":"task_022_comprehensive_decision_hard_hard016","category":"comprehensive_decision","score":0.0,"tokens":882737,"cost_usd":0.207292,"elapsed_time":541.56,"gpr":0.875,"tpe":0.253,"ee":0.2143},{"task_id":"task_023_comprehensive_decision_hard_hard017","category":"comprehensive_decision","score":0.0,"tokens":57024,"cost_usd":0.010399,"elapsed_time":182.53,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_024_comprehensive_decision_hard_hard018","category":"comprehensive_decision","score":0.0,"tokens":218861,"cost_usd":0.107646,"elapsed_time":138.0,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_025_comprehensive_decision_hard_hard019","category":"comprehensive_decision","score":0.0,"tokens":1647698,"cost_usd":0.339484,"elapsed_time":473.48,"gpr":0.0,"tpe":0.0,"ee":0.3529},{"task_id":"task_026_comprehensive_decision_medium_medium001","category":"comprehensive_decision","score":0.0,"tokens":856924,"cost_usd":0.163063,"elapsed_time":391.5,"gpr":0.125,"tpe":1.0,"ee":0.2083},{"task_id":"task_027_comprehensive_decision_medium_medium002","category":"comprehensive_decision","score":0.0,"tokens":181443,"cost_usd":0.078698,"elapsed_time":174.01,"gpr":0.1429,"tpe":1.0,"ee":0.625},{"task_id":"task_028_comprehensive_decision_medium_medium003","category":"comprehensive_decision","score":0.0,"tokens":155128,"cost_usd":0.034143,"elapsed_time":119.02,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_029_comprehensive_decision_medium_medium004","category":"comprehensive_decision","score":0.0,"tokens":2418200,"cost_usd":0.384197,"elapsed_time":457.47,"gpr":0.375,"tpe":0.7223,"ee":0.1163},{"task_id":"task_030_comprehensive_decision_medium_medium005","category":"comprehensive_decision","score":1.0,"tokens":401474,"cost_usd":0.068932,"elapsed_time":254.22,"gpr":1.0,"tpe":0.1243,"ee":0.1852},{"task_id":"task_031_comprehensive_decision_medium_medium006","category":"comprehensive_decision","score":0.0,"tokens":648387,"cost_usd":0.279421,"elapsed_time":120.3,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_032_comprehensive_decision_medium_medium007","category":"comprehensive_decision","score":0.0,"tokens":1513488,"cost_usd":0.278137,"elapsed_time":298.03,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_033_comprehensive_decision_medium_medium008","category":"comprehensive_decision","score":1.0,"tokens":387602,"cost_usd":0.077394,"elapsed_time":175.64,"gpr":1.0,"tpe":0.5316,"ee":0.3333},{"task_id":"task_034_comprehensive_decision_medium_medium009","category":"comprehensive_decision","score":1.0,"tokens":538784,"cost_usd":0.208717,"elapsed_time":119.8,"gpr":1.0,"tpe":0.9367,"ee":0.7143},{"task_id":"task_035_comprehensive_decision_medium_medium010","category":"comprehensive_decision","score":0.0,"tokens":281899,"cost_usd":0.092952,"elapsed_time":131.84,"gpr":0.0,"tpe":0.0,"ee":0.3846},{"task_id":"task_036_comprehensive_decision_medium_medium011","category":"comprehensive_decision","score":1.0,"tokens":1978128,"cost_usd":0.328693,"elapsed_time":407.62,"gpr":1.0,"tpe":0.0929,"ee":0.125},{"task_id":"task_037_comprehensive_decision_medium_medium012","category":"comprehensive_decision","score":0.0,"tokens":1327934,"cost_usd":0.260905,"elapsed_time":663.02,"gpr":0.25,"tpe":0.0471,"ee":0.1316},{"task_id":"task_038_comprehensive_decision_medium_medium013","category":"comprehensive_decision","score":1.0,"tokens":1534422,"cost_usd":0.246968,"elapsed_time":542.77,"gpr":1.0,"tpe":0.3958,"ee":0.0877},{"task_id":"task_039_comprehensive_decision_medium_medium014","category":"comprehensive_decision","score":0.0,"tokens":1039031,"cost_usd":0.193695,"elapsed_time":309.96,"gpr":0.5,"tpe":0.1668,"ee":0.2},{"task_id":"task_040_comprehensive_decision_medium_medium015","category":"comprehensive_decision","score":1.0,"tokens":211859,"cost_usd":0.091794,"elapsed_time":348.54,"gpr":1.0,"tpe":0.828,"ee":0.5556},{"task_id":"task_041_comprehensive_decision_medium_medium016","category":"comprehensive_decision","score":0.0,"tokens":1346918,"cost_usd":0.296538,"elapsed_time":488.62,"gpr":0.0,"tpe":0.0,"ee":0.1515},{"task_id":"task_042_comprehensive_decision_medium_medium017","category":"comprehensive_decision","score":0.0,"tokens":1012466,"cost_usd":0.227378,"elapsed_time":303.21,"gpr":0.0,"tpe":0.0,"ee":0.1786},{"task_id":"task_043_comprehensive_decision_medium_medium018","category":"comprehensive_decision","score":0.0,"tokens":599574,"cost_usd":0.274416,"elapsed_time":172.02,"gpr":0.0,"tpe":0.0,"ee":0.4167},{"task_id":"task_044_comprehensive_decision_medium_medium019","category":"comprehensive_decision","score":1.0,"tokens":1287836,"cost_usd":0.248228,"elapsed_time":444.07,"gpr":0.3333,"tpe":0.1094,"ee":0.1667},{"task_id":"task_045_comprehensive_decision_medium_medium020","category":"comprehensive_decision","score":0.0,"tokens":288915,"cost_usd":0.103329,"elapsed_time":123.78,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_046_comprehensive_decision_medium_medium021","category":"comprehensive_decision","score":0.0,"tokens":447321,"cost_usd":0.21316,"elapsed_time":160.08,"gpr":0.0,"tpe":0.0,"ee":0.4545},{"task_id":"task_047_comprehensive_decision_medium_medium022","category":"comprehensive_decision","score":0.0,"tokens":650624,"cost_usd":0.299178,"elapsed_time":125.96,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_048_comprehensive_decision_medium_medium023","category":"comprehensive_decision","score":1.0,"tokens":1017750,"cost_usd":0.17601,"elapsed_time":191.64,"gpr":1.0,"tpe":0.1568,"ee":0.1667},{"task_id":"task_049_comprehensive_decision_medium_medium024","category":"comprehensive_decision","score":1.0,"tokens":759951,"cost_usd":0.163829,"elapsed_time":355.25,"gpr":1.0,"tpe":0.2059,"ee":0.25},{"task_id":"task_050_comprehensive_decision_medium_medium025","category":"comprehensive_decision","score":0.0,"tokens":727694,"cost_usd":0.114989,"elapsed_time":303.67,"gpr":0.0,"tpe":0.0,"ee":0.1389},{"task_id":"task_051_comprehensive_decision_medium_medium026","category":"comprehensive_decision","score":1.0,"tokens":130082,"cost_usd":0.03277,"elapsed_time":74.5,"gpr":1.0,"tpe":0.729,"ee":0.625},{"task_id":"task_052_comprehensive_decision_medium_medium027","category":"comprehensive_decision","score":1.0,"tokens":1541837,"cost_usd":0.303115,"elapsed_time":510.49,"gpr":1.0,"tpe":0.0646,"ee":0.1613},{"task_id":"task_053_comprehensive_decision_medium_medium028","category":"comprehensive_decision","score":0.0,"tokens":423124,"cost_usd":0.221221,"elapsed_time":143.29,"gpr":0.0,"tpe":0.0,"ee":0.625},{"task_id":"task_054_comprehensive_decision_medium_medium029","category":"comprehensive_decision","score":0.0,"tokens":939956,"cost_usd":0.229697,"elapsed_time":132.05,"gpr":0.5,"tpe":0.9,"ee":0.5},{"task_id":"task_055_comprehensive_decision_medium_medium030","category":"comprehensive_decision","score":0.0,"tokens":883116,"cost_usd":0.205003,"elapsed_time":246.56,"gpr":0.1429,"tpe":1.0,"ee":0.4167},{"task_id":"task_056_comprehensive_decision_medium_medium031","category":"comprehensive_decision","score":0.0,"tokens":3767758,"cost_usd":0.709954,"elapsed_time":444.57,"gpr":0.6667,"tpe":0.0646,"ee":0.1111},{"task_id":"task_057_comprehensive_decision_medium_medium032","category":"comprehensive_decision","score":0.0,"tokens":801641,"cost_usd":0.187876,"elapsed_time":228.75,"gpr":0.0,"tpe":0.0,"ee":0.3125},{"task_id":"task_058_comprehensive_decision_medium_medium033","category":"comprehensive_decision","score":0.0,"tokens":44553,"cost_usd":0.007604,"elapsed_time":87.96,"gpr":0.25,"tpe":1.0,"ee":0.8333},{"task_id":"task_059_comprehensive_decision_medium_medium034","category":"comprehensive_decision","score":0.0,"tokens":758623,"cost_usd":0.169421,"elapsed_time":244.31,"gpr":0.0,"tpe":0.0,"ee":0.2381},{"task_id":"task_060_comprehensive_decision_medium_medium035","category":"comprehensive_decision","score":0.0,"tokens":667984,"cost_usd":0.238528,"elapsed_time":159.28,"gpr":0.0,"tpe":0.0,"ee":0.4545},{"task_id":"task_061_comprehensive_decision_medium_medium036","category":"comprehensive_decision","score":0.0,"tokens":73610,"cost_usd":0.016345,"elapsed_time":130.58,"gpr":0.0,"tpe":0.0,"ee":0.7143},{"task_id":"task_062_comprehensive_decision_medium_medium037","category":"comprehensive_decision","score":0.0,"tokens":205685,"cost_usd":0.0667,"elapsed_time":143.41,"gpr":0.0,"tpe":0.0,"ee":0.4167},{"task_id":"task_063_comprehensive_decision_medium_medium038","category":"comprehensive_decision","score":0.0,"tokens":3526254,"cost_usd":0.572121,"elapsed_time":1072.49,"gpr":0.3333,"tpe":0.0424,"ee":0.0519},{"task_id":"task_064_comprehensive_decision_medium_medium039","category":"comprehensive_decision","score":1.0,"tokens":375837,"cost_usd":0.096388,"elapsed_time":239.52,"gpr":1.0,"tpe":0.6827,"ee":0.3333},{"task_id":"task_065_comprehensive_decision_medium_medium040","category":"comprehensive_decision","score":0.0,"tokens":543808,"cost_usd":0.244775,"elapsed_time":163.74,"gpr":0.0,"tpe":0.0,"ee":0.4545},{"task_id":"task_066_comprehensive_decision_medium_medium041","category":"comprehensive_decision","score":1.0,"tokens":1160863,"cost_usd":0.191549,"elapsed_time":476.41,"gpr":1.0,"tpe":0.0424,"ee":0.1429},{"task_id":"task_067_comprehensive_decision_medium_medium042","category":"comprehensive_decision","score":0.0,"tokens":280848,"cost_usd":0.094386,"elapsed_time":140.51,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_068_comprehensive_decision_medium_medium043","category":"comprehensive_decision","score":0.0,"tokens":182134,"cost_usd":0.068585,"elapsed_time":163.51,"gpr":0.0,"tpe":0.0,"ee":0.5714},{"task_id":"task_069_comprehensive_decision_medium_medium044","category":"comprehensive_decision","score":0.0,"tokens":619235,"cost_usd":0.127855,"elapsed_time":149.49,"gpr":0.0,"tpe":0.0,"ee":0.3571},{"task_id":"task_070_comprehensive_decision_medium_medium045","category":"comprehensive_decision","score":0.0,"tokens":484522,"cost_usd":0.140848,"elapsed_time":234.73,"gpr":0.0,"tpe":0.0,"ee":0.3571},{"task_id":"task_071_enterprise_industry_analysis_easy_easy001","category":"enterprise_industry_analysis","score":0.0,"tokens":2045686,"cost_usd":0.300221,"elapsed_time":408.5,"gpr":0.0,"tpe":0.0,"ee":0.0652},{"task_id":"task_072_enterprise_industry_analysis_easy_easy002","category":"enterprise_industry_analysis","score":0.0,"tokens":2084219,"cost_usd":0.366108,"elapsed_time":247.36,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_073_enterprise_industry_analysis_easy_easy003","category":"enterprise_industry_analysis","score":1.0,"tokens":1098774,"cost_usd":0.216054,"elapsed_time":273.95,"gpr":1.0,"tpe":0.1094,"ee":0.125},{"task_id":"task_074_enterprise_industry_analysis_easy_easy004","category":"enterprise_industry_analysis","score":1.0,"tokens":776336,"cost_usd":0.159653,"elapsed_time":289.39,"gpr":1.0,"tpe":0.1668,"ee":0.15},{"task_id":"task_075_enterprise_industry_analysis_easy_easy005","category":"enterprise_industry_analysis","score":0.0,"tokens":1618947,"cost_usd":0.301357,"elapsed_time":173.28,"gpr":0.0,"tpe":0.0,"ee":0.1667},{"task_id":"task_076_enterprise_industry_analysis_easy_easy006","category":"enterprise_industry_analysis","score":1.0,"tokens":2016444,"cost_usd":0.347522,"elapsed_time":236.25,"gpr":1.0,"tpe":0.1346,"ee":0.1154},{"task_id":"task_077_enterprise_industry_analysis_easy_easy007","category":"enterprise_industry_analysis","score":0.0,"tokens":237285,"cost_usd":0.063129,"elapsed_time":155.33,"gpr":0.6667,"tpe":0.81,"ee":0.2308},{"task_id":"task_078_enterprise_industry_analysis_easy_easy008","category":"enterprise_industry_analysis","score":0.0,"tokens":2606154,"cost_usd":0.397253,"elapsed_time":504.81,"gpr":0.0,"tpe":0.0,"ee":0.0566},{"task_id":"task_079_enterprise_industry_analysis_easy_easy009","category":"enterprise_industry_analysis","score":0.0,"tokens":924722,"cost_usd":0.190099,"elapsed_time":345.77,"gpr":0.0,"tpe":0.0,"ee":0.1111},{"task_id":"task_080_enterprise_industry_analysis_easy_easy010","category":"enterprise_industry_analysis","score":1.0,"tokens":202175,"cost_usd":0.037836,"elapsed_time":147.76,"gpr":1.0,"tpe":0.3371,"ee":0.2143},{"task_id":"task_081_enterprise_industry_analysis_easy_easy011","category":"enterprise_industry_analysis","score":0.0,"tokens":523474,"cost_usd":0.111348,"elapsed_time":204.47,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_082_enterprise_industry_analysis_easy_easy012","category":"enterprise_industry_analysis","score":0.0,"tokens":422510,"cost_usd":0.090403,"elapsed_time":258.54,"gpr":0.0,"tpe":0.0,"ee":0.1667},{"task_id":"task_083_enterprise_industry_analysis_easy_easy013","category":"enterprise_industry_analysis","score":1.0,"tokens":883856,"cost_usd":0.181077,"elapsed_time":174.08,"gpr":1.0,"tpe":0.2967,"ee":0.1667},{"task_id":"task_084_enterprise_industry_analysis_easy_easy014","category":"enterprise_industry_analysis","score":1.0,"tokens":683840,"cost_usd":0.131182,"elapsed_time":202.43,"gpr":1.0,"tpe":0.2288,"ee":0.1765},{"task_id":"task_085_enterprise_industry_analysis_easy_easy015","category":"enterprise_industry_analysis","score":0.0,"tokens":647658,"cost_usd":0.243431,"elapsed_time":166.82,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_086_enterprise_industry_analysis_easy_easy016","category":"enterprise_industry_analysis","score":0.0,"tokens":1337463,"cost_usd":0.259962,"elapsed_time":457.9,"gpr":0.6667,"tpe":0.1426,"ee":0.0938},{"task_id":"task_087_enterprise_industry_analysis_easy_easy017","category":"enterprise_industry_analysis","score":0.0,"tokens":908481,"cost_usd":0.28528,"elapsed_time":229.12,"gpr":0.3333,"tpe":0.3487,"ee":0.2},{"task_id":"task_088_enterprise_industry_analysis_easy_easy018","category":"enterprise_industry_analysis","score":0.0,"tokens":566952,"cost_usd":0.118748,"elapsed_time":178.36,"gpr":0.0,"tpe":0.0,"ee":0.2143},{"task_id":"task_089_enterprise_industry_analysis_easy_easy019","category":"enterprise_industry_analysis","score":0.0,"tokens":139074,"cost_usd":0.032836,"elapsed_time":147.54,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_090_enterprise_industry_analysis_easy_easy020","category":"enterprise_industry_analysis","score":0.0,"tokens":653011,"cost_usd":0.285812,"elapsed_time":208.32,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_091_enterprise_industry_analysis_easy_easy021","category":"enterprise_industry_analysis","score":0.0,"tokens":346994,"cost_usd":0.059752,"elapsed_time":299.34,"gpr":0.3333,"tpe":0.1351,"ee":0.1304},{"task_id":"task_092_enterprise_industry_analysis_easy_easy022","category":"enterprise_industry_analysis","score":0.0,"tokens":1313574,"cost_usd":0.218981,"elapsed_time":654.08,"gpr":0.0,"tpe":0.0,"ee":0.0638},{"task_id":"task_093_enterprise_industry_analysis_easy_easy023","category":"enterprise_industry_analysis","score":1.0,"tokens":1639153,"cost_usd":0.297295,"elapsed_time":297.23,"gpr":1.0,"tpe":0.2059,"ee":0.1667},{"task_id":"task_094_enterprise_industry_analysis_easy_easy024","category":"enterprise_industry_analysis","score":1.0,"tokens":2102549,"cost_usd":0.409893,"elapsed_time":336.64,"gpr":1.0,"tpe":0.0545,"ee":0.0938},{"task_id":"task_095_enterprise_industry_analysis_easy_easy025","category":"enterprise_industry_analysis","score":1.0,"tokens":1259997,"cost_usd":0.292557,"elapsed_time":312.58,"gpr":1.0,"tpe":0.1501,"ee":0.1429},{"task_id":"task_096_enterprise_industry_analysis_easy_easy026","category":"enterprise_industry_analysis","score":0.0,"tokens":2635863,"cost_usd":0.480384,"elapsed_time":386.34,"gpr":0.3333,"tpe":0.2059,"ee":0.0938},{"task_id":"task_097_enterprise_industry_analysis_easy_easy027","category":"enterprise_industry_analysis","score":0.0,"tokens":751679,"cost_usd":0.118914,"elapsed_time":589.83,"gpr":0.3333,"tpe":0.0718,"ee":0.0789},{"task_id":"task_098_enterprise_industry_analysis_easy_easy028","category":"enterprise_industry_analysis","score":0.0,"tokens":270141,"cost_usd":0.057849,"elapsed_time":187.94,"gpr":0.0,"tpe":0.0,"ee":0.2143},{"task_id":"task_099_enterprise_industry_analysis_easy_easy029","category":"enterprise_industry_analysis","score":0.0,"tokens":1699325,"cost_usd":0.26691,"elapsed_time":476.69,"gpr":0.0,"tpe":0.0,"ee":0.0857},{"task_id":"task_100_enterprise_industry_analysis_easy_easy030","category":"enterprise_industry_analysis","score":0.0,"tokens":684181,"cost_usd":0.14028,"elapsed_time":260.58,"gpr":0.0,"tpe":0.0,"ee":0.1765},{"task_id":"task_101_enterprise_industry_analysis_easy_easy031","category":"enterprise_industry_analysis","score":0.0,"tokens":89563,"cost_usd":0.014589,"elapsed_time":163.28,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_102_enterprise_industry_analysis_easy_easy032","category":"enterprise_industry_analysis","score":0.0,"tokens":2864358,"cost_usd":0.436753,"elapsed_time":602.48,"gpr":0.3333,"tpe":0.0343,"ee":0.0714},{"task_id":"task_103_enterprise_industry_analysis_easy_easy033","category":"enterprise_industry_analysis","score":0.0,"tokens":481142,"cost_usd":0.080053,"elapsed_time":331.02,"gpr":0.0,"tpe":0.0,"ee":0.125},{"task_id":"task_104_enterprise_industry_analysis_easy_easy034","category":"enterprise_industry_analysis","score":0.0,"tokens":1529134,"cost_usd":0.293834,"elapsed_time":265.0,"gpr":0.3333,"tpe":0.2288,"ee":0.1765},{"task_id":"task_105_enterprise_industry_analysis_easy_easy035","category":"enterprise_industry_analysis","score":1.0,"tokens":268853,"cost_usd":0.046909,"elapsed_time":269.29,"gpr":1.0,"tpe":0.2288,"ee":0.1765},{"task_id":"task_106_enterprise_industry_analysis_easy_easy036","category":"enterprise_industry_analysis","score":1.0,"tokens":132591,"cost_usd":0.021395,"elapsed_time":132.18,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_107_enterprise_industry_analysis_easy_easy037","category":"enterprise_industry_analysis","score":1.0,"tokens":226152,"cost_usd":0.117436,"elapsed_time":70.99,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_108_enterprise_industry_analysis_easy_easy038","category":"enterprise_industry_analysis","score":1.0,"tokens":476583,"cost_usd":0.147262,"elapsed_time":242.66,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_109_enterprise_industry_analysis_easy_easy039","category":"enterprise_industry_analysis","score":0.0,"tokens":181623,"cost_usd":0.037216,"elapsed_time":155.13,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_110_enterprise_industry_analysis_easy_easy040","category":"enterprise_industry_analysis","score":1.0,"tokens":158174,"cost_usd":0.026196,"elapsed_time":175.06,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_111_enterprise_industry_analysis_easy_easy041","category":"enterprise_industry_analysis","score":0.0,"tokens":951214,"cost_usd":0.24358,"elapsed_time":214.86,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_112_enterprise_industry_analysis_easy_easy042","category":"enterprise_industry_analysis","score":0.0,"tokens":1296918,"cost_usd":0.245552,"elapsed_time":312.11,"gpr":0.0,"tpe":0.0,"ee":0.1364},{"task_id":"task_113_enterprise_industry_analysis_easy_easy043","category":"enterprise_industry_analysis","score":1.0,"tokens":607859,"cost_usd":0.102824,"elapsed_time":514.71,"gpr":1.0,"tpe":0.0424,"ee":0.0909},{"task_id":"task_114_enterprise_industry_analysis_easy_easy044","category":"enterprise_industry_analysis","score":1.0,"tokens":101898,"cost_usd":0.018496,"elapsed_time":85.77,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_115_enterprise_industry_analysis_easy_easy045","category":"enterprise_industry_analysis","score":0.0,"tokens":1427681,"cost_usd":0.24598,"elapsed_time":428.87,"gpr":0.0,"tpe":0.0,"ee":0.12},{"task_id":"task_116_enterprise_industry_analysis_easy_easy046","category":"enterprise_industry_analysis","score":0.0,"tokens":975993,"cost_usd":0.206643,"elapsed_time":299.78,"gpr":0.6667,"tpe":0.2542,"ee":0.1875},{"task_id":"task_117_enterprise_industry_analysis_easy_easy047","category":"enterprise_industry_analysis","score":1.0,"tokens":309002,"cost_usd":0.053376,"elapsed_time":266.88,"gpr":1.0,"tpe":0.2059,"ee":0.1667},{"task_id":"task_118_enterprise_industry_analysis_easy_easy048","category":"enterprise_industry_analysis","score":0.0,"tokens":160421,"cost_usd":0.035624,"elapsed_time":198.71,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_119_enterprise_industry_analysis_easy_easy049","category":"enterprise_industry_analysis","score":0.0,"tokens":301800,"cost_usd":0.058893,"elapsed_time":203.53,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_120_enterprise_industry_analysis_easy_easy050","category":"enterprise_industry_analysis","score":0.0,"tokens":402740,"cost_usd":0.078322,"elapsed_time":189.82,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_121_enterprise_industry_analysis_easy_easy051","category":"enterprise_industry_analysis","score":1.0,"tokens":155646,"cost_usd":0.027445,"elapsed_time":175.3,"gpr":1.0,"tpe":0.3874,"ee":0.25},{"task_id":"task_122_enterprise_industry_analysis_easy_easy052","category":"enterprise_industry_analysis","score":1.0,"tokens":93029,"cost_usd":0.017944,"elapsed_time":102.4,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_123_enterprise_industry_analysis_easy_easy053","category":"enterprise_industry_analysis","score":1.0,"tokens":615805,"cost_usd":0.124121,"elapsed_time":242.84,"gpr":1.0,"tpe":0.3138,"ee":0.2143},{"task_id":"task_124_enterprise_industry_analysis_easy_easy054","category":"enterprise_industry_analysis","score":1.0,"tokens":507857,"cost_usd":0.09734,"elapsed_time":291.71,"gpr":1.0,"tpe":0.1501,"ee":0.1429},{"task_id":"task_125_enterprise_industry_analysis_easy_easy055","category":"enterprise_industry_analysis","score":1.0,"tokens":533876,"cost_usd":0.085939,"elapsed_time":409.9,"gpr":1.0,"tpe":0.0798,"ee":0.1111},{"task_id":"task_126_enterprise_industry_analysis_easy_easy056","category":"enterprise_industry_analysis","score":0.0,"tokens":849183,"cost_usd":0.235113,"elapsed_time":274.06,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_127_enterprise_industry_analysis_easy_easy057","category":"enterprise_industry_analysis","score":1.0,"tokens":350926,"cost_usd":0.129863,"elapsed_time":106.12,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_128_enterprise_industry_analysis_easy_easy058","category":"enterprise_industry_analysis","score":0.0,"tokens":906759,"cost_usd":0.225492,"elapsed_time":176.18,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_129_enterprise_industry_analysis_easy_easy059","category":"enterprise_industry_analysis","score":1.0,"tokens":738503,"cost_usd":0.156349,"elapsed_time":302.47,"gpr":1.0,"tpe":0.1853,"ee":0.1579},{"task_id":"task_130_enterprise_industry_analysis_easy_easy060","category":"enterprise_industry_analysis","score":1.0,"tokens":104444,"cost_usd":0.021453,"elapsed_time":118.66,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_131_enterprise_industry_analysis_easy_easy061","category":"enterprise_industry_analysis","score":0.0,"tokens":277013,"cost_usd":0.099581,"elapsed_time":152.58,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_132_enterprise_industry_analysis_easy_easy062","category":"enterprise_industry_analysis","score":1.0,"tokens":248089,"cost_usd":0.126099,"elapsed_time":136.22,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_133_enterprise_industry_analysis_easy_easy063","category":"enterprise_industry_analysis","score":0.0,"tokens":183604,"cost_usd":0.064928,"elapsed_time":150.71,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_134_enterprise_industry_analysis_easy_easy064","category":"enterprise_industry_analysis","score":0.0,"tokens":160096,"cost_usd":0.061316,"elapsed_time":148.07,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_135_enterprise_industry_analysis_easy_easy065","category":"enterprise_industry_analysis","score":1.0,"tokens":68304,"cost_usd":0.022965,"elapsed_time":84.9,"gpr":1.0,"tpe":0.9667,"ee":0.75},{"task_id":"task_136_enterprise_industry_analysis_easy_easy066","category":"enterprise_industry_analysis","score":1.0,"tokens":278587,"cost_usd":0.056175,"elapsed_time":154.85,"gpr":1.0,"tpe":0.5216,"ee":0.2727},{"task_id":"task_137_enterprise_industry_analysis_easy_easy067","category":"enterprise_industry_analysis","score":1.0,"tokens":530799,"cost_usd":0.155422,"elapsed_time":136.87,"gpr":1.0,"tpe":0.7587,"ee":0.4286},{"task_id":"task_138_enterprise_industry_analysis_easy_easy068","category":"enterprise_industry_analysis","score":0.0,"tokens":176976,"cost_usd":0.0821,"elapsed_time":95.95,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_139_enterprise_industry_analysis_easy_easy069","category":"enterprise_industry_analysis","score":0.0,"tokens":184217,"cost_usd":0.041031,"elapsed_time":176.4,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_140_enterprise_industry_analysis_easy_easy070","category":"enterprise_industry_analysis","score":0.0,"tokens":186247,"cost_usd":0.037888,"elapsed_time":180.26,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_141_enterprise_industry_analysis_easy_easy071","category":"enterprise_industry_analysis","score":1.0,"tokens":120683,"cost_usd":0.02663,"elapsed_time":110.11,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_142_enterprise_industry_analysis_easy_easy072","category":"enterprise_industry_analysis","score":1.0,"tokens":735845,"cost_usd":0.181233,"elapsed_time":159.91,"gpr":1.0,"tpe":0.5708,"ee":0.3333},{"task_id":"task_143_enterprise_industry_analysis_easy_easy073","category":"enterprise_industry_analysis","score":0.0,"tokens":709066,"cost_usd":0.272416,"elapsed_time":344.58,"gpr":0.6667,"tpe":0.729,"ee":0.4286},{"task_id":"task_144_enterprise_industry_analysis_easy_easy074","category":"enterprise_industry_analysis","score":0.0,"tokens":599793,"cost_usd":0.237587,"elapsed_time":383.89,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_145_enterprise_industry_analysis_easy_easy075","category":"enterprise_industry_analysis","score":1.0,"tokens":59809,"cost_usd":0.010737,"elapsed_time":87.2,"gpr":1.0,"tpe":0.87,"ee":0.6},{"task_id":"task_146_enterprise_industry_analysis_easy_easy076","category":"enterprise_industry_analysis","score":1.0,"tokens":451366,"cost_usd":0.163396,"elapsed_time":98.76,"gpr":1.0,"tpe":0.783,"ee":0.5},{"task_id":"task_147_enterprise_industry_analysis_easy_easy077","category":"enterprise_industry_analysis","score":1.0,"tokens":330445,"cost_usd":0.097292,"elapsed_time":98.0,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_148_enterprise_industry_analysis_easy_easy078","category":"enterprise_industry_analysis","score":1.0,"tokens":167170,"cost_usd":0.052764,"elapsed_time":82.44,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_149_enterprise_industry_analysis_easy_easy079","category":"enterprise_industry_analysis","score":1.0,"tokens":85645,"cost_usd":0.018062,"elapsed_time":87.47,"gpr":1.0,"tpe":0.783,"ee":0.5},{"task_id":"task_150_enterprise_industry_analysis_easy_easy080","category":"enterprise_industry_analysis","score":1.0,"tokens":70444,"cost_usd":0.012217,"elapsed_time":106.75,"gpr":1.0,"tpe":0.843,"ee":0.5},{"task_id":"task_151_enterprise_industry_analysis_easy_easy081","category":"enterprise_industry_analysis","score":1.0,"tokens":458271,"cost_usd":0.169013,"elapsed_time":100.91,"gpr":1.0,"tpe":0.756,"ee":0.5},{"task_id":"task_152_enterprise_industry_analysis_easy_easy082","category":"enterprise_industry_analysis","score":1.0,"tokens":151074,"cost_usd":0.038431,"elapsed_time":128.53,"gpr":1.0,"tpe":0.7587,"ee":0.4286},{"task_id":"task_153_enterprise_industry_analysis_easy_easy083","category":"enterprise_industry_analysis","score":0.0,"tokens":322842,"cost_usd":0.101305,"elapsed_time":181.0,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_154_enterprise_industry_analysis_easy_easy084","category":"enterprise_industry_analysis","score":1.0,"tokens":166764,"cost_usd":0.052342,"elapsed_time":152.99,"gpr":1.0,"tpe":0.8002,"ee":0.375},{"task_id":"task_155_enterprise_industry_analysis_easy_easy085","category":"enterprise_industry_analysis","score":1.0,"tokens":902036,"cost_usd":0.201055,"elapsed_time":165.11,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_156_enterprise_industry_analysis_easy_easy086","category":"enterprise_industry_analysis","score":1.0,"tokens":476827,"cost_usd":0.118726,"elapsed_time":138.04,"gpr":1.0,"tpe":0.5137,"ee":0.3},{"task_id":"task_157_enterprise_industry_analysis_easy_easy088","category":"enterprise_industry_analysis","score":1.0,"tokens":222976,"cost_usd":0.036465,"elapsed_time":207.93,"gpr":1.0,"tpe":0.2542,"ee":0.1333},{"task_id":"task_158_enterprise_industry_analysis_easy_easy090","category":"enterprise_industry_analysis","score":0.0,"tokens":33005,"cost_usd":0.005667,"elapsed_time":153.01,"gpr":0.0,"tpe":0.0,"ee":0.75},{"task_id":"task_159_enterprise_industry_analysis_easy_easy091","category":"enterprise_industry_analysis","score":0.0,"tokens":273765,"cost_usd":0.085549,"elapsed_time":256.13,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_160_enterprise_industry_analysis_easy_easy092","category":"enterprise_industry_analysis","score":0.0,"tokens":406246,"cost_usd":0.121562,"elapsed_time":226.11,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_161_enterprise_industry_analysis_easy_easy093","category":"enterprise_industry_analysis","score":1.0,"tokens":221802,"cost_usd":0.038525,"elapsed_time":190.04,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_162_enterprise_industry_analysis_easy_easy094","category":"enterprise_industry_analysis","score":0.0,"tokens":818845,"cost_usd":0.246584,"elapsed_time":467.62,"gpr":0.0,"tpe":0.0,"ee":0.2143},{"task_id":"task_163_enterprise_industry_analysis_easy_easy095","category":"enterprise_industry_analysis","score":0.0,"tokens":1030411,"cost_usd":0.170715,"elapsed_time":385.47,"gpr":0.0,"tpe":0.0,"ee":0.1034},{"task_id":"task_164_enterprise_industry_analysis_easy_easy096","category":"enterprise_industry_analysis","score":1.0,"tokens":204218,"cost_usd":0.044575,"elapsed_time":241.94,"gpr":1.0,"tpe":0.3874,"ee":0.25},{"task_id":"task_165_enterprise_industry_analysis_easy_easy097","category":"enterprise_industry_analysis","score":0.0,"tokens":1097388,"cost_usd":0.255592,"elapsed_time":219.61,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_166_enterprise_industry_analysis_easy_easy098","category":"enterprise_industry_analysis","score":1.0,"tokens":524401,"cost_usd":0.133108,"elapsed_time":132.67,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_167_enterprise_industry_analysis_easy_easy099","category":"enterprise_industry_analysis","score":0.0,"tokens":459460,"cost_usd":0.092876,"elapsed_time":206.47,"gpr":0.0,"tpe":0.0,"ee":0.1579},{"task_id":"task_168_enterprise_industry_analysis_easy_easy100","category":"enterprise_industry_analysis","score":0.0,"tokens":280515,"cost_usd":0.04664,"elapsed_time":180.99,"gpr":0.0,"tpe":0.0,"ee":0.1667},{"task_id":"task_169_enterprise_industry_analysis_easy_easy101","category":"enterprise_industry_analysis","score":0.0,"tokens":349083,"cost_usd":0.057375,"elapsed_time":238.41,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_170_enterprise_industry_analysis_easy_easy102","category":"enterprise_industry_analysis","score":1.0,"tokens":123121,"cost_usd":0.019505,"elapsed_time":76.43,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_171_enterprise_industry_analysis_easy_easy103","category":"enterprise_industry_analysis","score":1.0,"tokens":163983,"cost_usd":0.030021,"elapsed_time":149.24,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_172_enterprise_industry_analysis_easy_easy104","category":"enterprise_industry_analysis","score":1.0,"tokens":176862,"cost_usd":0.030652,"elapsed_time":178.47,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_173_enterprise_industry_analysis_easy_easy105","category":"enterprise_industry_analysis","score":0.0,"tokens":257256,"cost_usd":0.042773,"elapsed_time":205.25,"gpr":0.0,"tpe":0.0,"ee":0.1765},{"task_id":"task_174_enterprise_industry_analysis_easy_easy106","category":"enterprise_industry_analysis","score":1.0,"tokens":398266,"cost_usd":0.063218,"elapsed_time":307.0,"gpr":1.0,"tpe":0.0886,"ee":0.1154},{"task_id":"task_175_enterprise_industry_analysis_easy_easy107","category":"enterprise_industry_analysis","score":0.0,"tokens":655756,"cost_usd":0.118189,"elapsed_time":291.21,"gpr":0.0,"tpe":0.0,"ee":0.1154},{"task_id":"task_176_enterprise_industry_analysis_easy_easy108","category":"enterprise_industry_analysis","score":0.0,"tokens":779538,"cost_usd":0.186984,"elapsed_time":162.4,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_177_enterprise_industry_analysis_easy_easy109","category":"enterprise_industry_analysis","score":0.0,"tokens":957849,"cost_usd":0.176223,"elapsed_time":283.39,"gpr":0.0,"tpe":0.0,"ee":0.1304},{"task_id":"task_178_enterprise_industry_analysis_easy_easy110","category":"enterprise_industry_analysis","score":0.0,"tokens":169417,"cost_usd":0.033048,"elapsed_time":110.45,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_179_enterprise_industry_analysis_easy_easy111","category":"enterprise_industry_analysis","score":0.0,"tokens":817282,"cost_usd":0.125798,"elapsed_time":375.86,"gpr":0.0,"tpe":0.0,"ee":0.0833},{"task_id":"task_180_enterprise_industry_analysis_easy_easy112","category":"enterprise_industry_analysis","score":0.0,"tokens":482541,"cost_usd":0.116197,"elapsed_time":164.59,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_181_enterprise_industry_analysis_easy_easy113","category":"enterprise_industry_analysis","score":0.0,"tokens":568067,"cost_usd":0.287729,"elapsed_time":198.68,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_182_enterprise_industry_analysis_easy_easy114","category":"enterprise_industry_analysis","score":0.0,"tokens":284137,"cost_usd":0.093254,"elapsed_time":140.2,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_183_enterprise_industry_analysis_easy_easy115","category":"enterprise_industry_analysis","score":0.0,"tokens":262621,"cost_usd":0.056615,"elapsed_time":162.77,"gpr":0.0,"tpe":0.0,"ee":0.2143},{"task_id":"task_184_enterprise_industry_analysis_easy_easy116","category":"enterprise_industry_analysis","score":0.0,"tokens":554385,"cost_usd":0.17649,"elapsed_time":175.07,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_185_enterprise_industry_analysis_easy_easy117","category":"enterprise_industry_analysis","score":0.0,"tokens":965570,"cost_usd":0.216226,"elapsed_time":181.52,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_186_enterprise_industry_analysis_medium_medium001","category":"enterprise_industry_analysis","score":1.0,"tokens":189262,"cost_usd":0.034129,"elapsed_time":152.99,"gpr":1.0,"tpe":0.4783,"ee":0.3636},{"task_id":"task_187_enterprise_industry_analysis_medium_medium002","category":"enterprise_industry_analysis","score":1.0,"tokens":2672853,"cost_usd":0.47318,"elapsed_time":298.09,"gpr":1.0,"tpe":0.1658,"ee":0.1481},{"task_id":"task_188_enterprise_industry_analysis_medium_medium003","category":"enterprise_industry_analysis","score":0.0,"tokens":817736,"cost_usd":0.156972,"elapsed_time":271.99,"gpr":0.0,"tpe":0.0,"ee":0.1905},{"task_id":"task_189_enterprise_industry_analysis_medium_medium004","category":"enterprise_industry_analysis","score":0.0,"tokens":2048810,"cost_usd":0.350178,"elapsed_time":444.8,"gpr":0.0,"tpe":0.0,"ee":0.1176},{"task_id":"task_190_enterprise_industry_analysis_medium_medium005","category":"enterprise_industry_analysis","score":1.0,"tokens":178745,"cost_usd":0.039026,"elapsed_time":91.51,"gpr":1.0,"tpe":0.729,"ee":0.5714},{"task_id":"task_191_enterprise_industry_analysis_medium_medium006","category":"enterprise_industry_analysis","score":0.0,"tokens":2705445,"cost_usd":0.513784,"elapsed_time":542.51,"gpr":0.0,"tpe":0.0,"ee":0.16},{"task_id":"task_192_enterprise_industry_analysis_medium_medium007","category":"enterprise_industry_analysis","score":0.0,"tokens":209277,"cost_usd":0.038796,"elapsed_time":129.43,"gpr":0.0,"tpe":0.0,"ee":0.3636},{"task_id":"task_193_enterprise_industry_analysis_medium_medium008","category":"enterprise_industry_analysis","score":0.0,"tokens":585411,"cost_usd":0.224141,"elapsed_time":149.77,"gpr":0.0,"tpe":0.0,"ee":0.4444},{"task_id":"task_194_enterprise_industry_analysis_medium_medium009","category":"enterprise_industry_analysis","score":1.0,"tokens":594107,"cost_usd":0.114992,"elapsed_time":236.5,"gpr":1.0,"tpe":0.2959,"ee":0.2105},{"task_id":"task_195_enterprise_industry_analysis_medium_medium010","category":"enterprise_industry_analysis","score":0.0,"tokens":472045,"cost_usd":0.084656,"elapsed_time":190.59,"gpr":0.0,"tpe":0.0,"ee":0.2667},{"task_id":"task_196_enterprise_industry_analysis_medium_medium011","category":"enterprise_industry_analysis","score":0.0,"tokens":950776,"cost_usd":0.184707,"elapsed_time":352.84,"gpr":0.2,"tpe":0.9,"ee":0.1111},{"task_id":"task_197_enterprise_industry_analysis_medium_medium012","category":"enterprise_industry_analysis","score":0.0,"tokens":1728196,"cost_usd":0.409797,"elapsed_time":273.63,"gpr":0.0,"tpe":0.0,"ee":0.1481},{"task_id":"task_198_enterprise_industry_analysis_medium_medium013","category":"enterprise_industry_analysis","score":0.0,"tokens":1141743,"cost_usd":0.364735,"elapsed_time":169.35,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_199_enterprise_industry_analysis_medium_medium014","category":"enterprise_industry_analysis","score":1.0,"tokens":892038,"cost_usd":0.148316,"elapsed_time":289.54,"gpr":1.0,"tpe":0.2773,"ee":0.1786},{"task_id":"task_200_enterprise_industry_analysis_medium_medium015","category":"enterprise_industry_analysis","score":0.0,"tokens":342088,"cost_usd":0.075239,"elapsed_time":176.52,"gpr":0.0,"tpe":0.0,"ee":0.3846},{"task_id":"task_201_enterprise_industry_analysis_medium_medium016","category":"enterprise_industry_analysis","score":0.0,"tokens":143012,"cost_usd":0.034066,"elapsed_time":116.15,"gpr":0.4,"tpe":0.905,"ee":0.5},{"task_id":"task_202_enterprise_industry_analysis_medium_medium017","category":"enterprise_industry_analysis","score":0.0,"tokens":547287,"cost_usd":0.182167,"elapsed_time":170.39,"gpr":0.0,"tpe":0.0,"ee":0.3571},{"task_id":"task_203_enterprise_industry_analysis_medium_medium018","category":"enterprise_industry_analysis","score":1.0,"tokens":1076381,"cost_usd":0.219405,"elapsed_time":189.53,"gpr":1.0,"tpe":0.3874,"ee":0.3571},{"task_id":"task_204_enterprise_industry_analysis_medium_medium019","category":"enterprise_industry_analysis","score":0.0,"tokens":2060322,"cost_usd":0.330443,"elapsed_time":699.31,"gpr":0.0,"tpe":0.0,"ee":0.0862},{"task_id":"task_205_enterprise_industry_analysis_medium_medium020","category":"enterprise_industry_analysis","score":1.0,"tokens":100963,"cost_usd":0.025064,"elapsed_time":76.76,"gpr":1.0,"tpe":0.729,"ee":0.625},{"task_id":"task_206_enterprise_industry_analysis_medium_medium021","category":"enterprise_industry_analysis","score":0.0,"tokens":1220234,"cost_usd":0.305685,"elapsed_time":249.33,"gpr":0.0,"tpe":0.0,"ee":0.2381},{"task_id":"task_207_enterprise_industry_analysis_medium_medium022","category":"enterprise_industry_analysis","score":1.0,"tokens":327258,"cost_usd":0.131189,"elapsed_time":78.02,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_208_enterprise_industry_analysis_medium_medium023","category":"enterprise_industry_analysis","score":1.0,"tokens":1040929,"cost_usd":0.180742,"elapsed_time":564.54,"gpr":1.0,"tpe":0.3058,"ee":0.1471},{"task_id":"task_209_enterprise_industry_analysis_medium_medium024","category":"enterprise_industry_analysis","score":0.0,"tokens":291715,"cost_usd":0.085355,"elapsed_time":156.37,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_210_enterprise_industry_analysis_medium_medium025","category":"enterprise_industry_analysis","score":1.0,"tokens":1109011,"cost_usd":0.186064,"elapsed_time":380.62,"gpr":1.0,"tpe":0.1785,"ee":0.1562},{"task_id":"task_211_enterprise_industry_analysis_medium_medium026","category":"enterprise_industry_analysis","score":0.0,"tokens":1200397,"cost_usd":0.248399,"elapsed_time":283.71,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_212_enterprise_industry_analysis_medium_medium027","category":"enterprise_industry_analysis","score":0.0,"tokens":859888,"cost_usd":0.161504,"elapsed_time":421.39,"gpr":0.8,"tpe":0.5329,"ee":0.2174},{"task_id":"task_213_enterprise_industry_analysis_medium_medium028","category":"enterprise_industry_analysis","score":1.0,"tokens":1627430,"cost_usd":0.34518,"elapsed_time":199.65,"gpr":1.0,"tpe":0.6115,"ee":0.2941},{"task_id":"task_214_enterprise_industry_analysis_medium_medium029","category":"enterprise_industry_analysis","score":1.0,"tokens":250363,"cost_usd":0.050987,"elapsed_time":218.71,"gpr":1.0,"tpe":0.5661,"ee":0.3846},{"task_id":"task_215_enterprise_industry_analysis_medium_medium030","category":"enterprise_industry_analysis","score":1.0,"tokens":446426,"cost_usd":0.075903,"elapsed_time":351.17,"gpr":1.0,"tpe":0.0798,"ee":0.1724},{"task_id":"task_216_enterprise_industry_analysis_medium_medium031","category":"enterprise_industry_analysis","score":1.0,"tokens":541188,"cost_usd":0.086764,"elapsed_time":351.42,"gpr":1.0,"tpe":0.1603,"ee":0.1515},{"task_id":"task_217_enterprise_industry_analysis_medium_medium032","category":"enterprise_industry_analysis","score":1.0,"tokens":257030,"cost_usd":0.047704,"elapsed_time":192.94,"gpr":1.0,"tpe":0.6359,"ee":0.3571},{"task_id":"task_218_enterprise_industry_analysis_medium_medium033","category":"enterprise_industry_analysis","score":1.0,"tokens":197479,"cost_usd":0.042856,"elapsed_time":138.99,"gpr":1.0,"tpe":0.6855,"ee":0.5},{"task_id":"task_219_enterprise_industry_analysis_medium_medium034","category":"enterprise_industry_analysis","score":0.0,"tokens":716483,"cost_usd":0.320187,"elapsed_time":211.47,"gpr":0.0,"tpe":0.0,"ee":0.3571},{"task_id":"task_220_enterprise_industry_analysis_medium_medium035","category":"enterprise_industry_analysis","score":0.0,"tokens":1086172,"cost_usd":0.278165,"elapsed_time":198.12,"gpr":0.6,"tpe":0.795,"ee":0.4167},{"task_id":"task_221_enterprise_industry_analysis_medium_medium036","category":"enterprise_industry_analysis","score":0.0,"tokens":1488368,"cost_usd":0.331599,"elapsed_time":329.76,"gpr":0.4,"tpe":0.2059,"ee":0.2381},{"task_id":"task_222_enterprise_industry_analysis_medium_medium037","category":"enterprise_industry_analysis","score":1.0,"tokens":357183,"cost_usd":0.08923,"elapsed_time":253.89,"gpr":1.0,"tpe":0.8775,"ee":0.3571},{"task_id":"task_223_enterprise_industry_analysis_medium_medium038","category":"enterprise_industry_analysis","score":0.0,"tokens":358966,"cost_usd":0.116398,"elapsed_time":415.64,"gpr":0.4,"tpe":0.5905,"ee":0.4545},{"task_id":"task_224_enterprise_industry_analysis_medium_medium039","category":"enterprise_industry_analysis","score":1.0,"tokens":603910,"cost_usd":0.207457,"elapsed_time":221.56,"gpr":1.0,"tpe":0.8293,"ee":0.5},{"task_id":"task_225_enterprise_industry_analysis_medium_medium040","category":"enterprise_industry_analysis","score":1.0,"tokens":514693,"cost_usd":0.105124,"elapsed_time":226.95,"gpr":1.0,"tpe":0.5867,"ee":0.3333},{"task_id":"task_226_enterprise_industry_analysis_medium_medium041","category":"enterprise_industry_analysis","score":1.0,"tokens":142920,"cost_usd":0.038546,"elapsed_time":97.87,"gpr":1.0,"tpe":0.9458,"ee":0.625},{"task_id":"task_227_enterprise_industry_analysis_medium_medium042","category":"enterprise_industry_analysis","score":0.0,"tokens":168423,"cost_usd":0.067331,"elapsed_time":145.03,"gpr":0.0,"tpe":0.0,"ee":0.625},{"task_id":"task_228_enterprise_industry_analysis_medium_medium043","category":"enterprise_industry_analysis","score":0.0,"tokens":701059,"cost_usd":0.126257,"elapsed_time":195.77,"gpr":0.6,"tpe":0.9,"ee":0.2778},{"task_id":"task_229_enterprise_industry_analysis_medium_medium044","category":"enterprise_industry_analysis","score":1.0,"tokens":219716,"cost_usd":0.067316,"elapsed_time":342.7,"gpr":1.0,"tpe":0.5905,"ee":0.5},{"task_id":"task_230_enterprise_industry_analysis_medium_medium045","category":"enterprise_industry_analysis","score":0.0,"tokens":934274,"cost_usd":0.209211,"elapsed_time":444.74,"gpr":0.0,"tpe":0.0,"ee":0.3125},{"task_id":"task_231_enterprise_industry_analysis_medium_medium046","category":"enterprise_industry_analysis","score":1.0,"tokens":522486,"cost_usd":0.097523,"elapsed_time":264.64,"gpr":1.0,"tpe":0.2339,"ee":0.2632},{"task_id":"task_232_enterprise_industry_analysis_medium_medium047","category":"enterprise_industry_analysis","score":1.0,"tokens":340197,"cost_usd":0.054495,"elapsed_time":361.3,"gpr":1.0,"tpe":0.1783,"ee":0.2273},{"task_id":"task_233_enterprise_industry_analysis_medium_medium048","category":"enterprise_industry_analysis","score":1.0,"tokens":367616,"cost_usd":0.063174,"elapsed_time":296.08,"gpr":1.0,"tpe":0.5235,"ee":0.25},{"task_id":"task_234_enterprise_industry_analysis_medium_medium049","category":"enterprise_industry_analysis","score":1.0,"tokens":448229,"cost_usd":0.086038,"elapsed_time":376.99,"gpr":1.0,"tpe":0.4575,"ee":0.25},{"task_id":"task_235_enterprise_industry_analysis_medium_medium050","category":"enterprise_industry_analysis","score":1.0,"tokens":868761,"cost_usd":0.148322,"elapsed_time":838.66,"gpr":1.0,"tpe":0.0355,"ee":0.1316},{"task_id":"task_236_enterprise_industry_analysis_medium_medium051","category":"enterprise_industry_analysis","score":0.0,"tokens":343404,"cost_usd":0.07319,"elapsed_time":189.13,"gpr":0.6,"tpe":0.4695,"ee":0.3125},{"task_id":"task_237_enterprise_industry_analysis_medium_medium052","category":"enterprise_industry_analysis","score":0.0,"tokens":741785,"cost_usd":0.226541,"elapsed_time":210.5,"gpr":0.6,"tpe":0.8102,"ee":0.3846},{"task_id":"task_238_enterprise_industry_analysis_medium_medium053","category":"enterprise_industry_analysis","score":1.0,"tokens":595894,"cost_usd":0.154786,"elapsed_time":181.02,"gpr":1.0,"tpe":0.6561,"ee":0.5556},{"task_id":"task_239_enterprise_industry_analysis_medium_medium054","category":"enterprise_industry_analysis","score":0.0,"tokens":1254029,"cost_usd":0.237404,"elapsed_time":461.12,"gpr":0.4,"tpe":0.0646,"ee":0.1429},{"task_id":"task_240_enterprise_industry_analysis_medium_medium055","category":"enterprise_industry_analysis","score":1.0,"tokens":290428,"cost_usd":0.058689,"elapsed_time":280.24,"gpr":1.0,"tpe":0.6247,"ee":0.3333},{"task_id":"task_241_enterprise_industry_analysis_medium_medium056","category":"enterprise_industry_analysis","score":0.0,"tokens":710287,"cost_usd":0.129497,"elapsed_time":338.78,"gpr":0.6,"tpe":0.1587,"ee":0.2083},{"task_id":"task_242_enterprise_industry_analysis_medium_medium057","category":"enterprise_industry_analysis","score":1.0,"tokens":273329,"cost_usd":0.052018,"elapsed_time":189.53,"gpr":1.0,"tpe":0.3936,"ee":0.3125},{"task_id":"task_243_enterprise_industry_analysis_medium_medium058","category":"enterprise_industry_analysis","score":0.0,"tokens":363313,"cost_usd":0.064263,"elapsed_time":192.84,"gpr":0.8,"tpe":0.4668,"ee":0.2941},{"task_id":"task_244_enterprise_industry_analysis_medium_medium059","category":"enterprise_industry_analysis","score":1.0,"tokens":185169,"cost_usd":0.037372,"elapsed_time":157.11,"gpr":1.0,"tpe":0.5232,"ee":0.4167},{"task_id":"task_245_enterprise_industry_analysis_medium_medium060","category":"enterprise_industry_analysis","score":1.0,"tokens":293993,"cost_usd":0.057854,"elapsed_time":165.71,"gpr":1.0,"tpe":0.4046,"ee":0.3571},{"task_id":"task_246_enterprise_industry_analysis_medium_medium061","category":"enterprise_industry_analysis","score":1.0,"tokens":283107,"cost_usd":0.052687,"elapsed_time":130.95,"gpr":1.0,"tpe":0.5635,"ee":0.3846},{"task_id":"task_247_enterprise_industry_analysis_medium_medium062","category":"enterprise_industry_analysis","score":1.0,"tokens":593661,"cost_usd":0.092277,"elapsed_time":280.35,"gpr":1.0,"tpe":0.0474,"ee":0.1389},{"task_id":"task_248_enterprise_industry_analysis_medium_medium063","category":"enterprise_industry_analysis","score":1.0,"tokens":641381,"cost_usd":0.129068,"elapsed_time":197.02,"gpr":1.0,"tpe":0.3932,"ee":0.2353},{"task_id":"task_249_enterprise_industry_analysis_medium_medium064","category":"enterprise_industry_analysis","score":1.0,"tokens":758261,"cost_usd":0.137796,"elapsed_time":332.44,"gpr":1.0,"tpe":0.2535,"ee":0.1667},{"task_id":"task_250_enterprise_industry_analysis_medium_medium065","category":"enterprise_industry_analysis","score":1.0,"tokens":139516,"cost_usd":0.027267,"elapsed_time":95.36,"gpr":1.0,"tpe":0.7805,"ee":0.4},{"task_id":"task_251_enterprise_industry_analysis_medium_medium066","category":"enterprise_industry_analysis","score":1.0,"tokens":334896,"cost_usd":0.060689,"elapsed_time":195.19,"gpr":1.0,"tpe":0.2824,"ee":0.25},{"task_id":"task_252_enterprise_industry_analysis_medium_medium067","category":"enterprise_industry_analysis","score":1.0,"tokens":2347579,"cost_usd":0.332541,"elapsed_time":246.21,"gpr":1.0,"tpe":0.0861,"ee":0.1333},{"task_id":"task_253_enterprise_industry_analysis_medium_medium068","category":"enterprise_industry_analysis","score":0.0,"tokens":458980,"cost_usd":0.075494,"elapsed_time":277.13,"gpr":0.75,"tpe":0.1356,"ee":0.1667},{"task_id":"task_254_enterprise_industry_analysis_medium_medium069","category":"enterprise_industry_analysis","score":0.0,"tokens":428241,"cost_usd":0.137789,"elapsed_time":165.28,"gpr":0.75,"tpe":0.3487,"ee":0.2667},{"task_id":"task_255_enterprise_industry_analysis_medium_medium070","category":"enterprise_industry_analysis","score":0.0,"tokens":168525,"cost_usd":0.033935,"elapsed_time":81.62,"gpr":0.3333,"tpe":0.81,"ee":0.5714},{"task_id":"task_256_enterprise_industry_analysis_medium_medium071","category":"enterprise_industry_analysis","score":0.0,"tokens":1072927,"cost_usd":0.165655,"elapsed_time":370.39,"gpr":0.0,"tpe":0.0,"ee":0.1053},{"task_id":"task_257_enterprise_industry_analysis_medium_medium072","category":"enterprise_industry_analysis","score":0.0,"tokens":1046043,"cost_usd":0.18787,"elapsed_time":269.4,"gpr":0.25,"tpe":0.1853,"ee":0.1739},{"task_id":"task_258_enterprise_industry_analysis_medium_medium073","category":"enterprise_industry_analysis","score":0.0,"tokens":913297,"cost_usd":0.137825,"elapsed_time":592.92,"gpr":0.5,"tpe":0.0804,"ee":0.0769},{"task_id":"task_259_enterprise_industry_analysis_medium_medium074","category":"enterprise_industry_analysis","score":1.0,"tokens":1817075,"cost_usd":0.308815,"elapsed_time":304.45,"gpr":1.0,"tpe":0.2145,"ee":0.16},{"task_id":"task_260_enterprise_industry_analysis_medium_medium075","category":"enterprise_industry_analysis","score":1.0,"tokens":1004567,"cost_usd":0.15737,"elapsed_time":264.02,"gpr":1.0,"tpe":0.2041,"ee":0.1905},{"task_id":"task_261_enterprise_industry_analysis_medium_medium076","category":"enterprise_industry_analysis","score":0.0,"tokens":934412,"cost_usd":0.168658,"elapsed_time":177.67,"gpr":0.5,"tpe":0.5049,"ee":0.2},{"task_id":"task_262_enterprise_industry_analysis_medium_medium077","category":"enterprise_industry_analysis","score":1.0,"tokens":186165,"cost_usd":0.062637,"elapsed_time":78.08,"gpr":1.0,"tpe":0.6679,"ee":0.4444},{"task_id":"task_263_enterprise_industry_analysis_medium_medium078","category":"enterprise_industry_analysis","score":1.0,"tokens":185012,"cost_usd":0.036048,"elapsed_time":220.22,"gpr":1.0,"tpe":0.4597,"ee":0.3077},{"task_id":"task_264_enterprise_industry_analysis_medium_medium079","category":"enterprise_industry_analysis","score":1.0,"tokens":419529,"cost_usd":0.073835,"elapsed_time":216.78,"gpr":1.0,"tpe":0.5064,"ee":0.2105},{"task_id":"task_265_enterprise_industry_analysis_medium_medium080","category":"enterprise_industry_analysis","score":0.0,"tokens":250601,"cost_usd":0.05609,"elapsed_time":203.84,"gpr":0.25,"tpe":0.6561,"ee":0.5},{"task_id":"task_266_enterprise_industry_analysis_medium_medium081","category":"enterprise_industry_analysis","score":1.0,"tokens":261526,"cost_usd":0.04645,"elapsed_time":221.76,"gpr":1.0,"tpe":0.7011,"ee":0.2667},{"task_id":"task_267_enterprise_industry_analysis_medium_medium082","category":"enterprise_industry_analysis","score":0.0,"tokens":174294,"cost_usd":0.054849,"elapsed_time":68.68,"gpr":0.0,"tpe":0.0,"ee":0.8},{"task_id":"task_268_enterprise_industry_analysis_medium_medium083","category":"enterprise_industry_analysis","score":0.0,"tokens":733983,"cost_usd":0.120519,"elapsed_time":305.42,"gpr":0.0,"tpe":0.0,"ee":0.1333},{"task_id":"task_269_enterprise_industry_analysis_medium_medium084","category":"enterprise_industry_analysis","score":0.0,"tokens":909935,"cost_usd":0.165384,"elapsed_time":204.8,"gpr":0.25,"tpe":0.729,"ee":0.2222},{"task_id":"task_270_enterprise_industry_analysis_medium_medium085","category":"enterprise_industry_analysis","score":1.0,"tokens":506044,"cost_usd":0.09847,"elapsed_time":278.99,"gpr":1.0,"tpe":0.3304,"ee":0.1667},{"task_id":"task_271_enterprise_industry_analysis_medium_medium086","category":"enterprise_industry_analysis","score":0.0,"tokens":894806,"cost_usd":0.216051,"elapsed_time":196.36,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_272_enterprise_industry_analysis_medium_medium087","category":"enterprise_industry_analysis","score":0.0,"tokens":362432,"cost_usd":0.060485,"elapsed_time":237.83,"gpr":0.75,"tpe":0.5916,"ee":0.2857},{"task_id":"task_273_enterprise_industry_analysis_medium_medium088","category":"enterprise_industry_analysis","score":1.0,"tokens":1659708,"cost_usd":0.291751,"elapsed_time":816.69,"gpr":1.0,"tpe":0.0203,"ee":0.0976},{"task_id":"task_274_enterprise_industry_analysis_medium_medium089","category":"enterprise_industry_analysis","score":0.0,"tokens":556078,"cost_usd":0.183334,"elapsed_time":282.23,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_275_enterprise_industry_analysis_medium_medium090","category":"enterprise_industry_analysis","score":0.0,"tokens":1070984,"cost_usd":0.172787,"elapsed_time":501.42,"gpr":0.0,"tpe":0.0,"ee":0.129},{"task_id":"task_276_enterprise_industry_analysis_medium_medium091","category":"enterprise_industry_analysis","score":0.0,"tokens":264980,"cost_usd":0.046927,"elapsed_time":315.94,"gpr":1.0,"tpe":0.1668,"ee":0.1905},{"task_id":"task_277_enterprise_industry_analysis_medium_medium092","category":"enterprise_industry_analysis","score":1.0,"tokens":2610784,"cost_usd":0.415962,"elapsed_time":833.37,"gpr":1.0,"tpe":0.0237,"ee":0.0833},{"task_id":"task_278_enterprise_industry_analysis_medium_medium093","category":"enterprise_industry_analysis","score":0.0,"tokens":945219,"cost_usd":0.158605,"elapsed_time":514.38,"gpr":0.5,"tpe":0.2415,"ee":0.1481},{"task_id":"task_279_enterprise_industry_analysis_medium_medium094","category":"enterprise_industry_analysis","score":0.0,"tokens":2129528,"cost_usd":0.307278,"elapsed_time":901.04,"gpr":0.75,"tpe":0.0021,"ee":0.0519},{"task_id":"task_280_enterprise_industry_analysis_medium_medium095","category":"enterprise_industry_analysis","score":0.0,"tokens":293685,"cost_usd":0.057505,"elapsed_time":223.17,"gpr":0.75,"tpe":0.3757,"ee":0.2222},{"task_id":"task_281_enterprise_industry_analysis_medium_medium096","category":"enterprise_industry_analysis","score":1.0,"tokens":231790,"cost_usd":0.044036,"elapsed_time":231.33,"gpr":1.0,"tpe":0.5351,"ee":0.2857},{"task_id":"task_282_enterprise_industry_analysis_medium_medium097","category":"enterprise_industry_analysis","score":1.0,"tokens":284007,"cost_usd":0.100307,"elapsed_time":139.98,"gpr":1.0,"tpe":0.8145,"ee":0.5714},{"task_id":"task_283_enterprise_industry_analysis_medium_medium098","category":"enterprise_industry_analysis","score":0.0,"tokens":1980780,"cost_usd":0.304582,"elapsed_time":805.09,"gpr":0.5,"tpe":0.0036,"ee":0.0541},{"task_id":"task_284_enterprise_industry_analysis_medium_medium099","category":"enterprise_industry_analysis","score":0.0,"tokens":383951,"cost_usd":0.092921,"elapsed_time":263.16,"gpr":0.75,"tpe":0.3383,"ee":0.25},{"task_id":"task_285_enterprise_industry_analysis_medium_medium100","category":"enterprise_industry_analysis","score":0.0,"tokens":678870,"cost_usd":0.113886,"elapsed_time":385.43,"gpr":0.0,"tpe":0.0,"ee":0.1081},{"task_id":"task_286_enterprise_industry_analysis_medium_medium101","category":"enterprise_industry_analysis","score":1.0,"tokens":278326,"cost_usd":0.055399,"elapsed_time":179.18,"gpr":1.0,"tpe":0.4594,"ee":0.3077},{"task_id":"task_287_enterprise_industry_analysis_medium_medium102","category":"enterprise_industry_analysis","score":0.0,"tokens":993725,"cost_usd":0.199085,"elapsed_time":192.53,"gpr":0.5,"tpe":0.2556,"ee":0.1905},{"task_id":"task_288_enterprise_industry_analysis_medium_medium103","category":"enterprise_industry_analysis","score":0.0,"tokens":267501,"cost_usd":0.055307,"elapsed_time":120.97,"gpr":0.0,"tpe":0.0,"ee":0.3077},{"task_id":"task_289_enterprise_industry_analysis_medium_medium104","category":"enterprise_industry_analysis","score":0.0,"tokens":722790,"cost_usd":0.153079,"elapsed_time":253.91,"gpr":0.25,"tpe":0.0798,"ee":0.1333},{"task_id":"task_290_enterprise_industry_analysis_medium_medium105","category":"enterprise_industry_analysis","score":0.0,"tokens":1483752,"cost_usd":0.271073,"elapsed_time":222.7,"gpr":0.25,"tpe":0.1668,"ee":0.1739},{"task_id":"task_291_enterprise_industry_analysis_medium_medium106","category":"enterprise_industry_analysis","score":0.0,"tokens":202225,"cost_usd":0.037689,"elapsed_time":111.43,"gpr":0.25,"tpe":0.5905,"ee":0.4},{"task_id":"task_292_enterprise_industry_analysis_medium_medium107","category":"enterprise_industry_analysis","score":0.0,"tokens":237482,"cost_usd":0.095838,"elapsed_time":97.75,"gpr":0.25,"tpe":0.729,"ee":0.4},{"task_id":"task_293_enterprise_industry_analysis_medium_medium108","category":"enterprise_industry_analysis","score":0.0,"tokens":691934,"cost_usd":0.150224,"elapsed_time":202.64,"gpr":0.0,"tpe":0.0,"ee":0.1739},{"task_id":"task_294_enterprise_industry_analysis_medium_medium109","category":"enterprise_industry_analysis","score":0.0,"tokens":805431,"cost_usd":0.173782,"elapsed_time":285.78,"gpr":0.25,"tpe":0.0798,"ee":0.1429},{"task_id":"task_295_enterprise_industry_analysis_medium_medium110","category":"enterprise_industry_analysis","score":0.0,"tokens":370000,"cost_usd":0.097608,"elapsed_time":119.07,"gpr":0.25,"tpe":0.4783,"ee":0.25},{"task_id":"task_296_enterprise_industry_analysis_medium_medium111","category":"enterprise_industry_analysis","score":0.0,"tokens":187941,"cost_usd":0.03065,"elapsed_time":133.77,"gpr":0.25,"tpe":0.5905,"ee":0.3333},{"task_id":"task_297_enterprise_industry_policy_analysis_easy_easy067","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":47033,"cost_usd":0.008833,"elapsed_time":41.18,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_298_enterprise_industry_policy_analysis_easy_easy068","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":430877,"cost_usd":0.098617,"elapsed_time":119.56,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_299_enterprise_industry_policy_analysis_easy_easy069","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":113112,"cost_usd":0.021384,"elapsed_time":99.97,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_300_enterprise_industry_policy_analysis_easy_easy070","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":94395,"cost_usd":0.01576,"elapsed_time":87.7,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_301_enterprise_industry_policy_analysis_easy_easy071","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":146108,"cost_usd":0.034564,"elapsed_time":84.46,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_302_enterprise_industry_policy_analysis_easy_easy072","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":96633,"cost_usd":0.015496,"elapsed_time":56.11,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_303_enterprise_industry_policy_analysis_easy_easy073","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":192577,"cost_usd":0.033342,"elapsed_time":144.94,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_304_enterprise_industry_policy_analysis_easy_easy074","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":422811,"cost_usd":0.072469,"elapsed_time":116.6,"gpr":0.0,"tpe":0.0,"ee":0.2143},{"task_id":"task_305_enterprise_industry_policy_analysis_easy_easy075","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":112404,"cost_usd":0.024557,"elapsed_time":63.09,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_306_enterprise_industry_policy_analysis_easy_easy076","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":612571,"cost_usd":0.171144,"elapsed_time":103.81,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_307_enterprise_industry_policy_analysis_medium_medium001","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":169973,"cost_usd":0.085001,"elapsed_time":86.94,"gpr":0.2,"tpe":1.0,"ee":0.8333},{"task_id":"task_308_enterprise_industry_policy_analysis_medium_medium002","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":174622,"cost_usd":0.08667,"elapsed_time":48.83,"gpr":1.0,"tpe":1.0,"ee":1.25},{"task_id":"task_309_enterprise_industry_policy_analysis_medium_medium003","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":104014,"cost_usd":0.028232,"elapsed_time":102.45,"gpr":0.8,"tpe":1.0,"ee":1.0},{"task_id":"task_310_enterprise_industry_policy_analysis_medium_medium004","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":209738,"cost_usd":0.042273,"elapsed_time":135.44,"gpr":0.4,"tpe":0.4305,"ee":0.3571},{"task_id":"task_311_enterprise_industry_policy_analysis_medium_medium005","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":2596888,"cost_usd":0.384841,"elapsed_time":300.59,"gpr":1.0,"tpe":0.599,"ee":0.2083},{"task_id":"task_312_enterprise_industry_policy_analysis_medium_medium006","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":436552,"cost_usd":0.131052,"elapsed_time":140.15,"gpr":0.4,"tpe":0.7452,"ee":0.3571},{"task_id":"task_313_enterprise_industry_policy_analysis_medium_medium007","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":420990,"cost_usd":0.131942,"elapsed_time":156.44,"gpr":0.6,"tpe":0.9667,"ee":0.4167},{"task_id":"task_314_enterprise_industry_policy_analysis_medium_medium008","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":305923,"cost_usd":0.055902,"elapsed_time":166.89,"gpr":1.0,"tpe":0.302,"ee":0.2941},{"task_id":"task_315_enterprise_industry_policy_analysis_medium_medium009","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":938913,"cost_usd":0.311906,"elapsed_time":153.4,"gpr":0.4,"tpe":1.0,"ee":0.3846},{"task_id":"task_316_enterprise_industry_policy_analysis_medium_medium010","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":995675,"cost_usd":0.214881,"elapsed_time":145.86,"gpr":0.0,"tpe":0.0,"ee":0.2941},{"task_id":"task_317_enterprise_industry_policy_analysis_medium_medium011","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":139192,"cost_usd":0.065752,"elapsed_time":83.32,"gpr":0.0,"tpe":0.0,"ee":0.8333},{"task_id":"task_318_enterprise_industry_policy_analysis_medium_medium012","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":737943,"cost_usd":0.157434,"elapsed_time":122.52,"gpr":0.6,"tpe":0.81,"ee":0.625},{"task_id":"task_319_enterprise_industry_policy_analysis_medium_medium013","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":169484,"cost_usd":0.083935,"elapsed_time":83.27,"gpr":0.2,"tpe":1.0,"ee":0.8333},{"task_id":"task_320_enterprise_industry_policy_analysis_medium_medium014","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1868616,"cost_usd":0.418046,"elapsed_time":205.51,"gpr":1.0,"tpe":0.5781,"ee":0.2778},{"task_id":"task_321_enterprise_industry_policy_analysis_medium_medium015","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":614835,"cost_usd":0.119458,"elapsed_time":293.89,"gpr":1.0,"tpe":0.4643,"ee":0.1515},{"task_id":"task_322_enterprise_industry_policy_analysis_medium_medium016","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":384663,"cost_usd":0.134915,"elapsed_time":138.18,"gpr":1.0,"tpe":0.7937,"ee":0.5556},{"task_id":"task_323_enterprise_industry_policy_analysis_medium_medium017","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":548142,"cost_usd":0.185081,"elapsed_time":153.72,"gpr":0.8,"tpe":0.914,"ee":0.4545},{"task_id":"task_324_enterprise_industry_policy_analysis_medium_medium018","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":292909,"cost_usd":0.051376,"elapsed_time":139.7,"gpr":1.0,"tpe":0.8015,"ee":0.3571},{"task_id":"task_325_enterprise_industry_policy_analysis_medium_medium019","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":188025,"cost_usd":0.073166,"elapsed_time":114.78,"gpr":0.0,"tpe":0.0,"ee":0.5556},{"task_id":"task_326_enterprise_industry_policy_analysis_medium_medium020","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":33530,"cost_usd":0.00636,"elapsed_time":89.81,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_327_enterprise_industry_policy_analysis_medium_medium021","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":10933,"cost_usd":0.002015,"elapsed_time":84.82,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_328_enterprise_industry_policy_analysis_medium_medium022","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":0,"cost_usd":0.0,"elapsed_time":65.65,"gpr":0.0,"tpe":0.0,"ee":1.25},{"task_id":"task_329_enterprise_industry_policy_analysis_medium_medium023","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":322144,"cost_usd":0.104944,"elapsed_time":186.38,"gpr":0.2,"tpe":0.4305,"ee":0.3846},{"task_id":"task_330_enterprise_industry_policy_analysis_medium_medium024","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":0,"cost_usd":0.0,"elapsed_time":30.43,"gpr":0.0,"tpe":0.0,"ee":1.25},{"task_id":"task_331_enterprise_industry_policy_analysis_medium_medium025","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":329205,"cost_usd":0.113208,"elapsed_time":181.6,"gpr":1.0,"tpe":0.9,"ee":0.5},{"task_id":"task_332_enterprise_industry_policy_analysis_medium_medium026","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":417544,"cost_usd":0.113294,"elapsed_time":157.7,"gpr":1.0,"tpe":0.7584,"ee":0.4545},{"task_id":"task_333_enterprise_industry_policy_analysis_medium_medium027","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1138445,"cost_usd":0.199734,"elapsed_time":372.21,"gpr":1.0,"tpe":0.2973,"ee":0.1667},{"task_id":"task_334_enterprise_industry_policy_analysis_medium_medium028","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":470683,"cost_usd":0.141983,"elapsed_time":147.07,"gpr":0.6,"tpe":0.8261,"ee":0.4167},{"task_id":"task_335_enterprise_industry_policy_analysis_medium_medium029","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":634725,"cost_usd":0.172631,"elapsed_time":115.89,"gpr":1.0,"tpe":0.6853,"ee":0.5556},{"task_id":"task_336_enterprise_industry_policy_analysis_medium_medium030","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":70253,"cost_usd":0.013032,"elapsed_time":121.49,"gpr":0.4,"tpe":0.855,"ee":0.7143},{"task_id":"task_337_enterprise_industry_policy_analysis_medium_medium031","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":666252,"cost_usd":0.218671,"elapsed_time":137.12,"gpr":1.0,"tpe":0.7128,"ee":0.5},{"task_id":"task_338_enterprise_industry_policy_analysis_medium_medium032","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":459046,"cost_usd":0.079205,"elapsed_time":305.54,"gpr":1.0,"tpe":0.3065,"ee":0.1538},{"task_id":"task_339_enterprise_industry_policy_analysis_medium_medium033","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":956344,"cost_usd":0.162148,"elapsed_time":419.14,"gpr":0.75,"tpe":0.1161,"ee":0.0833},{"task_id":"task_340_enterprise_industry_policy_analysis_medium_medium034","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":316760,"cost_usd":0.118611,"elapsed_time":114.4,"gpr":0.5,"tpe":1.0,"ee":0.5},{"task_id":"task_341_enterprise_industry_policy_analysis_medium_medium035","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":464072,"cost_usd":0.139096,"elapsed_time":166.16,"gpr":0.75,"tpe":0.81,"ee":0.4444},{"task_id":"task_342_enterprise_industry_policy_analysis_medium_medium036","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":156987,"cost_usd":0.043161,"elapsed_time":140.57,"gpr":0.0,"tpe":0.0,"ee":0.3636},{"task_id":"task_343_enterprise_industry_policy_analysis_medium_medium037","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1415199,"cost_usd":0.310754,"elapsed_time":273.92,"gpr":0.75,"tpe":0.6145,"ee":0.3077},{"task_id":"task_344_enterprise_industry_policy_analysis_medium_medium038","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":343450,"cost_usd":0.107907,"elapsed_time":155.25,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_345_enterprise_industry_policy_analysis_medium_medium039","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":174865,"cost_usd":0.043353,"elapsed_time":123.44,"gpr":0.0,"tpe":0.0,"ee":0.4444},{"task_id":"task_346_enterprise_industry_policy_analysis_medium_medium040","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":407415,"cost_usd":0.101341,"elapsed_time":162.28,"gpr":0.5,"tpe":0.5905,"ee":0.25},{"task_id":"task_347_enterprise_industry_policy_analysis_medium_medium041","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":33487,"cost_usd":0.006101,"elapsed_time":63.06,"gpr":0.0,"tpe":0.0,"ee":0.8},{"task_id":"task_348_enterprise_industry_policy_analysis_medium_medium042","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":460989,"cost_usd":0.112521,"elapsed_time":105.53,"gpr":0.25,"tpe":0.9,"ee":0.4444},{"task_id":"task_349_enterprise_industry_policy_analysis_medium_medium043","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1135527,"cost_usd":0.379352,"elapsed_time":322.67,"gpr":1.0,"tpe":0.5303,"ee":0.2857},{"task_id":"task_350_enterprise_industry_policy_analysis_medium_medium044","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":461311,"cost_usd":0.11651,"elapsed_time":140.24,"gpr":0.75,"tpe":0.9033,"ee":0.6667},{"task_id":"task_351_enterprise_industry_policy_analysis_medium_medium045","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":469329,"cost_usd":0.162273,"elapsed_time":129.58,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_352_enterprise_industry_policy_analysis_medium_medium046","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":360665,"cost_usd":0.116799,"elapsed_time":156.87,"gpr":0.0,"tpe":0.0,"ee":0.4444},{"task_id":"task_353_enterprise_industry_policy_analysis_medium_medium047","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":433177,"cost_usd":0.095023,"elapsed_time":138.77,"gpr":1.0,"tpe":0.8598,"ee":0.5714},{"task_id":"task_354_enterprise_industry_policy_analysis_medium_medium048","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":484458,"cost_usd":0.176216,"elapsed_time":174.56,"gpr":0.5,"tpe":0.6561,"ee":0.3636},{"task_id":"task_355_enterprise_industry_policy_analysis_medium_medium049","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":73592,"cost_usd":0.025219,"elapsed_time":125.25,"gpr":0.0,"tpe":0.0,"ee":0.8},{"task_id":"task_356_enterprise_industry_policy_analysis_medium_medium050","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1004230,"cost_usd":0.197214,"elapsed_time":182.06,"gpr":0.5,"tpe":0.3312,"ee":0.2222},{"task_id":"task_357_enterprise_industry_policy_analysis_medium_medium051","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":224961,"cost_usd":0.07469,"elapsed_time":69.88,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_358_enterprise_industry_policy_analysis_medium_medium052","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2527474,"cost_usd":0.454326,"elapsed_time":543.73,"gpr":0.5,"tpe":0.043,"ee":0.087},{"task_id":"task_359_enterprise_industry_policy_analysis_medium_medium053","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":146268,"cost_usd":0.029881,"elapsed_time":158.06,"gpr":0.75,"tpe":0.5905,"ee":0.4},{"task_id":"task_360_enterprise_industry_policy_analysis_medium_medium054","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":365997,"cost_usd":0.102727,"elapsed_time":101.98,"gpr":0.5,"tpe":0.9,"ee":0.6667},{"task_id":"task_361_enterprise_industry_policy_analysis_medium_medium055","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":169782,"cost_usd":0.084422,"elapsed_time":102.94,"gpr":0.0,"tpe":0.0,"ee":0.6667},{"task_id":"task_362_enterprise_industry_policy_analysis_medium_medium056","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1743101,"cost_usd":0.325169,"elapsed_time":273.05,"gpr":0.0,"tpe":0.0,"ee":0.2222},{"task_id":"task_363_enterprise_industry_policy_analysis_medium_medium057","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":202760,"cost_usd":0.059598,"elapsed_time":123.45,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_364_enterprise_industry_policy_analysis_medium_medium058","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":782166,"cost_usd":0.240585,"elapsed_time":163.76,"gpr":0.5,"tpe":0.4305,"ee":0.3333},{"task_id":"task_365_enterprise_industry_policy_analysis_medium_medium059","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":311158,"cost_usd":0.059181,"elapsed_time":195.66,"gpr":0.5,"tpe":0.7452,"ee":0.2105},{"task_id":"task_366_enterprise_industry_policy_analysis_medium_medium060","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":390608,"cost_usd":0.135153,"elapsed_time":138.53,"gpr":0.25,"tpe":0.6561,"ee":0.4444},{"task_id":"task_367_enterprise_industry_policy_analysis_medium_medium061","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1283449,"cost_usd":0.262985,"elapsed_time":148.48,"gpr":0.0,"tpe":0.0,"ee":0.3077},{"task_id":"task_368_enterprise_industry_policy_analysis_medium_medium062","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":483785,"cost_usd":0.125288,"elapsed_time":181.75,"gpr":0.75,"tpe":0.9667,"ee":0.2857},{"task_id":"task_369_enterprise_industry_policy_analysis_medium_medium063","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1468193,"cost_usd":0.227223,"elapsed_time":435.9,"gpr":0.25,"tpe":0.0424,"ee":0.069},{"task_id":"task_370_enterprise_industry_policy_analysis_medium_medium064","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":189894,"cost_usd":0.069418,"elapsed_time":100.96,"gpr":0.25,"tpe":0.9,"ee":0.5},{"task_id":"task_371_enterprise_industry_policy_analysis_medium_medium065","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":461357,"cost_usd":0.194488,"elapsed_time":139.78,"gpr":0.0,"tpe":0.0,"ee":0.3636},{"task_id":"task_372_enterprise_industry_policy_analysis_medium_medium066","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":544080,"cost_usd":0.118046,"elapsed_time":84.83,"gpr":0.0,"tpe":0.0,"ee":0.5714},{"task_id":"task_373_hypothesis_verification_hard_hard001","category":"hypothesis_verification","score":0.0,"tokens":785526,"cost_usd":0.154892,"elapsed_time":421.6,"gpr":0.7778,"tpe":0.2542,"ee":0.2414},{"task_id":"task_374_hypothesis_verification_hard_hard002","category":"hypothesis_verification","score":1.0,"tokens":340729,"cost_usd":0.073828,"elapsed_time":213.36,"gpr":1.0,"tpe":0.6753,"ee":0.2941},{"task_id":"task_375_hypothesis_verification_hard_hard003","category":"hypothesis_verification","score":0.0,"tokens":623469,"cost_usd":0.134424,"elapsed_time":389.89,"gpr":0.0,"tpe":0.0,"ee":0.2963},{"task_id":"task_376_hypothesis_verification_hard_hard004","category":"hypothesis_verification","score":0.0,"tokens":2095281,"cost_usd":0.441587,"elapsed_time":350.76,"gpr":0.1,"tpe":0.81,"ee":0.3684},{"task_id":"task_377_hypothesis_verification_hard_hard005","category":"hypothesis_verification","score":0.0,"tokens":274245,"cost_usd":0.116567,"elapsed_time":142.12,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_378_hypothesis_verification_hard_hard006","category":"hypothesis_verification","score":0.0,"tokens":669669,"cost_usd":0.194372,"elapsed_time":163.44,"gpr":0.0,"tpe":0.0,"ee":0.7},{"task_id":"task_379_hypothesis_verification_hard_hard007","category":"hypothesis_verification","score":0.0,"tokens":1616568,"cost_usd":0.54507,"elapsed_time":1153.7,"gpr":0.1875,"tpe":0.81,"ee":0.3913},{"task_id":"task_380_hypothesis_verification_hard_hard008","category":"hypothesis_verification","score":0.0,"tokens":395079,"cost_usd":0.152013,"elapsed_time":110.56,"gpr":0.0,"tpe":0.0,"ee":0.875},{"task_id":"task_381_hypothesis_verification_hard_hard009","category":"hypothesis_verification","score":0.0,"tokens":1064236,"cost_usd":0.208913,"elapsed_time":411.71,"gpr":0.0,"tpe":0.0,"ee":0.2143},{"task_id":"task_382_hypothesis_verification_hard_hard010","category":"hypothesis_verification","score":0.0,"tokens":1562689,"cost_usd":0.293392,"elapsed_time":587.22,"gpr":0.0667,"tpe":1.0,"ee":0.2258},{"task_id":"task_383_hypothesis_verification_hard_hard011","category":"hypothesis_verification","score":0.0,"tokens":1023327,"cost_usd":0.195049,"elapsed_time":591.28,"gpr":0.1538,"tpe":0.81,"ee":0.2121},{"task_id":"task_384_hypothesis_verification_hard_hard012","category":"hypothesis_verification","score":0.0,"tokens":2597943,"cost_usd":0.47299,"elapsed_time":523.76,"gpr":0.0,"tpe":0.0,"ee":0.2353},{"task_id":"task_385_hypothesis_verification_hard_hard013","category":"hypothesis_verification","score":0.0,"tokens":2030759,"cost_usd":0.366166,"elapsed_time":672.5,"gpr":0.0,"tpe":0.0,"ee":0.1944},{"task_id":"task_386_hypothesis_verification_hard_hard014","category":"hypothesis_verification","score":0.0,"tokens":1642703,"cost_usd":0.335798,"elapsed_time":655.48,"gpr":0.25,"tpe":0.3786,"ee":0.1935},{"task_id":"task_387_hypothesis_verification_hard_hard015","category":"hypothesis_verification","score":0.0,"tokens":795130,"cost_usd":0.142472,"elapsed_time":252.39,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_388_hypothesis_verification_medium_medium001","category":"hypothesis_verification","score":0.0,"tokens":1900308,"cost_usd":0.359358,"elapsed_time":419.31,"gpr":0.0,"tpe":0.0,"ee":0.2258},{"task_id":"task_389_hypothesis_verification_medium_medium002","category":"hypothesis_verification","score":0.0,"tokens":981574,"cost_usd":0.223632,"elapsed_time":568.94,"gpr":0.0,"tpe":0.0,"ee":0.28},{"task_id":"task_390_hypothesis_verification_medium_medium003","category":"hypothesis_verification","score":0.0,"tokens":4579669,"cost_usd":0.86028,"elapsed_time":785.44,"gpr":0.0,"tpe":0.0,"ee":0.1277},{"task_id":"task_391_hypothesis_verification_medium_medium004","category":"hypothesis_verification","score":0.0,"tokens":256086,"cost_usd":0.046184,"elapsed_time":156.6,"gpr":0.2,"tpe":1.0,"ee":0.5},{"task_id":"task_392_hypothesis_verification_medium_medium005","category":"hypothesis_verification","score":0.0,"tokens":253643,"cost_usd":0.050831,"elapsed_time":186.63,"gpr":0.4,"tpe":0.4329,"ee":0.4375},{"task_id":"task_393_hypothesis_verification_medium_medium006","category":"hypothesis_verification","score":0.0,"tokens":1442769,"cost_usd":0.282259,"elapsed_time":344.11,"gpr":0.4,"tpe":0.2339,"ee":0.1765},{"task_id":"task_394_hypothesis_verification_medium_medium007","category":"hypothesis_verification","score":0.0,"tokens":3235381,"cost_usd":0.733089,"elapsed_time":663.41,"gpr":0.1111,"tpe":0.2059,"ee":0.1522},{"task_id":"task_395_hypothesis_verification_medium_medium008","category":"hypothesis_verification","score":0.0,"tokens":769211,"cost_usd":0.133287,"elapsed_time":335.18,"gpr":0.0,"tpe":0.0,"ee":0.3077},{"task_id":"task_396_hypothesis_verification_medium_medium009","category":"hypothesis_verification","score":0.0,"tokens":909173,"cost_usd":0.199185,"elapsed_time":187.22,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_397_hypothesis_verification_medium_medium010","category":"hypothesis_verification","score":0.0,"tokens":200543,"cost_usd":0.074264,"elapsed_time":145.3,"gpr":0.0,"tpe":0.0,"ee":0.875},{"task_id":"task_398_hypothesis_verification_medium_medium011","category":"hypothesis_verification","score":1.0,"tokens":1996747,"cost_usd":0.345761,"elapsed_time":517.5,"gpr":0.8333,"tpe":0.211,"ee":0.2069},{"task_id":"task_399_hypothesis_verification_medium_medium012","category":"hypothesis_verification","score":1.0,"tokens":422938,"cost_usd":0.095842,"elapsed_time":265.46,"gpr":1.0,"tpe":0.6036,"ee":0.4615},{"task_id":"task_400_hypothesis_verification_medium_medium013","category":"hypothesis_verification","score":0.0,"tokens":1145316,"cost_usd":0.221575,"elapsed_time":623.68,"gpr":0.1667,"tpe":0.2059,"ee":0.2258},{"task_id":"task_401_hypothesis_verification_medium_medium014","category":"hypothesis_verification","score":0.0,"tokens":1108692,"cost_usd":0.23048,"elapsed_time":381.4,"gpr":0.4,"tpe":0.485,"ee":0.2963},{"task_id":"task_402_industry_planning_hard_hard001","category":"industry_planning","score":0.5,"tokens":2127190,"cost_usd":0.348451,"elapsed_time":416.34,"gpr":0.1111,"tpe":0.4783,"ee":0.2121},{"task_id":"task_403_industry_planning_hard_hard002","category":"industry_planning","score":1.0,"tokens":2817056,"cost_usd":0.474091,"elapsed_time":528.27,"gpr":0.0,"tpe":0.0,"ee":0.1944},{"task_id":"task_404_industry_planning_hard_hard003","category":"industry_planning","score":0.0,"tokens":22734,"cost_usd":0.004591,"elapsed_time":106.3,"gpr":0.0,"tpe":0.0,"ee":2.0},{"task_id":"task_405_industry_planning_hard_hard004","category":"industry_planning","score":0.0,"tokens":45949,"cost_usd":0.00987,"elapsed_time":95.49,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_406_industry_planning_hard_hard005","category":"industry_planning","score":0.0,"tokens":690723,"cost_usd":0.171367,"elapsed_time":277.77,"gpr":0.3333,"tpe":0.6876,"ee":0.5},{"task_id":"task_407_industry_planning_hard_hard006","category":"industry_planning","score":0.0,"tokens":238520,"cost_usd":0.062818,"elapsed_time":127.73,"gpr":0.0833,"tpe":1.0,"ee":0.7},{"task_id":"task_408_industry_planning_hard_hard007","category":"industry_planning","score":0.0,"tokens":118730,"cost_usd":0.029667,"elapsed_time":119.03,"gpr":0.4,"tpe":0.7421,"ee":0.6},{"task_id":"task_409_industry_planning_hard_hard008","category":"industry_planning","score":0.0,"tokens":251415,"cost_usd":0.059326,"elapsed_time":148.22,"gpr":0.625,"tpe":0.848,"ee":0.5455},{"task_id":"task_410_industry_planning_hard_hard009","category":"industry_planning","score":0.0,"tokens":124592,"cost_usd":0.027595,"elapsed_time":130.75,"gpr":0.0,"tpe":0.0,"ee":0.6667},{"task_id":"task_411_industry_planning_hard_hard010","category":"industry_planning","score":0.0,"tokens":1649700,"cost_usd":0.440137,"elapsed_time":344.91,"gpr":0.3333,"tpe":0.2398,"ee":0.2609},{"task_id":"task_412_industry_planning_hard_hard011","category":"industry_planning","score":0.0,"tokens":342662,"cost_usd":0.116166,"elapsed_time":148.04,"gpr":0.1667,"tpe":0.5314,"ee":0.5},{"task_id":"task_413_industry_planning_hard_hard012","category":"industry_planning","score":1.0,"tokens":1685396,"cost_usd":0.409395,"elapsed_time":382.64,"gpr":1.0,"tpe":0.2647,"ee":0.2609},{"task_id":"task_414_industry_planning_hard_hard013","category":"industry_planning","score":0.0,"tokens":258230,"cost_usd":0.056075,"elapsed_time":124.7,"gpr":0.0,"tpe":0.0,"ee":0.7},{"task_id":"task_415_industry_planning_hard_hard014","category":"industry_planning","score":0.0,"tokens":3670511,"cost_usd":1.085573,"elapsed_time":494.06,"gpr":0.0,"tpe":0.0,"ee":0.1087},{"task_id":"task_416_industry_planning_medium_medium001","category":"industry_planning","score":1.0,"tokens":1172174,"cost_usd":0.225091,"elapsed_time":521.83,"gpr":1.0,"tpe":0.0769,"ee":0.2222},{"task_id":"task_417_industry_planning_medium_medium002","category":"industry_planning","score":0.0,"tokens":2739575,"cost_usd":0.520058,"elapsed_time":496.8,"gpr":0.0,"tpe":0.0,"ee":0.3077},{"task_id":"task_418_industry_planning_medium_medium003","category":"industry_planning","score":0.0,"tokens":736272,"cost_usd":0.148062,"elapsed_time":354.42,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_419_industry_planning_medium_medium004","category":"industry_planning","score":1.0,"tokens":3828027,"cost_usd":0.6621,"elapsed_time":624.33,"gpr":1.0,"tpe":0.4315,"ee":0.1795},{"task_id":"task_420_industry_planning_medium_medium005","category":"industry_planning","score":0.0,"tokens":57051,"cost_usd":0.010407,"elapsed_time":87.75,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_421_industry_planning_medium_medium006","category":"industry_planning","score":0.0,"tokens":319184,"cost_usd":0.131763,"elapsed_time":139.69,"gpr":0.0,"tpe":0.0,"ee":0.8},{"task_id":"task_422_industry_planning_medium_medium007","category":"industry_planning","score":0.0,"tokens":1695156,"cost_usd":0.256178,"elapsed_time":350.57,"gpr":0.0,"tpe":0.0,"ee":0.2857},{"task_id":"task_423_industry_planning_medium_medium008","category":"industry_planning","score":1.0,"tokens":1978011,"cost_usd":0.313981,"elapsed_time":805.17,"gpr":1.0,"tpe":0.0278,"ee":0.1905},{"task_id":"task_424_industry_planning_medium_medium009","category":"industry_planning","score":0.0,"tokens":2561142,"cost_usd":0.607526,"elapsed_time":545.77,"gpr":0.1,"tpe":0.1351,"ee":0.2581},{"task_id":"task_425_industry_planning_medium_medium010","category":"industry_planning","score":1.0,"tokens":749162,"cost_usd":0.160493,"elapsed_time":390.1,"gpr":1.0,"tpe":0.3138,"ee":0.381},{"task_id":"task_426_industry_planning_medium_medium011","category":"industry_planning","score":1.0,"tokens":1390165,"cost_usd":0.264729,"elapsed_time":700.21,"gpr":1.0,"tpe":0.2499,"ee":0.2667},{"task_id":"task_427_industry_planning_medium_medium012","category":"industry_planning","score":0.0,"tokens":90929,"cost_usd":0.024757,"elapsed_time":128.45,"gpr":0.0,"tpe":0.0,"ee":1.1429},{"task_id":"task_428_industry_planning_medium_medium013","category":"industry_planning","score":0.0,"tokens":2206196,"cost_usd":0.410105,"elapsed_time":1014.62,"gpr":0.1667,"tpe":0.3874,"ee":0.2286},{"task_id":"task_429_industry_planning_medium_medium014","category":"industry_planning","score":0.0,"tokens":246385,"cost_usd":0.076101,"elapsed_time":136.03,"gpr":0.0,"tpe":0.0,"ee":1.2857},{"task_id":"task_430_international_comparison_hard_hard001","category":"international_comparison","score":0.5,"tokens":2339486,"cost_usd":0.746297,"elapsed_time":432.77,"gpr":0.3,"tpe":0.5416,"ee":0.2069},{"task_id":"task_431_international_comparison_hard_hard002","category":"international_comparison","score":0.0,"tokens":1336058,"cost_usd":0.593883,"elapsed_time":282.78,"gpr":0.0,"tpe":0.0,"ee":0.3158},{"task_id":"task_432_international_comparison_hard_hard003","category":"international_comparison","score":0.0,"tokens":710013,"cost_usd":0.18131,"elapsed_time":388.42,"gpr":0.3333,"tpe":0.5066,"ee":0.24},{"task_id":"task_433_international_comparison_hard_hard004","category":"international_comparison","score":0.0,"tokens":1655346,"cost_usd":0.694488,"elapsed_time":257.91,"gpr":0.0,"tpe":0.0,"ee":0.3529},{"task_id":"task_434_international_comparison_hard_hard005","category":"international_comparison","score":0.6667,"tokens":1754908,"cost_usd":0.358814,"elapsed_time":449.07,"gpr":0.9091,"tpe":0.0825,"ee":0.1875},{"task_id":"task_435_international_comparison_hard_hard006","category":"international_comparison","score":0.25,"tokens":1906737,"cost_usd":0.681537,"elapsed_time":278.74,"gpr":0.1667,"tpe":0.3487,"ee":0.4118},{"task_id":"task_436_international_comparison_hard_hard007","category":"international_comparison","score":0.25,"tokens":1880544,"cost_usd":0.688395,"elapsed_time":244.52,"gpr":0.3636,"tpe":0.4783,"ee":0.5},{"task_id":"task_437_international_comparison_hard_hard008","category":"international_comparison","score":0.5,"tokens":506196,"cost_usd":0.159823,"elapsed_time":243.08,"gpr":0.1667,"tpe":0.3874,"ee":0.375},{"task_id":"task_438_international_comparison_hard_hard009","category":"international_comparison","score":0.0,"tokens":1764521,"cost_usd":0.96457,"elapsed_time":497.64,"gpr":0,"tpe":0.0,"ee":0},{"task_id":"task_439_international_comparison_hard_hard010","category":"international_comparison","score":0.25,"tokens":2069611,"cost_usd":0.87441,"elapsed_time":270.83,"gpr":0.3333,"tpe":0.6144,"ee":0.3},{"task_id":"task_440_international_comparison_hard_hard011","category":"international_comparison","score":0.25,"tokens":1385321,"cost_usd":0.560239,"elapsed_time":192.94,"gpr":0.6,"tpe":0.6522,"ee":0.4615},{"task_id":"task_441_international_comparison_hard_hard012","category":"international_comparison","score":0.5,"tokens":899548,"cost_usd":0.267957,"elapsed_time":239.47,"gpr":0.4286,"tpe":0.5916,"ee":0.4},{"task_id":"task_442_international_comparison_hard_hard013","category":"international_comparison","score":0.25,"tokens":1340418,"cost_usd":0.511307,"elapsed_time":198.78,"gpr":0.6,"tpe":0.6876,"ee":0.5},{"task_id":"task_443_international_comparison_hard_hard014","category":"international_comparison","score":0.5,"tokens":1019983,"cost_usd":0.284492,"elapsed_time":258.06,"gpr":0.3333,"tpe":0.3138,"ee":0.3529},{"task_id":"task_444_international_comparison_medium_medium001","category":"international_comparison","score":1.0,"tokens":242808,"cost_usd":0.056598,"elapsed_time":132.62,"gpr":1.0,"tpe":0.5851,"ee":0.3636},{"task_id":"task_445_international_comparison_medium_medium002","category":"international_comparison","score":0.0,"tokens":1196382,"cost_usd":0.333731,"elapsed_time":299.2,"gpr":0.1667,"tpe":0.729,"ee":0.1667},{"task_id":"task_446_international_comparison_medium_medium003","category":"international_comparison","score":0.5,"tokens":1397563,"cost_usd":0.363361,"elapsed_time":210.49,"gpr":0.5,"tpe":0.4883,"ee":0.2},{"task_id":"task_447_international_comparison_medium_medium004","category":"international_comparison","score":0.0,"tokens":2022551,"cost_usd":0.8145,"elapsed_time":236.26,"gpr":0.6667,"tpe":0.3376,"ee":0.1818},{"task_id":"task_448_international_comparison_medium_medium005","category":"international_comparison","score":0.33,"tokens":2781231,"cost_usd":0.990294,"elapsed_time":291.91,"gpr":0.5,"tpe":0.1959,"ee":0.1818},{"task_id":"task_449_international_comparison_medium_medium006","category":"international_comparison","score":0.0,"tokens":1583366,"cost_usd":0.623841,"elapsed_time":300.62,"gpr":0.1667,"tpe":1.0,"ee":0.2174},{"task_id":"task_450_international_comparison_medium_medium007","category":"international_comparison","score":0.0,"tokens":3172177,"cost_usd":0.960792,"elapsed_time":367.73,"gpr":0.5,"tpe":0.0798,"ee":0.1429},{"task_id":"task_451_international_comparison_medium_medium008","category":"international_comparison","score":1.0,"tokens":1714186,"cost_usd":0.639715,"elapsed_time":253.08,"gpr":1.0,"tpe":0.3382,"ee":0.2105},{"task_id":"task_452_international_comparison_medium_medium009","category":"international_comparison","score":0.5,"tokens":1205303,"cost_usd":0.260395,"elapsed_time":240.38,"gpr":0.1429,"tpe":0.2824,"ee":0.2632},{"task_id":"task_453_international_comparison_medium_medium010","category":"international_comparison","score":0.0,"tokens":1945207,"cost_usd":0.838146,"elapsed_time":241.56,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_454_international_comparison_medium_medium011","category":"international_comparison","score":1.0,"tokens":414834,"cost_usd":0.113988,"elapsed_time":163.85,"gpr":1.0,"tpe":0.4457,"ee":0.2857},{"task_id":"task_455_international_comparison_medium_medium012","category":"international_comparison","score":1.0,"tokens":563465,"cost_usd":0.183623,"elapsed_time":114.76,"gpr":1.0,"tpe":0.6968,"ee":0.5556},{"task_id":"task_456_international_comparison_medium_medium013","category":"international_comparison","score":0.33,"tokens":319622,"cost_usd":0.125506,"elapsed_time":186.88,"gpr":0.5,"tpe":0.4783,"ee":0.3636},{"task_id":"task_457_international_comparison_medium_medium014","category":"international_comparison","score":0.5,"tokens":2458178,"cost_usd":0.521446,"elapsed_time":592.18,"gpr":0.375,"tpe":0.6696,"ee":0.0926},{"task_id":"task_458_international_comparison_medium_medium015","category":"international_comparison","score":0.0,"tokens":2182510,"cost_usd":0.603283,"elapsed_time":511.41,"gpr":0.1429,"tpe":0.729,"ee":0.0943},{"task_id":"task_459_international_comparison_medium_medium016","category":"international_comparison","score":0.5,"tokens":1161848,"cost_usd":0.59838,"elapsed_time":215.27,"gpr":0.1667,"tpe":1.0,"ee":0.4167},{"task_id":"task_460_international_comparison_medium_medium017","category":"international_comparison","score":0.25,"tokens":1286065,"cost_usd":0.26173,"elapsed_time":404.11,"gpr":0.5,"tpe":0.0424,"ee":0.119},{"task_id":"task_461_international_comparison_medium_medium018","category":"international_comparison","score":0.5,"tokens":2137383,"cost_usd":0.78705,"elapsed_time":321.24,"gpr":0.4286,"tpe":0.2428,"ee":0.1538},{"task_id":"task_462_international_comparison_medium_medium019","category":"international_comparison","score":1.0,"tokens":1300139,"cost_usd":0.45411,"elapsed_time":344.44,"gpr":1.0,"tpe":0.1351,"ee":0.1739},{"task_id":"task_463_international_comparison_medium_medium020","category":"international_comparison","score":0.0,"tokens":3543701,"cost_usd":1.379958,"elapsed_time":356.19,"gpr":0.0,"tpe":0.0,"ee":0.1724},{"task_id":"task_464_international_comparison_medium_medium021","category":"international_comparison","score":0.0,"tokens":3196266,"cost_usd":0.952128,"elapsed_time":327.96,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_465_international_comparison_medium_medium022","category":"international_comparison","score":0.0,"tokens":3723981,"cost_usd":1.087803,"elapsed_time":432.3,"gpr":0.4545,"tpe":0.4048,"ee":0.1351},{"task_id":"task_466_international_comparison_medium_medium023","category":"international_comparison","score":0.0,"tokens":1846667,"cost_usd":0.472929,"elapsed_time":312.11,"gpr":0.3333,"tpe":0.905,"ee":0.2273},{"task_id":"task_467_international_comparison_medium_medium024","category":"international_comparison","score":0.0,"tokens":1042526,"cost_usd":0.28112,"elapsed_time":301.36,"gpr":1.0,"tpe":0.3591,"ee":0.2381},{"task_id":"task_468_international_comparison_medium_medium025","category":"international_comparison","score":0.5,"tokens":1175766,"cost_usd":0.25718,"elapsed_time":306.87,"gpr":0.6667,"tpe":0.2623,"ee":0.1429},{"task_id":"task_469_risk_assessment_hard_hard001","category":"risk_assessment","score":0.0,"tokens":1029295,"cost_usd":0.206912,"elapsed_time":316.7,"gpr":0.0,"tpe":0.0,"ee":0.2121},{"task_id":"task_470_risk_assessment_hard_hard002","category":"risk_assessment","score":1.0,"tokens":529925,"cost_usd":0.121624,"elapsed_time":290.67,"gpr":1.0,"tpe":0.6167,"ee":0.4667},{"task_id":"task_471_risk_assessment_hard_hard003","category":"risk_assessment","score":0.0,"tokens":33708,"cost_usd":0.006443,"elapsed_time":73.51,"gpr":0.0909,"tpe":1.0,"ee":1.4},{"task_id":"task_472_risk_assessment_hard_hard004","category":"risk_assessment","score":0.0,"tokens":68737,"cost_usd":0.013096,"elapsed_time":139.82,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_473_risk_assessment_hard_hard005","category":"risk_assessment","score":0.0,"tokens":272956,"cost_usd":0.067604,"elapsed_time":301.86,"gpr":0.0,"tpe":0.0,"ee":0.6364},{"task_id":"task_474_risk_assessment_hard_hard006","category":"risk_assessment","score":0.0,"tokens":57693,"cost_usd":0.011352,"elapsed_time":171.42,"gpr":0.0,"tpe":0.0,"ee":1.1667},{"task_id":"task_475_risk_assessment_hard_hard007","category":"risk_assessment","score":0.0,"tokens":257536,"cost_usd":0.088043,"elapsed_time":108.22,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_476_risk_assessment_hard_hard008","category":"risk_assessment","score":0.0,"tokens":2315189,"cost_usd":0.382031,"elapsed_time":579.62,"gpr":0.5,"tpe":1.0,"ee":0.1842},{"task_id":"task_477_risk_assessment_hard_hard009","category":"risk_assessment","score":0.0,"tokens":822726,"cost_usd":0.172963,"elapsed_time":349.34,"gpr":0.75,"tpe":0.4305,"ee":0.3684},{"task_id":"task_478_risk_assessment_hard_hard010","category":"risk_assessment","score":0.0,"tokens":877546,"cost_usd":0.192231,"elapsed_time":1025.97,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_479_risk_assessment_hard_hard011","category":"risk_assessment","score":0.0,"tokens":33550,"cost_usd":0.006021,"elapsed_time":95.29,"gpr":0.0,"tpe":0.0,"ee":1.4},{"task_id":"task_480_risk_assessment_hard_hard012","category":"risk_assessment","score":0.0,"tokens":201946,"cost_usd":0.043746,"elapsed_time":163.14,"gpr":1.0,"tpe":0.6561,"ee":0.5833},{"task_id":"task_481_risk_assessment_hard_hard013","category":"risk_assessment","score":0.0,"tokens":248710,"cost_usd":0.051749,"elapsed_time":120.3,"gpr":0.0,"tpe":0.0,"ee":0.8182},{"task_id":"task_482_risk_assessment_medium_medium001","category":"risk_assessment","score":0.0,"tokens":503361,"cost_usd":0.124838,"elapsed_time":118.22,"gpr":0.3333,"tpe":0.9,"ee":0.75},{"task_id":"task_483_risk_assessment_medium_medium002","category":"risk_assessment","score":0.0,"tokens":428930,"cost_usd":0.11473,"elapsed_time":103.48,"gpr":0.0,"tpe":0.0,"ee":0.6667},{"task_id":"task_484_risk_assessment_medium_medium003","category":"risk_assessment","score":0.0,"tokens":107765,"cost_usd":0.018234,"elapsed_time":129.86,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_485_risk_assessment_medium_medium004","category":"risk_assessment","score":0.5,"tokens":582657,"cost_usd":0.229282,"elapsed_time":183.2,"gpr":0.0,"tpe":0.0,"ee":0.4167},{"task_id":"task_486_risk_assessment_medium_medium005","category":"risk_assessment","score":0.0,"tokens":3840648,"cost_usd":0.695268,"elapsed_time":1069.81,"gpr":0.125,"tpe":0.9,"ee":0.0962},{"task_id":"task_487_risk_assessment_medium_medium006","category":"risk_assessment","score":0.0,"tokens":1473617,"cost_usd":0.290987,"elapsed_time":653.67,"gpr":0.1818,"tpe":0.2828,"ee":0.1724},{"task_id":"task_488_risk_assessment_medium_medium007","category":"risk_assessment","score":0.0,"tokens":3129572,"cost_usd":0.50769,"elapsed_time":814.41,"gpr":0.1667,"tpe":1.0,"ee":0.1429},{"task_id":"task_489_risk_assessment_medium_medium008","category":"risk_assessment","score":0.0,"tokens":33634,"cost_usd":0.006843,"elapsed_time":106.13,"gpr":0.0,"tpe":0.0,"ee":1.2},{"task_id":"task_490_risk_assessment_medium_medium009","category":"risk_assessment","score":1.0,"tokens":2405968,"cost_usd":0.506804,"elapsed_time":788.81,"gpr":0.0,"tpe":0.0,"ee":0.1765},{"task_id":"task_491_risk_assessment_medium_medium010","category":"risk_assessment","score":0.0,"tokens":2003896,"cost_usd":0.375026,"elapsed_time":1059.57,"gpr":0.0,"tpe":0.0,"ee":0.1818},{"task_id":"task_492_risk_assessment_medium_medium011","category":"risk_assessment","score":0.0,"tokens":270813,"cost_usd":0.139569,"elapsed_time":109.02,"gpr":0.0,"tpe":0.0,"ee":0.8571}]},{"model":"kimi-k2.5","acc":0.2627,"ee":0.3301,"gpr":0.1918,"tpe":0.2512,"efficiency":{"total_tokens":465073576,"total_cost_usd":160.25,"total_requests":7883},"category_scores":{"comprehensive_decision":{"acc":0.0466,"ee":0.3452,"gpr":0.1218,"tpe":0.1549,"count":70},"enterprise_industry_analysis":{"acc":0.3451,"ee":0.3318,"gpr":0.2124,"tpe":0.2523,"count":226},"enterprise_industry_policy_analysis":{"acc":0.3684,"ee":0.3363,"gpr":0.3573,"tpe":0.3797,"count":76},"hypothesis_verification":{"acc":0.0345,"ee":0.25,"gpr":0.0775,"tpe":0.1278,"count":29},"industry_planning":{"acc":0.1786,"ee":0.3513,"gpr":0.0767,"tpe":0.2419,"count":28},"international_comparison":{"acc":0.2606,"ee":0.2705,"gpr":0.2421,"tpe":0.3389,"count":39},"risk_assessment":{"acc":0.1597,"ee":0.3371,"gpr":0.1003,"tpe":0.2987,"count":24}},"tasks":[{"task_id":"task_001_comprehensive_decision_easy_easy001","category":"comprehensive_decision","score":0.0,"tokens":534105,"cost_usd":0.212334,"elapsed_time":183.84,"gpr":0.0,"tpe":0.0,"ee":0.125},{"task_id":"task_002_comprehensive_decision_easy_easy002","category":"comprehensive_decision","score":0.0,"tokens":51039,"cost_usd":0.022291,"elapsed_time":64.07,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_003_comprehensive_decision_easy_easy003","category":"comprehensive_decision","score":0.0,"tokens":904830,"cost_usd":0.357544,"elapsed_time":310.66,"gpr":0.25,"tpe":0.9,"ee":0.4286},{"task_id":"task_004_comprehensive_decision_easy_easy004","category":"comprehensive_decision","score":0.0,"tokens":9867,"cost_usd":0.004341,"elapsed_time":29.76,"gpr":0.0,"tpe":0.0,"ee":2.0},{"task_id":"task_005_comprehensive_decision_easy_easy005","category":"comprehensive_decision","score":0.0,"tokens":1545878,"cost_usd":0.578313,"elapsed_time":463.76,"gpr":0.5,"tpe":0.4657,"ee":0.1875},{"task_id":"task_006_comprehensive_decision_easy_easy006","category":"comprehensive_decision","score":0.0,"tokens":98477,"cost_usd":0.04303,"elapsed_time":93.44,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_007_comprehensive_decision_hard_hard001","category":"comprehensive_decision","score":0.0,"tokens":3674435,"cost_usd":1.458315,"elapsed_time":1161.64,"gpr":0.0,"tpe":0.0,"ee":0.2917},{"task_id":"task_008_comprehensive_decision_hard_hard002","category":"comprehensive_decision","score":0.0,"tokens":10034,"cost_usd":0.004415,"elapsed_time":31.5,"gpr":0.0,"tpe":0.0,"ee":7.0},{"task_id":"task_009_comprehensive_decision_hard_hard003","category":"comprehensive_decision","score":0.0,"tokens":841607,"cost_usd":0.370792,"elapsed_time":251.91,"gpr":0.0,"tpe":0.0,"ee":1.1667},{"task_id":"task_010_comprehensive_decision_hard_hard004","category":"comprehensive_decision","score":0.0,"tokens":1236900,"cost_usd":0.470863,"elapsed_time":375.29,"gpr":0.0,"tpe":0.0,"ee":0.875},{"task_id":"task_011_comprehensive_decision_hard_hard005","category":"comprehensive_decision","score":0.0,"tokens":489920,"cost_usd":0.193568,"elapsed_time":232.62,"gpr":0.0,"tpe":0.0,"ee":0.6364},{"task_id":"task_012_comprehensive_decision_hard_hard006","category":"comprehensive_decision","score":0.0,"tokens":10005,"cost_usd":0.004402,"elapsed_time":55.57,"gpr":0.0,"tpe":0.0,"ee":7.0},{"task_id":"task_013_comprehensive_decision_hard_hard007","category":"comprehensive_decision","score":0.0,"tokens":73540,"cost_usd":0.025136,"elapsed_time":119.97,"gpr":0.1667,"tpe":1.0,"ee":1.4},{"task_id":"task_014_comprehensive_decision_hard_hard008","category":"comprehensive_decision","score":0.0,"tokens":214733,"cost_usd":0.087878,"elapsed_time":143.26,"gpr":0.0,"tpe":0.0,"ee":0.7},{"task_id":"task_015_comprehensive_decision_hard_hard009","category":"comprehensive_decision","score":0.0,"tokens":2528138,"cost_usd":0.815877,"elapsed_time":1207.59,"gpr":0.1429,"tpe":0.2059,"ee":0.2963},{"task_id":"task_016_comprehensive_decision_hard_hard010","category":"comprehensive_decision","score":0.0,"tokens":694412,"cost_usd":0.304066,"elapsed_time":255.9,"gpr":0.0,"tpe":0.0,"ee":1.4},{"task_id":"task_017_comprehensive_decision_hard_hard011","category":"comprehensive_decision","score":0.0,"tokens":402184,"cost_usd":0.171981,"elapsed_time":196.81,"gpr":0.0,"tpe":0.0,"ee":0.7778},{"task_id":"task_018_comprehensive_decision_hard_hard012","category":"comprehensive_decision","score":0.0,"tokens":990540,"cost_usd":0.343391,"elapsed_time":284.56,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_019_comprehensive_decision_hard_hard013","category":"comprehensive_decision","score":0.0,"tokens":10011,"cost_usd":0.004211,"elapsed_time":33.35,"gpr":0.0,"tpe":0.0,"ee":3.5},{"task_id":"task_020_comprehensive_decision_hard_hard014","category":"comprehensive_decision","score":0.0,"tokens":324346,"cost_usd":0.121616,"elapsed_time":361.05,"gpr":0.5714,"tpe":0.4342,"ee":0.4118},{"task_id":"task_021_comprehensive_decision_hard_hard015","category":"comprehensive_decision","score":0.0,"tokens":1474446,"cost_usd":0.563833,"elapsed_time":583.93,"gpr":0.8571,"tpe":0.95,"ee":0.8},{"task_id":"task_022_comprehensive_decision_hard_hard016","category":"comprehensive_decision","score":0.0,"tokens":790394,"cost_usd":0.345132,"elapsed_time":309.7,"gpr":0.125,"tpe":0.729,"ee":0.6667},{"task_id":"task_023_comprehensive_decision_hard_hard017","category":"comprehensive_decision","score":0.0,"tokens":144200,"cost_usd":0.05359,"elapsed_time":77.22,"gpr":0.0,"tpe":0.0,"ee":0.5455},{"task_id":"task_024_comprehensive_decision_hard_hard018","category":"comprehensive_decision","score":0.0,"tokens":1671218,"cost_usd":0.55965,"elapsed_time":592.31,"gpr":1.0,"tpe":0.3471,"ee":0.3889},{"task_id":"task_025_comprehensive_decision_hard_hard019","category":"comprehensive_decision","score":0.0,"tokens":19448,"cost_usd":0.008766,"elapsed_time":23.66,"gpr":0.0,"tpe":0.0,"ee":3.0},{"task_id":"task_026_comprehensive_decision_medium_medium001","category":"comprehensive_decision","score":0.0,"tokens":91939,"cost_usd":0.03831,"elapsed_time":101.33,"gpr":0.0,"tpe":0.0,"ee":0.625},{"task_id":"task_027_comprehensive_decision_medium_medium002","category":"comprehensive_decision","score":0.0,"tokens":1279271,"cost_usd":0.483132,"elapsed_time":364.5,"gpr":0.4286,"tpe":0.786,"ee":0.5556},{"task_id":"task_028_comprehensive_decision_medium_medium003","category":"comprehensive_decision","score":0.0,"tokens":29989,"cost_usd":0.013406,"elapsed_time":60.27,"gpr":0.0,"tpe":0.0,"ee":1.6667},{"task_id":"task_029_comprehensive_decision_medium_medium004","category":"comprehensive_decision","score":0.0,"tokens":43106,"cost_usd":0.015477,"elapsed_time":60.41,"gpr":0.0,"tpe":0.0,"ee":1.25},{"task_id":"task_030_comprehensive_decision_medium_medium005","category":"comprehensive_decision","score":0.0,"tokens":19222,"cost_usd":0.008715,"elapsed_time":18.63,"gpr":0.0,"tpe":0.0,"ee":2.5},{"task_id":"task_031_comprehensive_decision_medium_medium006","category":"comprehensive_decision","score":0.0,"tokens":71831,"cost_usd":0.026333,"elapsed_time":93.59,"gpr":0.0,"tpe":0.0,"ee":0.5714},{"task_id":"task_032_comprehensive_decision_medium_medium007","category":"comprehensive_decision","score":0.0,"tokens":56137,"cost_usd":0.019626,"elapsed_time":90.51,"gpr":0.0,"tpe":0.0,"ee":0.7143},{"task_id":"task_033_comprehensive_decision_medium_medium008","category":"comprehensive_decision","score":0.0,"tokens":314619,"cost_usd":0.133532,"elapsed_time":204.69,"gpr":0.3333,"tpe":0.81,"ee":0.625},{"task_id":"task_034_comprehensive_decision_medium_medium009","category":"comprehensive_decision","score":0.0,"tokens":32779,"cost_usd":0.014629,"elapsed_time":47.59,"gpr":0.0,"tpe":0.0,"ee":1.6667},{"task_id":"task_035_comprehensive_decision_medium_medium010","category":"comprehensive_decision","score":0.5,"tokens":136620,"cost_usd":0.045595,"elapsed_time":125.54,"gpr":0.25,"tpe":0.729,"ee":0.4545},{"task_id":"task_036_comprehensive_decision_medium_medium011","category":"comprehensive_decision","score":0.0,"tokens":91207,"cost_usd":0.036546,"elapsed_time":115.65,"gpr":0.0,"tpe":0.0,"ee":0.5556},{"task_id":"task_037_comprehensive_decision_medium_medium012","category":"comprehensive_decision","score":0.0,"tokens":269200,"cost_usd":0.103559,"elapsed_time":114.16,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_038_comprehensive_decision_medium_medium013","category":"comprehensive_decision","score":0.0,"tokens":40297,"cost_usd":0.014542,"elapsed_time":55.9,"gpr":0.0,"tpe":0.0,"ee":1.25},{"task_id":"task_039_comprehensive_decision_medium_medium014","category":"comprehensive_decision","score":0.0,"tokens":29200,"cost_usd":0.013043,"elapsed_time":24.44,"gpr":0.0,"tpe":0.0,"ee":1.6667},{"task_id":"task_040_comprehensive_decision_medium_medium015","category":"comprehensive_decision","score":0.0,"tokens":19534,"cost_usd":0.008679,"elapsed_time":44.66,"gpr":0.0,"tpe":0.0,"ee":2.5},{"task_id":"task_041_comprehensive_decision_medium_medium016","category":"comprehensive_decision","score":0.0,"tokens":466796,"cost_usd":0.17943,"elapsed_time":288.84,"gpr":0.0,"tpe":0.0,"ee":0.1852},{"task_id":"task_042_comprehensive_decision_medium_medium017","category":"comprehensive_decision","score":0.0,"tokens":94412,"cost_usd":0.041288,"elapsed_time":80.23,"gpr":0.0,"tpe":0.0,"ee":0.5556},{"task_id":"task_043_comprehensive_decision_medium_medium018","category":"comprehensive_decision","score":0.0,"tokens":355891,"cost_usd":0.139927,"elapsed_time":102.73,"gpr":0.0,"tpe":0.0,"ee":0.7143},{"task_id":"task_044_comprehensive_decision_medium_medium019","category":"comprehensive_decision","score":0.0,"tokens":44694,"cost_usd":0.020116,"elapsed_time":44.68,"gpr":0.0,"tpe":0.0,"ee":1.25},{"task_id":"task_045_comprehensive_decision_medium_medium020","category":"comprehensive_decision","score":0.0,"tokens":9870,"cost_usd":0.004343,"elapsed_time":33.12,"gpr":0.0,"tpe":0.0,"ee":5.0},{"task_id":"task_046_comprehensive_decision_medium_medium021","category":"comprehensive_decision","score":0.0,"tokens":63478,"cost_usd":0.026275,"elapsed_time":68.48,"gpr":0.0,"tpe":0.0,"ee":0.8333},{"task_id":"task_047_comprehensive_decision_medium_medium022","category":"comprehensive_decision","score":0.0,"tokens":580063,"cost_usd":0.202466,"elapsed_time":279.02,"gpr":0.0,"tpe":0.0,"ee":0.1562},{"task_id":"task_048_comprehensive_decision_medium_medium023","category":"comprehensive_decision","score":0.0,"tokens":1049513,"cost_usd":0.428779,"elapsed_time":359.75,"gpr":0.25,"tpe":0.5314,"ee":0.25},{"task_id":"task_049_comprehensive_decision_medium_medium024","category":"comprehensive_decision","score":0.0,"tokens":576285,"cost_usd":0.252653,"elapsed_time":349.76,"gpr":0.0,"tpe":0.0,"ee":0.7143},{"task_id":"task_050_comprehensive_decision_medium_medium025","category":"comprehensive_decision","score":0.0,"tokens":310572,"cost_usd":0.11585,"elapsed_time":169.52,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_051_comprehensive_decision_medium_medium026","category":"comprehensive_decision","score":0.0,"tokens":19562,"cost_usd":0.008692,"elapsed_time":26.03,"gpr":0.0,"tpe":0.0,"ee":2.5},{"task_id":"task_052_comprehensive_decision_medium_medium027","category":"comprehensive_decision","score":1.0,"tokens":2274289,"cost_usd":0.919135,"elapsed_time":723.68,"gpr":1.0,"tpe":0.3874,"ee":0.3571},{"task_id":"task_053_comprehensive_decision_medium_medium028","category":"comprehensive_decision","score":0.7647,"tokens":1619703,"cost_usd":0.55309,"elapsed_time":796.15,"gpr":0.5,"tpe":0.0343,"ee":0.1064},{"task_id":"task_054_comprehensive_decision_medium_medium029","category":"comprehensive_decision","score":0.0,"tokens":1654319,"cost_usd":0.585941,"elapsed_time":470.94,"gpr":0.5,"tpe":0.9,"ee":0.3846},{"task_id":"task_055_comprehensive_decision_medium_medium030","category":"comprehensive_decision","score":0.0,"tokens":711375,"cost_usd":0.246406,"elapsed_time":254.6,"gpr":0.8571,"tpe":0.368,"ee":0.2273},{"task_id":"task_056_comprehensive_decision_medium_medium031","category":"comprehensive_decision","score":0.0,"tokens":19634,"cost_usd":0.006869,"elapsed_time":48.08,"gpr":0.0,"tpe":0.0,"ee":2.0},{"task_id":"task_057_comprehensive_decision_medium_medium032","category":"comprehensive_decision","score":0.0,"tokens":3169899,"cost_usd":1.14849,"elapsed_time":874.5,"gpr":0.7143,"tpe":0.4219,"ee":0.25},{"task_id":"task_058_comprehensive_decision_medium_medium033","category":"comprehensive_decision","score":0.0,"tokens":102254,"cost_usd":0.038561,"elapsed_time":90.05,"gpr":0.5,"tpe":0.7695,"ee":0.5556},{"task_id":"task_059_comprehensive_decision_medium_medium034","category":"comprehensive_decision","score":0.0,"tokens":86405,"cost_usd":0.03841,"elapsed_time":72.96,"gpr":0.0,"tpe":0.0,"ee":1.25},{"task_id":"task_060_comprehensive_decision_medium_medium035","category":"comprehensive_decision","score":0.0,"tokens":29017,"cost_usd":0.011111,"elapsed_time":33.62,"gpr":0.0,"tpe":0.0,"ee":1.6667},{"task_id":"task_061_comprehensive_decision_medium_medium036","category":"comprehensive_decision","score":0.0,"tokens":19648,"cost_usd":0.008842,"elapsed_time":56.21,"gpr":0.0,"tpe":0.0,"ee":2.5},{"task_id":"task_062_comprehensive_decision_medium_medium037","category":"comprehensive_decision","score":0.0,"tokens":422203,"cost_usd":0.149333,"elapsed_time":153.91,"gpr":0.0,"tpe":0.0,"ee":0.3571},{"task_id":"task_063_comprehensive_decision_medium_medium038","category":"comprehensive_decision","score":0.0,"tokens":2993583,"cost_usd":1.112372,"elapsed_time":794.28,"gpr":0.0,"tpe":0.0,"ee":0.2105},{"task_id":"task_064_comprehensive_decision_medium_medium039","category":"comprehensive_decision","score":0.0,"tokens":3625330,"cost_usd":1.302776,"elapsed_time":1115.96,"gpr":0.0,"tpe":0.0,"ee":0.1818},{"task_id":"task_065_comprehensive_decision_medium_medium040","category":"comprehensive_decision","score":0.0,"tokens":123084,"cost_usd":0.052904,"elapsed_time":103.28,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_066_comprehensive_decision_medium_medium041","category":"comprehensive_decision","score":0.0,"tokens":601246,"cost_usd":0.206217,"elapsed_time":298.68,"gpr":0.3333,"tpe":0.1501,"ee":0.2174},{"task_id":"task_067_comprehensive_decision_medium_medium042","category":"comprehensive_decision","score":0.0,"tokens":940637,"cost_usd":0.332029,"elapsed_time":294.79,"gpr":0.0,"tpe":0.0,"ee":0.2632},{"task_id":"task_068_comprehensive_decision_medium_medium043","category":"comprehensive_decision","score":1.0,"tokens":304364,"cost_usd":0.11559,"elapsed_time":208.42,"gpr":1.0,"tpe":0.4305,"ee":0.3333},{"task_id":"task_069_comprehensive_decision_medium_medium044","category":"comprehensive_decision","score":0.0,"tokens":91507,"cost_usd":0.034968,"elapsed_time":207.35,"gpr":0.0,"tpe":0.0,"ee":0.5556},{"task_id":"task_070_comprehensive_decision_medium_medium045","category":"comprehensive_decision","score":0.0,"tokens":483638,"cost_usd":0.17485,"elapsed_time":202.13,"gpr":0.0,"tpe":0.0,"ee":0.2778},{"task_id":"task_071_enterprise_industry_analysis_easy_easy001","category":"enterprise_industry_analysis","score":0.0,"tokens":19154,"cost_usd":0.008624,"elapsed_time":25.73,"gpr":0.0,"tpe":0.0,"ee":1.5},{"task_id":"task_072_enterprise_industry_analysis_easy_easy002","category":"enterprise_industry_analysis","score":0.0,"tokens":129461,"cost_usd":0.046441,"elapsed_time":135.73,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_073_enterprise_industry_analysis_easy_easy003","category":"enterprise_industry_analysis","score":0.0,"tokens":39167,"cost_usd":0.0159,"elapsed_time":113.99,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_074_enterprise_industry_analysis_easy_easy004","category":"enterprise_industry_analysis","score":0.0,"tokens":997508,"cost_usd":0.343515,"elapsed_time":341.53,"gpr":0.0,"tpe":0.0,"ee":0.15},{"task_id":"task_075_enterprise_industry_analysis_easy_easy005","category":"enterprise_industry_analysis","score":0.0,"tokens":388691,"cost_usd":0.167577,"elapsed_time":308.09,"gpr":0.0,"tpe":0.0,"ee":0.75},{"task_id":"task_076_enterprise_industry_analysis_easy_easy006","category":"enterprise_industry_analysis","score":1.0,"tokens":71976,"cost_usd":0.032285,"elapsed_time":81.78,"gpr":1.0,"tpe":0.783,"ee":0.5},{"task_id":"task_077_enterprise_industry_analysis_easy_easy007","category":"enterprise_industry_analysis","score":0.0,"tokens":49521,"cost_usd":0.018099,"elapsed_time":85.33,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_078_enterprise_industry_analysis_easy_easy008","category":"enterprise_industry_analysis","score":0.0,"tokens":103631,"cost_usd":0.043756,"elapsed_time":79.44,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_079_enterprise_industry_analysis_easy_easy009","category":"enterprise_industry_analysis","score":0.0,"tokens":47136,"cost_usd":0.020933,"elapsed_time":50.34,"gpr":0.0,"tpe":0.0,"ee":0.75},{"task_id":"task_080_enterprise_industry_analysis_easy_easy010","category":"enterprise_industry_analysis","score":0.0,"tokens":224699,"cost_usd":0.097097,"elapsed_time":106.61,"gpr":0.3333,"tpe":0.9,"ee":0.5},{"task_id":"task_081_enterprise_industry_analysis_easy_easy011","category":"enterprise_industry_analysis","score":0.0,"tokens":88969,"cost_usd":0.033959,"elapsed_time":98.71,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_082_enterprise_industry_analysis_easy_easy012","category":"enterprise_industry_analysis","score":0.0,"tokens":98273,"cost_usd":0.03441,"elapsed_time":121.54,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_083_enterprise_industry_analysis_easy_easy013","category":"enterprise_industry_analysis","score":0.0,"tokens":95047,"cost_usd":0.037185,"elapsed_time":125.97,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_084_enterprise_industry_analysis_easy_easy014","category":"enterprise_industry_analysis","score":0.0,"tokens":29467,"cost_usd":0.013223,"elapsed_time":28.93,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_085_enterprise_industry_analysis_easy_easy015","category":"enterprise_industry_analysis","score":0.0,"tokens":29339,"cost_usd":0.011559,"elapsed_time":52.5,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_086_enterprise_industry_analysis_easy_easy016","category":"enterprise_industry_analysis","score":1.0,"tokens":524305,"cost_usd":0.186051,"elapsed_time":335.45,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_087_enterprise_industry_analysis_easy_easy017","category":"enterprise_industry_analysis","score":0.0,"tokens":138450,"cost_usd":0.057345,"elapsed_time":112.13,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_088_enterprise_industry_analysis_easy_easy018","category":"enterprise_industry_analysis","score":0.0,"tokens":282629,"cost_usd":0.105554,"elapsed_time":132.33,"gpr":0.0,"tpe":0.0,"ee":0.2143},{"task_id":"task_089_enterprise_industry_analysis_easy_easy019","category":"enterprise_industry_analysis","score":1.0,"tokens":580907,"cost_usd":0.187816,"elapsed_time":322.38,"gpr":1.0,"tpe":0.0523,"ee":0.0968},{"task_id":"task_090_enterprise_industry_analysis_easy_easy020","category":"enterprise_industry_analysis","score":0.0,"tokens":84641,"cost_usd":0.035469,"elapsed_time":70.99,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_091_enterprise_industry_analysis_easy_easy021","category":"enterprise_industry_analysis","score":1.0,"tokens":472100,"cost_usd":0.186252,"elapsed_time":227.23,"gpr":1.0,"tpe":0.2457,"ee":0.1765},{"task_id":"task_092_enterprise_industry_analysis_easy_easy022","category":"enterprise_industry_analysis","score":0.0,"tokens":29351,"cost_usd":0.013278,"elapsed_time":96.34,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_093_enterprise_industry_analysis_easy_easy023","category":"enterprise_industry_analysis","score":0.0,"tokens":61002,"cost_usd":0.021058,"elapsed_time":56.87,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_094_enterprise_industry_analysis_easy_easy024","category":"enterprise_industry_analysis","score":0.0,"tokens":698905,"cost_usd":0.251085,"elapsed_time":846.52,"gpr":0.3333,"tpe":1.0,"ee":0.2},{"task_id":"task_095_enterprise_industry_analysis_easy_easy025","category":"enterprise_industry_analysis","score":0.0,"tokens":349369,"cost_usd":0.138194,"elapsed_time":136.26,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_096_enterprise_industry_analysis_easy_easy026","category":"enterprise_industry_analysis","score":0.0,"tokens":19182,"cost_usd":0.008671,"elapsed_time":17.46,"gpr":0.0,"tpe":0.0,"ee":1.5},{"task_id":"task_097_enterprise_industry_analysis_easy_easy027","category":"enterprise_industry_analysis","score":0.0,"tokens":306014,"cost_usd":0.111851,"elapsed_time":178.31,"gpr":0.3333,"tpe":0.1501,"ee":0.1364},{"task_id":"task_098_enterprise_industry_analysis_easy_easy028","category":"enterprise_industry_analysis","score":1.0,"tokens":52947,"cost_usd":0.024069,"elapsed_time":53.8,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_099_enterprise_industry_analysis_easy_easy029","category":"enterprise_industry_analysis","score":0.0,"tokens":1139465,"cost_usd":0.423672,"elapsed_time":362.95,"gpr":0.0,"tpe":0.0,"ee":0.12},{"task_id":"task_100_enterprise_industry_analysis_easy_easy030","category":"enterprise_industry_analysis","score":0.0,"tokens":1912580,"cost_usd":0.583073,"elapsed_time":540.56,"gpr":0.0,"tpe":0.0,"ee":0.1579},{"task_id":"task_101_enterprise_industry_analysis_easy_easy031","category":"enterprise_industry_analysis","score":0.0,"tokens":9872,"cost_usd":0.004344,"elapsed_time":22.51,"gpr":0.0,"tpe":0.0,"ee":3.0},{"task_id":"task_102_enterprise_industry_analysis_easy_easy032","category":"enterprise_industry_analysis","score":0.0,"tokens":19680,"cost_usd":0.006875,"elapsed_time":41.85,"gpr":0.0,"tpe":0.0,"ee":1.5},{"task_id":"task_103_enterprise_industry_analysis_easy_easy033","category":"enterprise_industry_analysis","score":0.0,"tokens":82089,"cost_usd":0.034398,"elapsed_time":64.32,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_104_enterprise_industry_analysis_easy_easy034","category":"enterprise_industry_analysis","score":0.0,"tokens":3124410,"cost_usd":1.111865,"elapsed_time":711.9,"gpr":0.0,"tpe":0.0,"ee":0.1429},{"task_id":"task_105_enterprise_industry_analysis_easy_easy035","category":"enterprise_industry_analysis","score":1.0,"tokens":97575,"cost_usd":0.038522,"elapsed_time":91.32,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_106_enterprise_industry_analysis_easy_easy036","category":"enterprise_industry_analysis","score":1.0,"tokens":75130,"cost_usd":0.028793,"elapsed_time":57.73,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_107_enterprise_industry_analysis_easy_easy037","category":"enterprise_industry_analysis","score":0.0,"tokens":91299,"cost_usd":0.04015,"elapsed_time":44.68,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_108_enterprise_industry_analysis_easy_easy038","category":"enterprise_industry_analysis","score":0.0,"tokens":1010263,"cost_usd":0.366933,"elapsed_time":423.0,"gpr":0.0,"tpe":0.0,"ee":0.1071},{"task_id":"task_109_enterprise_industry_analysis_easy_easy039","category":"enterprise_industry_analysis","score":1.0,"tokens":86449,"cost_usd":0.034076,"elapsed_time":74.69,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_110_enterprise_industry_analysis_easy_easy040","category":"enterprise_industry_analysis","score":1.0,"tokens":212380,"cost_usd":0.090147,"elapsed_time":772.81,"gpr":1.0,"tpe":0.1998,"ee":0.1579},{"task_id":"task_111_enterprise_industry_analysis_easy_easy041","category":"enterprise_industry_analysis","score":0.0,"tokens":9863,"cost_usd":0.00434,"elapsed_time":53.69,"gpr":0.0,"tpe":0.0,"ee":3.0},{"task_id":"task_112_enterprise_industry_analysis_easy_easy042","category":"enterprise_industry_analysis","score":1.0,"tokens":61100,"cost_usd":0.023398,"elapsed_time":73.34,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_113_enterprise_industry_analysis_easy_easy043","category":"enterprise_industry_analysis","score":1.0,"tokens":526992,"cost_usd":0.202914,"elapsed_time":146.78,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_114_enterprise_industry_analysis_easy_easy044","category":"enterprise_industry_analysis","score":1.0,"tokens":71119,"cost_usd":0.026964,"elapsed_time":56.8,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_115_enterprise_industry_analysis_easy_easy045","category":"enterprise_industry_analysis","score":0.0,"tokens":19233,"cost_usd":0.008775,"elapsed_time":123.72,"gpr":0.0,"tpe":0.0,"ee":1.5},{"task_id":"task_116_enterprise_industry_analysis_easy_easy046","category":"enterprise_industry_analysis","score":0.0,"tokens":55150,"cost_usd":0.018281,"elapsed_time":93.19,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_117_enterprise_industry_analysis_easy_easy047","category":"enterprise_industry_analysis","score":0.0,"tokens":19576,"cost_usd":0.008709,"elapsed_time":30.65,"gpr":0.0,"tpe":0.0,"ee":1.5},{"task_id":"task_118_enterprise_industry_analysis_easy_easy048","category":"enterprise_industry_analysis","score":0.0,"tokens":51105,"cost_usd":0.016787,"elapsed_time":111.16,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_119_enterprise_industry_analysis_easy_easy049","category":"enterprise_industry_analysis","score":0.0,"tokens":29716,"cost_usd":0.009486,"elapsed_time":66.51,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_120_enterprise_industry_analysis_easy_easy050","category":"enterprise_industry_analysis","score":1.0,"tokens":684422,"cost_usd":0.292621,"elapsed_time":406.48,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_121_enterprise_industry_analysis_easy_easy051","category":"enterprise_industry_analysis","score":0.0,"tokens":79361,"cost_usd":0.035216,"elapsed_time":394.31,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_122_enterprise_industry_analysis_easy_easy052","category":"enterprise_industry_analysis","score":0.0,"tokens":80984,"cost_usd":0.03416,"elapsed_time":174.89,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_123_enterprise_industry_analysis_easy_easy053","category":"enterprise_industry_analysis","score":0.0,"tokens":50561,"cost_usd":0.022789,"elapsed_time":78.68,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_124_enterprise_industry_analysis_easy_easy054","category":"enterprise_industry_analysis","score":0.0,"tokens":19340,"cost_usd":0.008817,"elapsed_time":69.7,"gpr":0.0,"tpe":0.0,"ee":1.5},{"task_id":"task_125_enterprise_industry_analysis_easy_easy055","category":"enterprise_industry_analysis","score":0.0,"tokens":34464,"cost_usd":0.01537,"elapsed_time":88.1,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_126_enterprise_industry_analysis_easy_easy056","category":"enterprise_industry_analysis","score":0.0,"tokens":985805,"cost_usd":0.385865,"elapsed_time":316.87,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_127_enterprise_industry_analysis_easy_easy057","category":"enterprise_industry_analysis","score":0.0,"tokens":70250,"cost_usd":0.02955,"elapsed_time":100.2,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_128_enterprise_industry_analysis_easy_easy058","category":"enterprise_industry_analysis","score":0.0,"tokens":19446,"cost_usd":0.006741,"elapsed_time":88.56,"gpr":0.0,"tpe":0.0,"ee":1.5},{"task_id":"task_129_enterprise_industry_analysis_easy_easy059","category":"enterprise_industry_analysis","score":0.0,"tokens":24814,"cost_usd":0.007398,"elapsed_time":55.07,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_130_enterprise_industry_analysis_easy_easy060","category":"enterprise_industry_analysis","score":0.0,"tokens":70087,"cost_usd":0.030951,"elapsed_time":111.47,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_131_enterprise_industry_analysis_easy_easy061","category":"enterprise_industry_analysis","score":0.0,"tokens":425999,"cost_usd":0.185052,"elapsed_time":257.34,"gpr":0.3333,"tpe":0.5314,"ee":0.3333},{"task_id":"task_132_enterprise_industry_analysis_easy_easy062","category":"enterprise_industry_analysis","score":0.0,"tokens":466920,"cost_usd":0.202307,"elapsed_time":162.43,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_133_enterprise_industry_analysis_easy_easy063","category":"enterprise_industry_analysis","score":0.0,"tokens":184505,"cost_usd":0.073044,"elapsed_time":166.41,"gpr":0.0,"tpe":0.0,"ee":0.1875},{"task_id":"task_134_enterprise_industry_analysis_easy_easy064","category":"enterprise_industry_analysis","score":0.0,"tokens":331512,"cost_usd":0.144616,"elapsed_time":149.31,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_135_enterprise_industry_analysis_easy_easy065","category":"enterprise_industry_analysis","score":1.0,"tokens":745072,"cost_usd":0.256997,"elapsed_time":442.98,"gpr":1.0,"tpe":0.1501,"ee":0.125},{"task_id":"task_136_enterprise_industry_analysis_easy_easy066","category":"enterprise_industry_analysis","score":1.0,"tokens":896962,"cost_usd":0.286268,"elapsed_time":255.96,"gpr":1.0,"tpe":0.575,"ee":0.3},{"task_id":"task_137_enterprise_industry_analysis_easy_easy067","category":"enterprise_industry_analysis","score":0.0,"tokens":92147,"cost_usd":0.03802,"elapsed_time":48.75,"gpr":0.6667,"tpe":0.81,"ee":0.5},{"task_id":"task_138_enterprise_industry_analysis_easy_easy068","category":"enterprise_industry_analysis","score":1.0,"tokens":1735005,"cost_usd":0.682203,"elapsed_time":633.32,"gpr":1.0,"tpe":0.3487,"ee":0.2308},{"task_id":"task_139_enterprise_industry_analysis_easy_easy069","category":"enterprise_industry_analysis","score":0.0,"tokens":167247,"cost_usd":0.07209,"elapsed_time":142.64,"gpr":0.0,"tpe":0.0,"ee":0.75},{"task_id":"task_140_enterprise_industry_analysis_easy_easy070","category":"enterprise_industry_analysis","score":0.0,"tokens":1201659,"cost_usd":0.463193,"elapsed_time":404.51,"gpr":0.3333,"tpe":0.4305,"ee":0.25},{"task_id":"task_141_enterprise_industry_analysis_easy_easy071","category":"enterprise_industry_analysis","score":1.0,"tokens":156037,"cost_usd":0.062959,"elapsed_time":87.8,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_142_enterprise_industry_analysis_easy_easy072","category":"enterprise_industry_analysis","score":1.0,"tokens":2661321,"cost_usd":0.984499,"elapsed_time":590.82,"gpr":1.0,"tpe":0.7046,"ee":0.2143},{"task_id":"task_143_enterprise_industry_analysis_easy_easy073","category":"enterprise_industry_analysis","score":0.0,"tokens":696777,"cost_usd":0.253666,"elapsed_time":234.67,"gpr":0.6667,"tpe":0.561,"ee":0.3},{"task_id":"task_144_enterprise_industry_analysis_easy_easy074","category":"enterprise_industry_analysis","score":0.0,"tokens":19670,"cost_usd":0.006914,"elapsed_time":50.38,"gpr":0.0,"tpe":0.0,"ee":1.5},{"task_id":"task_145_enterprise_industry_analysis_easy_easy075","category":"enterprise_industry_analysis","score":1.0,"tokens":339104,"cost_usd":0.130834,"elapsed_time":156.16,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_146_enterprise_industry_analysis_easy_easy076","category":"enterprise_industry_analysis","score":1.0,"tokens":74079,"cost_usd":0.025144,"elapsed_time":97.0,"gpr":1.0,"tpe":0.7317,"ee":0.4286},{"task_id":"task_147_enterprise_industry_analysis_easy_easy077","category":"enterprise_industry_analysis","score":1.0,"tokens":74002,"cost_usd":0.030773,"elapsed_time":52.93,"gpr":1.0,"tpe":0.8187,"ee":0.4286},{"task_id":"task_148_enterprise_industry_analysis_easy_easy078","category":"enterprise_industry_analysis","score":1.0,"tokens":73408,"cost_usd":0.026811,"elapsed_time":71.01,"gpr":1.0,"tpe":0.7317,"ee":0.4286},{"task_id":"task_149_enterprise_industry_analysis_easy_easy079","category":"enterprise_industry_analysis","score":0.0,"tokens":667972,"cost_usd":0.26256,"elapsed_time":274.43,"gpr":1.0,"tpe":0.6342,"ee":0.375},{"task_id":"task_150_enterprise_industry_analysis_easy_easy080","category":"enterprise_industry_analysis","score":1.0,"tokens":658117,"cost_usd":0.249401,"elapsed_time":293.95,"gpr":1.0,"tpe":0.1501,"ee":0.1429},{"task_id":"task_151_enterprise_industry_analysis_easy_easy081","category":"enterprise_industry_analysis","score":0.0,"tokens":19590,"cost_usd":0.008713,"elapsed_time":46.7,"gpr":0.0,"tpe":0.0,"ee":1.5},{"task_id":"task_152_enterprise_industry_analysis_easy_easy082","category":"enterprise_industry_analysis","score":1.0,"tokens":61379,"cost_usd":0.027232,"elapsed_time":37.85,"gpr":1.0,"tpe":0.783,"ee":0.5},{"task_id":"task_153_enterprise_industry_analysis_easy_easy083","category":"enterprise_industry_analysis","score":0.0,"tokens":1333327,"cost_usd":0.475767,"elapsed_time":443.09,"gpr":0.0,"tpe":0.0,"ee":0.075},{"task_id":"task_154_enterprise_industry_analysis_easy_easy084","category":"enterprise_industry_analysis","score":1.0,"tokens":918315,"cost_usd":0.403086,"elapsed_time":237.45,"gpr":1.0,"tpe":0.8187,"ee":0.4286},{"task_id":"task_155_enterprise_industry_analysis_easy_easy085","category":"enterprise_industry_analysis","score":1.0,"tokens":661125,"cost_usd":0.260587,"elapsed_time":176.18,"gpr":1.0,"tpe":0.448,"ee":0.25},{"task_id":"task_156_enterprise_industry_analysis_easy_easy086","category":"enterprise_industry_analysis","score":1.0,"tokens":2181616,"cost_usd":0.852179,"elapsed_time":653.55,"gpr":1.0,"tpe":0.3034,"ee":0.2},{"task_id":"task_157_enterprise_industry_analysis_easy_easy088","category":"enterprise_industry_analysis","score":1.0,"tokens":34543,"cost_usd":0.014198,"elapsed_time":44.22,"gpr":1.0,"tpe":0.9,"ee":0.6667},{"task_id":"task_158_enterprise_industry_analysis_easy_easy090","category":"enterprise_industry_analysis","score":0.0,"tokens":153922,"cost_usd":0.056906,"elapsed_time":98.31,"gpr":0.6667,"tpe":0.5049,"ee":0.2727},{"task_id":"task_159_enterprise_industry_analysis_easy_easy091","category":"enterprise_industry_analysis","score":0.0,"tokens":34144,"cost_usd":0.013642,"elapsed_time":38.05,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_160_enterprise_industry_analysis_easy_easy092","category":"enterprise_industry_analysis","score":0.0,"tokens":474361,"cost_usd":0.182596,"elapsed_time":186.99,"gpr":0.3333,"tpe":0.5905,"ee":0.1667},{"task_id":"task_161_enterprise_industry_analysis_easy_easy093","category":"enterprise_industry_analysis","score":0.0,"tokens":324435,"cost_usd":0.115619,"elapsed_time":186.13,"gpr":1.0,"tpe":0.3138,"ee":0.2143},{"task_id":"task_162_enterprise_industry_analysis_easy_easy094","category":"enterprise_industry_analysis","score":1.0,"tokens":98196,"cost_usd":0.035777,"elapsed_time":131.64,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_163_enterprise_industry_analysis_easy_easy095","category":"enterprise_industry_analysis","score":1.0,"tokens":464415,"cost_usd":0.167728,"elapsed_time":708.21,"gpr":1.0,"tpe":0.2824,"ee":0.2},{"task_id":"task_164_enterprise_industry_analysis_easy_easy096","category":"enterprise_industry_analysis","score":1.0,"tokens":201570,"cost_usd":0.076284,"elapsed_time":139.64,"gpr":1.0,"tpe":0.2824,"ee":0.2},{"task_id":"task_165_enterprise_industry_analysis_easy_easy097","category":"enterprise_industry_analysis","score":1.0,"tokens":164356,"cost_usd":0.060791,"elapsed_time":194.62,"gpr":1.0,"tpe":0.3874,"ee":0.25},{"task_id":"task_166_enterprise_industry_analysis_easy_easy098","category":"enterprise_industry_analysis","score":1.0,"tokens":1183045,"cost_usd":0.41638,"elapsed_time":290.18,"gpr":1.0,"tpe":0.2495,"ee":0.1364},{"task_id":"task_167_enterprise_industry_analysis_easy_easy099","category":"enterprise_industry_analysis","score":0.0,"tokens":38617,"cost_usd":0.01328,"elapsed_time":50.76,"gpr":0.0,"tpe":0.0,"ee":0.75},{"task_id":"task_168_enterprise_industry_analysis_easy_easy100","category":"enterprise_industry_analysis","score":1.0,"tokens":147823,"cost_usd":0.049084,"elapsed_time":122.33,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_169_enterprise_industry_analysis_easy_easy101","category":"enterprise_industry_analysis","score":1.0,"tokens":39267,"cost_usd":0.013876,"elapsed_time":61.96,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_170_enterprise_industry_analysis_easy_easy102","category":"enterprise_industry_analysis","score":1.0,"tokens":62462,"cost_usd":0.02209,"elapsed_time":87.96,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_171_enterprise_industry_analysis_easy_easy103","category":"enterprise_industry_analysis","score":1.0,"tokens":53945,"cost_usd":0.019997,"elapsed_time":60.44,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_172_enterprise_industry_analysis_easy_easy104","category":"enterprise_industry_analysis","score":1.0,"tokens":171954,"cost_usd":0.064401,"elapsed_time":119.81,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_173_enterprise_industry_analysis_easy_easy105","category":"enterprise_industry_analysis","score":1.0,"tokens":61938,"cost_usd":0.02368,"elapsed_time":81.38,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_174_enterprise_industry_analysis_easy_easy106","category":"enterprise_industry_analysis","score":1.0,"tokens":265472,"cost_usd":0.094908,"elapsed_time":139.59,"gpr":1.0,"tpe":0.2824,"ee":0.2},{"task_id":"task_175_enterprise_industry_analysis_easy_easy107","category":"enterprise_industry_analysis","score":1.0,"tokens":124769,"cost_usd":0.047508,"elapsed_time":108.98,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_176_enterprise_industry_analysis_easy_easy108","category":"enterprise_industry_analysis","score":1.0,"tokens":182598,"cost_usd":0.072457,"elapsed_time":99.96,"gpr":1.0,"tpe":0.2288,"ee":0.1765},{"task_id":"task_177_enterprise_industry_analysis_easy_easy109","category":"enterprise_industry_analysis","score":1.0,"tokens":513546,"cost_usd":0.192386,"elapsed_time":232.3,"gpr":1.0,"tpe":0.1853,"ee":0.1579},{"task_id":"task_178_enterprise_industry_analysis_easy_easy110","category":"enterprise_industry_analysis","score":0.0,"tokens":215116,"cost_usd":0.076806,"elapsed_time":133.72,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_179_enterprise_industry_analysis_easy_easy111","category":"enterprise_industry_analysis","score":1.0,"tokens":71816,"cost_usd":0.022652,"elapsed_time":87.42,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_180_enterprise_industry_analysis_easy_easy112","category":"enterprise_industry_analysis","score":1.0,"tokens":86208,"cost_usd":0.032992,"elapsed_time":71.57,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_181_enterprise_industry_analysis_easy_easy113","category":"enterprise_industry_analysis","score":0.0,"tokens":213913,"cost_usd":0.075483,"elapsed_time":198.04,"gpr":0.0,"tpe":0.0,"ee":0.2143},{"task_id":"task_182_enterprise_industry_analysis_easy_easy114","category":"enterprise_industry_analysis","score":1.0,"tokens":67808,"cost_usd":0.023886,"elapsed_time":86.69,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_183_enterprise_industry_analysis_easy_easy115","category":"enterprise_industry_analysis","score":1.0,"tokens":1908046,"cost_usd":0.683645,"elapsed_time":620.49,"gpr":1.0,"tpe":0.0985,"ee":0.12},{"task_id":"task_184_enterprise_industry_analysis_easy_easy116","category":"enterprise_industry_analysis","score":0.0,"tokens":1731191,"cost_usd":0.590714,"elapsed_time":495.12,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_185_enterprise_industry_analysis_easy_easy117","category":"enterprise_industry_analysis","score":1.0,"tokens":190329,"cost_usd":0.075813,"elapsed_time":80.62,"gpr":1.0,"tpe":0.2824,"ee":0.2},{"task_id":"task_186_enterprise_industry_analysis_medium_medium001","category":"enterprise_industry_analysis","score":1.0,"tokens":2229168,"cost_usd":0.805918,"elapsed_time":925.43,"gpr":1.0,"tpe":0.0622,"ee":0.1176},{"task_id":"task_187_enterprise_industry_analysis_medium_medium002","category":"enterprise_industry_analysis","score":0.0,"tokens":185453,"cost_usd":0.07102,"elapsed_time":173.82,"gpr":0.75,"tpe":0.4225,"ee":0.2667},{"task_id":"task_188_enterprise_industry_analysis_medium_medium003","category":"enterprise_industry_analysis","score":0.0,"tokens":230838,"cost_usd":0.083382,"elapsed_time":150.74,"gpr":0.75,"tpe":0.6211,"ee":0.2857},{"task_id":"task_189_enterprise_industry_analysis_medium_medium004","category":"enterprise_industry_analysis","score":0.0,"tokens":3511197,"cost_usd":1.272122,"elapsed_time":1035.9,"gpr":0.5,"tpe":0.1041,"ee":0.125},{"task_id":"task_190_enterprise_industry_analysis_medium_medium005","category":"enterprise_industry_analysis","score":0.0,"tokens":389801,"cost_usd":0.138729,"elapsed_time":363.61,"gpr":0.0,"tpe":0.0,"ee":0.1538},{"task_id":"task_191_enterprise_industry_analysis_medium_medium006","category":"enterprise_industry_analysis","score":0.0,"tokens":2788419,"cost_usd":1.006363,"elapsed_time":1208.23,"gpr":0.0,"tpe":0.0,"ee":0.0816},{"task_id":"task_192_enterprise_industry_analysis_medium_medium007","category":"enterprise_industry_analysis","score":0.0,"tokens":490653,"cost_usd":0.170064,"elapsed_time":322.8,"gpr":0.4,"tpe":0.2403,"ee":0.1333},{"task_id":"task_193_enterprise_industry_analysis_medium_medium008","category":"enterprise_industry_analysis","score":1.0,"tokens":1228753,"cost_usd":0.44279,"elapsed_time":600.88,"gpr":0.6,"tpe":0.1542,"ee":0.1053},{"task_id":"task_194_enterprise_industry_analysis_medium_medium009","category":"enterprise_industry_analysis","score":0.0,"tokens":1919690,"cost_usd":0.796621,"elapsed_time":676.5,"gpr":0.2,"tpe":0.9,"ee":0.3077},{"task_id":"task_195_enterprise_industry_analysis_medium_medium010","category":"enterprise_industry_analysis","score":0.0,"tokens":907419,"cost_usd":0.363494,"elapsed_time":219.0,"gpr":0.4,"tpe":0.81,"ee":0.4},{"task_id":"task_196_enterprise_industry_analysis_medium_medium011","category":"enterprise_industry_analysis","score":1.0,"tokens":1231283,"cost_usd":0.444,"elapsed_time":523.49,"gpr":1.0,"tpe":0.4469,"ee":0.25},{"task_id":"task_197_enterprise_industry_analysis_medium_medium012","category":"enterprise_industry_analysis","score":1.0,"tokens":710358,"cost_usd":0.275173,"elapsed_time":471.98,"gpr":1.0,"tpe":0.1853,"ee":0.2},{"task_id":"task_198_enterprise_industry_analysis_medium_medium013","category":"enterprise_industry_analysis","score":0.0,"tokens":2204096,"cost_usd":0.930273,"elapsed_time":1207.56,"gpr":0.0,"tpe":0.0,"ee":0.2222},{"task_id":"task_199_enterprise_industry_analysis_medium_medium014","category":"enterprise_industry_analysis","score":0.0,"tokens":1263921,"cost_usd":0.478857,"elapsed_time":555.81,"gpr":0.2,"tpe":0.6561,"ee":0.2381},{"task_id":"task_200_enterprise_industry_analysis_medium_medium015","category":"enterprise_industry_analysis","score":0.0,"tokens":940377,"cost_usd":0.382794,"elapsed_time":468.32,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_201_enterprise_industry_analysis_medium_medium016","category":"enterprise_industry_analysis","score":0.0,"tokens":1663060,"cost_usd":0.667517,"elapsed_time":733.53,"gpr":0.8,"tpe":0.1074,"ee":0.1667},{"task_id":"task_202_enterprise_industry_analysis_medium_medium017","category":"enterprise_industry_analysis","score":1.0,"tokens":2033328,"cost_usd":0.81063,"elapsed_time":796.7,"gpr":1.0,"tpe":0.4108,"ee":0.2632},{"task_id":"task_203_enterprise_industry_analysis_medium_medium018","category":"enterprise_industry_analysis","score":0.0,"tokens":1350809,"cost_usd":0.471098,"elapsed_time":432.43,"gpr":0.6,"tpe":0.9367,"ee":0.2778},{"task_id":"task_204_enterprise_industry_analysis_medium_medium019","category":"enterprise_industry_analysis","score":0.0,"tokens":1344272,"cost_usd":0.517254,"elapsed_time":487.5,"gpr":0.2,"tpe":0.9,"ee":0.3125},{"task_id":"task_205_enterprise_industry_analysis_medium_medium020","category":"enterprise_industry_analysis","score":1.0,"tokens":364218,"cost_usd":0.160841,"elapsed_time":142.95,"gpr":1.0,"tpe":0.9058,"ee":0.625},{"task_id":"task_206_enterprise_industry_analysis_medium_medium021","category":"enterprise_industry_analysis","score":0.0,"tokens":1218202,"cost_usd":0.461592,"elapsed_time":841.43,"gpr":0.0,"tpe":0.0,"ee":0.2083},{"task_id":"task_207_enterprise_industry_analysis_medium_medium022","category":"enterprise_industry_analysis","score":1.0,"tokens":506496,"cost_usd":0.195127,"elapsed_time":507.87,"gpr":0.8,"tpe":0.3842,"ee":0.2174},{"task_id":"task_208_enterprise_industry_analysis_medium_medium023","category":"enterprise_industry_analysis","score":0.0,"tokens":825398,"cost_usd":0.326238,"elapsed_time":547.38,"gpr":0.8,"tpe":0.5582,"ee":0.3333},{"task_id":"task_209_enterprise_industry_analysis_medium_medium024","category":"enterprise_industry_analysis","score":1.0,"tokens":492556,"cost_usd":0.18886,"elapsed_time":562.01,"gpr":1.0,"tpe":0.3509,"ee":0.2632},{"task_id":"task_210_enterprise_industry_analysis_medium_medium025","category":"enterprise_industry_analysis","score":0.0,"tokens":561224,"cost_usd":0.204326,"elapsed_time":518.91,"gpr":0.4,"tpe":0.8645,"ee":0.2},{"task_id":"task_211_enterprise_industry_analysis_medium_medium026","category":"enterprise_industry_analysis","score":0.0,"tokens":1134187,"cost_usd":0.407398,"elapsed_time":606.01,"gpr":0.4,"tpe":0.4544,"ee":0.3846},{"task_id":"task_212_enterprise_industry_analysis_medium_medium027","category":"enterprise_industry_analysis","score":1.0,"tokens":878012,"cost_usd":0.247604,"elapsed_time":558.75,"gpr":1.0,"tpe":0.8303,"ee":0.4545},{"task_id":"task_213_enterprise_industry_analysis_medium_medium028","category":"enterprise_industry_analysis","score":0.0,"tokens":3573226,"cost_usd":1.28793,"elapsed_time":1207.45,"gpr":0.4,"tpe":0.2688,"ee":0.1613},{"task_id":"task_214_enterprise_industry_analysis_medium_medium029","category":"enterprise_industry_analysis","score":1.0,"tokens":416392,"cost_usd":0.162245,"elapsed_time":204.42,"gpr":1.0,"tpe":0.5314,"ee":0.4545},{"task_id":"task_215_enterprise_industry_analysis_medium_medium030","category":"enterprise_industry_analysis","score":1.0,"tokens":426868,"cost_usd":0.149143,"elapsed_time":347.55,"gpr":1.0,"tpe":0.2288,"ee":0.2632},{"task_id":"task_216_enterprise_industry_analysis_medium_medium031","category":"enterprise_industry_analysis","score":0.0,"tokens":1681747,"cost_usd":0.664079,"elapsed_time":903.63,"gpr":0.6,"tpe":0.2149,"ee":0.1613},{"task_id":"task_217_enterprise_industry_analysis_medium_medium032","category":"enterprise_industry_analysis","score":1.0,"tokens":825699,"cost_usd":0.303492,"elapsed_time":619.3,"gpr":1.0,"tpe":0.2701,"ee":0.1282},{"task_id":"task_218_enterprise_industry_analysis_medium_medium033","category":"enterprise_industry_analysis","score":0.0,"tokens":141289,"cost_usd":0.051386,"elapsed_time":193.42,"gpr":0.8,"tpe":0.817,"ee":0.625},{"task_id":"task_219_enterprise_industry_analysis_medium_medium034","category":"enterprise_industry_analysis","score":1.0,"tokens":489964,"cost_usd":0.16533,"elapsed_time":453.07,"gpr":1.0,"tpe":0.1827,"ee":0.2083},{"task_id":"task_220_enterprise_industry_analysis_medium_medium035","category":"enterprise_industry_analysis","score":0.0,"tokens":4753460,"cost_usd":1.620046,"elapsed_time":1207.55,"gpr":0.8,"tpe":0.3857,"ee":0.1429},{"task_id":"task_221_enterprise_industry_analysis_medium_medium036","category":"enterprise_industry_analysis","score":1.0,"tokens":1676790,"cost_usd":0.606466,"elapsed_time":648.98,"gpr":1.0,"tpe":0.0787,"ee":0.1613},{"task_id":"task_222_enterprise_industry_analysis_medium_medium037","category":"enterprise_industry_analysis","score":1.0,"tokens":1471953,"cost_usd":0.539657,"elapsed_time":538.35,"gpr":1.0,"tpe":0.3394,"ee":0.2174},{"task_id":"task_223_enterprise_industry_analysis_medium_medium038","category":"enterprise_industry_analysis","score":1.0,"tokens":423764,"cost_usd":0.171165,"elapsed_time":281.71,"gpr":1.0,"tpe":0.6444,"ee":0.5},{"task_id":"task_224_enterprise_industry_analysis_medium_medium039","category":"enterprise_industry_analysis","score":1.0,"tokens":259135,"cost_usd":0.106111,"elapsed_time":217.84,"gpr":1.0,"tpe":0.6817,"ee":0.3571},{"task_id":"task_225_enterprise_industry_analysis_medium_medium040","category":"enterprise_industry_analysis","score":0.0,"tokens":198637,"cost_usd":0.071451,"elapsed_time":192.37,"gpr":0.6,"tpe":0.7594,"ee":0.3846},{"task_id":"task_226_enterprise_industry_analysis_medium_medium041","category":"enterprise_industry_analysis","score":0.0,"tokens":1921771,"cost_usd":0.626646,"elapsed_time":471.53,"gpr":0.8,"tpe":0.8219,"ee":0.2632},{"task_id":"task_227_enterprise_industry_analysis_medium_medium042","category":"enterprise_industry_analysis","score":0.0,"tokens":4017234,"cost_usd":1.586289,"elapsed_time":981.98,"gpr":0.6,"tpe":0.6247,"ee":0.2083},{"task_id":"task_228_enterprise_industry_analysis_medium_medium043","category":"enterprise_industry_analysis","score":0.0,"tokens":665384,"cost_usd":0.198769,"elapsed_time":332.2,"gpr":0.4,"tpe":1.0,"ee":0.1724},{"task_id":"task_229_enterprise_industry_analysis_medium_medium044","category":"enterprise_industry_analysis","score":0.0,"tokens":558894,"cost_usd":0.192662,"elapsed_time":376.4,"gpr":0.8,"tpe":0.5505,"ee":0.3333},{"task_id":"task_230_enterprise_industry_analysis_medium_medium045","category":"enterprise_industry_analysis","score":0.0,"tokens":371037,"cost_usd":0.131329,"elapsed_time":269.67,"gpr":0.6,"tpe":0.3524,"ee":0.2778},{"task_id":"task_231_enterprise_industry_analysis_medium_medium046","category":"enterprise_industry_analysis","score":0.0,"tokens":485955,"cost_usd":0.160075,"elapsed_time":364.63,"gpr":0.2,"tpe":0.5314,"ee":0.25},{"task_id":"task_232_enterprise_industry_analysis_medium_medium047","category":"enterprise_industry_analysis","score":1.0,"tokens":974731,"cost_usd":0.343324,"elapsed_time":453.34,"gpr":1.0,"tpe":0.3752,"ee":0.2381},{"task_id":"task_233_enterprise_industry_analysis_medium_medium048","category":"enterprise_industry_analysis","score":0.0,"tokens":294853,"cost_usd":0.104137,"elapsed_time":214.35,"gpr":1.0,"tpe":0.467,"ee":0.3571},{"task_id":"task_234_enterprise_industry_analysis_medium_medium049","category":"enterprise_industry_analysis","score":0.0,"tokens":2528131,"cost_usd":0.946095,"elapsed_time":978.9,"gpr":0.6,"tpe":0.87,"ee":0.2778},{"task_id":"task_235_enterprise_industry_analysis_medium_medium050","category":"enterprise_industry_analysis","score":0.0,"tokens":1866367,"cost_usd":0.631407,"elapsed_time":646.48,"gpr":0.4,"tpe":1.0,"ee":0.2174},{"task_id":"task_236_enterprise_industry_analysis_medium_medium051","category":"enterprise_industry_analysis","score":0.0,"tokens":1272176,"cost_usd":0.494748,"elapsed_time":1130.65,"gpr":0.4,"tpe":0.9,"ee":0.2778},{"task_id":"task_237_enterprise_industry_analysis_medium_medium052","category":"enterprise_industry_analysis","score":0.0,"tokens":2218025,"cost_usd":0.923564,"elapsed_time":743.82,"gpr":0.8,"tpe":0.5905,"ee":0.2941},{"task_id":"task_238_enterprise_industry_analysis_medium_medium053","category":"enterprise_industry_analysis","score":0.0,"tokens":1632478,"cost_usd":0.547155,"elapsed_time":509.93,"gpr":0.8,"tpe":0.2122,"ee":0.1389},{"task_id":"task_239_enterprise_industry_analysis_medium_medium054","category":"enterprise_industry_analysis","score":1.0,"tokens":1476686,"cost_usd":0.617169,"elapsed_time":606.04,"gpr":1.0,"tpe":0.5208,"ee":0.4167},{"task_id":"task_240_enterprise_industry_analysis_medium_medium055","category":"enterprise_industry_analysis","score":1.0,"tokens":821386,"cost_usd":0.288128,"elapsed_time":313.68,"gpr":1.0,"tpe":0.2282,"ee":0.1562},{"task_id":"task_241_enterprise_industry_analysis_medium_medium056","category":"enterprise_industry_analysis","score":0.0,"tokens":477546,"cost_usd":0.200626,"elapsed_time":332.37,"gpr":0.0,"tpe":0.0,"ee":0.4545},{"task_id":"task_242_enterprise_industry_analysis_medium_medium057","category":"enterprise_industry_analysis","score":0.0,"tokens":1220114,"cost_usd":0.496824,"elapsed_time":450.88,"gpr":0.4,"tpe":0.5905,"ee":0.3846},{"task_id":"task_243_enterprise_industry_analysis_medium_medium058","category":"enterprise_industry_analysis","score":0.0,"tokens":113778,"cost_usd":0.045429,"elapsed_time":97.01,"gpr":0.2,"tpe":0.6561,"ee":0.5},{"task_id":"task_244_enterprise_industry_analysis_medium_medium059","category":"enterprise_industry_analysis","score":0.0,"tokens":69110,"cost_usd":0.028895,"elapsed_time":68.2,"gpr":0.0,"tpe":0.0,"ee":0.8333},{"task_id":"task_245_enterprise_industry_analysis_medium_medium060","category":"enterprise_industry_analysis","score":0.0,"tokens":19602,"cost_usd":0.008718,"elapsed_time":26.66,"gpr":0.0,"tpe":0.0,"ee":2.5},{"task_id":"task_246_enterprise_industry_analysis_medium_medium061","category":"enterprise_industry_analysis","score":1.0,"tokens":909168,"cost_usd":0.32388,"elapsed_time":375.76,"gpr":1.0,"tpe":0.5777,"ee":0.3571},{"task_id":"task_247_enterprise_industry_analysis_medium_medium062","category":"enterprise_industry_analysis","score":0.0,"tokens":155208,"cost_usd":0.054616,"elapsed_time":174.37,"gpr":0.0,"tpe":0.0,"ee":0.4545},{"task_id":"task_248_enterprise_industry_analysis_medium_medium063","category":"enterprise_industry_analysis","score":0.0,"tokens":39236,"cost_usd":0.013499,"elapsed_time":71.21,"gpr":0.0,"tpe":0.0,"ee":1.3333},{"task_id":"task_249_enterprise_industry_analysis_medium_medium064","category":"enterprise_industry_analysis","score":0.0,"tokens":20065,"cost_usd":0.007177,"elapsed_time":48.08,"gpr":0.0,"tpe":0.0,"ee":1.3333},{"task_id":"task_250_enterprise_industry_analysis_medium_medium065","category":"enterprise_industry_analysis","score":0.0,"tokens":315381,"cost_usd":0.104507,"elapsed_time":203.79,"gpr":0.25,"tpe":1.0,"ee":0.4444},{"task_id":"task_251_enterprise_industry_analysis_medium_medium066","category":"enterprise_industry_analysis","score":1.0,"tokens":1755083,"cost_usd":0.593943,"elapsed_time":456.45,"gpr":1.0,"tpe":0.3487,"ee":0.2857},{"task_id":"task_252_enterprise_industry_analysis_medium_medium067","category":"enterprise_industry_analysis","score":0.0,"tokens":51254,"cost_usd":0.016275,"elapsed_time":92.55,"gpr":0.0,"tpe":0.0,"ee":0.8},{"task_id":"task_253_enterprise_industry_analysis_medium_medium068","category":"enterprise_industry_analysis","score":0.0,"tokens":70632,"cost_usd":0.01573,"elapsed_time":159.93,"gpr":0.25,"tpe":1.0,"ee":0.4},{"task_id":"task_254_enterprise_industry_analysis_medium_medium069","category":"enterprise_industry_analysis","score":0.0,"tokens":925434,"cost_usd":0.38157,"elapsed_time":353.66,"gpr":0.25,"tpe":0.5314,"ee":0.2667},{"task_id":"task_255_enterprise_industry_analysis_medium_medium070","category":"enterprise_industry_analysis","score":0.0,"tokens":283702,"cost_usd":0.038933,"elapsed_time":753.13,"gpr":0.3333,"tpe":0.6561,"ee":0.4444},{"task_id":"task_256_enterprise_industry_analysis_medium_medium071","category":"enterprise_industry_analysis","score":0.0,"tokens":163115,"cost_usd":0.051366,"elapsed_time":156.62,"gpr":0.0,"tpe":0.0,"ee":0.4444},{"task_id":"task_257_enterprise_industry_analysis_medium_medium072","category":"enterprise_industry_analysis","score":1.0,"tokens":79970,"cost_usd":0.020851,"elapsed_time":126.86,"gpr":1.0,"tpe":0.7988,"ee":0.5},{"task_id":"task_258_enterprise_industry_analysis_medium_medium073","category":"enterprise_industry_analysis","score":1.0,"tokens":393368,"cost_usd":0.062807,"elapsed_time":127.11,"gpr":1.0,"tpe":0.7227,"ee":0.4444},{"task_id":"task_259_enterprise_industry_analysis_medium_medium074","category":"enterprise_industry_analysis","score":0.0,"tokens":2869173,"cost_usd":0.850628,"elapsed_time":1208.14,"gpr":1.0,"tpe":0.2308,"ee":0.1},{"task_id":"task_260_enterprise_industry_analysis_medium_medium075","category":"enterprise_industry_analysis","score":1.0,"tokens":491710,"cost_usd":0.168135,"elapsed_time":137.49,"gpr":0,"tpe":0.0,"ee":0.1818},{"task_id":"task_261_enterprise_industry_analysis_medium_medium076","category":"enterprise_industry_analysis","score":0.0,"tokens":312814,"cost_usd":0.099357,"elapsed_time":259.52,"gpr":0.75,"tpe":0.3556,"ee":0.1818},{"task_id":"task_262_enterprise_industry_analysis_medium_medium077","category":"enterprise_industry_analysis","score":0.0,"tokens":82775,"cost_usd":0.031003,"elapsed_time":76.48,"gpr":0.0,"tpe":0.0,"ee":0.5714},{"task_id":"task_263_enterprise_industry_analysis_medium_medium078","category":"enterprise_industry_analysis","score":0.0,"tokens":462098,"cost_usd":0.192078,"elapsed_time":176.16,"gpr":0.25,"tpe":0.81,"ee":0.5},{"task_id":"task_264_enterprise_industry_analysis_medium_medium079","category":"enterprise_industry_analysis","score":0.0,"tokens":290312,"cost_usd":0.120045,"elapsed_time":188.46,"gpr":0.25,"tpe":1.0,"ee":0.8},{"task_id":"task_265_enterprise_industry_analysis_medium_medium080","category":"enterprise_industry_analysis","score":0.0,"tokens":1074642,"cost_usd":0.398672,"elapsed_time":350.67,"gpr":0.0,"tpe":0.0,"ee":0.4444},{"task_id":"task_266_enterprise_industry_analysis_medium_medium081","category":"enterprise_industry_analysis","score":0.0,"tokens":32619,"cost_usd":0.008702,"elapsed_time":175.51,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_267_enterprise_industry_analysis_medium_medium082","category":"enterprise_industry_analysis","score":0.0,"tokens":128939,"cost_usd":0.045458,"elapsed_time":113.56,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_268_enterprise_industry_analysis_medium_medium083","category":"enterprise_industry_analysis","score":0.0,"tokens":157663,"cost_usd":0.050701,"elapsed_time":132.4,"gpr":0.25,"tpe":0.5905,"ee":0.3636},{"task_id":"task_269_enterprise_industry_analysis_medium_medium084","category":"enterprise_industry_analysis","score":1.0,"tokens":1666,"cost_usd":-0.02832,"elapsed_time":109.4,"gpr":1.0,"tpe":0.819,"ee":0.5},{"task_id":"task_270_enterprise_industry_analysis_medium_medium085","category":"enterprise_industry_analysis","score":0.0,"tokens":405680,"cost_usd":0.117416,"elapsed_time":242.34,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_271_enterprise_industry_analysis_medium_medium086","category":"enterprise_industry_analysis","score":0.0,"tokens":67588,"cost_usd":0.025928,"elapsed_time":60.15,"gpr":0.0,"tpe":0.0,"ee":1.3333},{"task_id":"task_272_enterprise_industry_analysis_medium_medium087","category":"enterprise_industry_analysis","score":1.0,"tokens":85145,"cost_usd":-0.033911,"elapsed_time":116.05,"gpr":1.0,"tpe":0.6902,"ee":0.25},{"task_id":"task_273_enterprise_industry_analysis_medium_medium088","category":"enterprise_industry_analysis","score":0.0,"tokens":112299,"cost_usd":0.024899,"elapsed_time":127.45,"gpr":0.5,"tpe":0.5344,"ee":0.3333},{"task_id":"task_274_enterprise_industry_analysis_medium_medium089","category":"enterprise_industry_analysis","score":1.0,"tokens":887112,"cost_usd":-0.021809,"elapsed_time":351.85,"gpr":1.0,"tpe":0.2591,"ee":0.1818},{"task_id":"task_275_enterprise_industry_analysis_medium_medium090","category":"enterprise_industry_analysis","score":0.0,"tokens":566634,"cost_usd":0.22945,"elapsed_time":185.2,"gpr":0.0,"tpe":0.0,"ee":0.5714},{"task_id":"task_276_enterprise_industry_analysis_medium_medium091","category":"enterprise_industry_analysis","score":0.0,"tokens":867037,"cost_usd":0.311107,"elapsed_time":300.71,"gpr":1.0,"tpe":0.4783,"ee":0.2222},{"task_id":"task_277_enterprise_industry_analysis_medium_medium092","category":"enterprise_industry_analysis","score":0.0,"tokens":19519,"cost_usd":0.008637,"elapsed_time":34.45,"gpr":0.0,"tpe":0.0,"ee":2.0},{"task_id":"task_278_enterprise_industry_analysis_medium_medium093","category":"enterprise_industry_analysis","score":0.0,"tokens":94828,"cost_usd":0.036053,"elapsed_time":68.11,"gpr":0.0,"tpe":0.0,"ee":0.8},{"task_id":"task_279_enterprise_industry_analysis_medium_medium094","category":"enterprise_industry_analysis","score":1.0,"tokens":836173,"cost_usd":0.250952,"elapsed_time":308.05,"gpr":1.0,"tpe":0.7219,"ee":0.2667},{"task_id":"task_280_enterprise_industry_analysis_medium_medium095","category":"enterprise_industry_analysis","score":0.0,"tokens":179975,"cost_usd":0.019877,"elapsed_time":149.31,"gpr":0.5,"tpe":0.6707,"ee":0.3077},{"task_id":"task_281_enterprise_industry_analysis_medium_medium096","category":"enterprise_industry_analysis","score":0.0,"tokens":146351,"cost_usd":0.037886,"elapsed_time":148.22,"gpr":0.25,"tpe":0.729,"ee":0.3077},{"task_id":"task_282_enterprise_industry_analysis_medium_medium097","category":"enterprise_industry_analysis","score":0.0,"tokens":1229247,"cost_usd":0.473078,"elapsed_time":370.73,"gpr":0.0,"tpe":0.0,"ee":0.2857},{"task_id":"task_283_enterprise_industry_analysis_medium_medium098","category":"enterprise_industry_analysis","score":0.0,"tokens":106439,"cost_usd":0.038957,"elapsed_time":76.19,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_284_enterprise_industry_analysis_medium_medium099","category":"enterprise_industry_analysis","score":0.0,"tokens":38622,"cost_usd":0.017133,"elapsed_time":35.86,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_285_enterprise_industry_analysis_medium_medium100","category":"enterprise_industry_analysis","score":0.0,"tokens":11253,"cost_usd":-0.024484,"elapsed_time":76.98,"gpr":0.75,"tpe":0.786,"ee":0.4},{"task_id":"task_286_enterprise_industry_analysis_medium_medium101","category":"enterprise_industry_analysis","score":0.0,"tokens":236683,"cost_usd":-0.045799,"elapsed_time":152.28,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_287_enterprise_industry_analysis_medium_medium102","category":"enterprise_industry_analysis","score":0.0,"tokens":204205,"cost_usd":0.080763,"elapsed_time":196.74,"gpr":0.0,"tpe":0.0,"ee":0.6667},{"task_id":"task_288_enterprise_industry_analysis_medium_medium103","category":"enterprise_industry_analysis","score":0.0,"tokens":633030,"cost_usd":0.241493,"elapsed_time":260.43,"gpr":0.25,"tpe":0.5314,"ee":0.3636},{"task_id":"task_289_enterprise_industry_analysis_medium_medium104","category":"enterprise_industry_analysis","score":0.0,"tokens":77075,"cost_usd":0.020334,"elapsed_time":105.37,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_290_enterprise_industry_analysis_medium_medium105","category":"enterprise_industry_analysis","score":0.0,"tokens":68403,"cost_usd":0.028472,"elapsed_time":46.76,"gpr":0.0,"tpe":0.0,"ee":0.6667},{"task_id":"task_291_enterprise_industry_analysis_medium_medium106","category":"enterprise_industry_analysis","score":0.0,"tokens":44438,"cost_usd":0.017917,"elapsed_time":45.33,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_292_enterprise_industry_analysis_medium_medium107","category":"enterprise_industry_analysis","score":0.0,"tokens":19586,"cost_usd":0.008701,"elapsed_time":27.24,"gpr":0.0,"tpe":0.0,"ee":2.0},{"task_id":"task_293_enterprise_industry_analysis_medium_medium108","category":"enterprise_industry_analysis","score":0.0,"tokens":997798,"cost_usd":0.213896,"elapsed_time":375.99,"gpr":0.25,"tpe":0.2824,"ee":0.2105},{"task_id":"task_294_enterprise_industry_analysis_medium_medium109","category":"enterprise_industry_analysis","score":1.0,"tokens":36408,"cost_usd":-0.014535,"elapsed_time":94.4,"gpr":1.0,"tpe":0.5626,"ee":0.4},{"task_id":"task_295_enterprise_industry_analysis_medium_medium110","category":"enterprise_industry_analysis","score":0.0,"tokens":387160,"cost_usd":0.090814,"elapsed_time":163.49,"gpr":0.25,"tpe":0.81,"ee":0.5},{"task_id":"task_296_enterprise_industry_analysis_medium_medium111","category":"enterprise_industry_analysis","score":0.0,"tokens":2170,"cost_usd":-0.060653,"elapsed_time":110.99,"gpr":0.5,"tpe":0.729,"ee":0.3636},{"task_id":"task_297_enterprise_industry_policy_analysis_easy_easy067","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":36152,"cost_usd":0.005994,"elapsed_time":51.49,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_298_enterprise_industry_policy_analysis_easy_easy068","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":178287,"cost_usd":0.056304,"elapsed_time":102.6,"gpr":1.0,"tpe":0.3874,"ee":0.25},{"task_id":"task_299_enterprise_industry_policy_analysis_easy_easy069","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":88813,"cost_usd":0.032789,"elapsed_time":56.79,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_300_enterprise_industry_policy_analysis_easy_easy070","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":182929,"cost_usd":0.075236,"elapsed_time":92.22,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_301_enterprise_industry_policy_analysis_easy_easy071","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":295346,"cost_usd":0.067042,"elapsed_time":142.53,"gpr":0.0,"tpe":0.0,"ee":0.1364},{"task_id":"task_302_enterprise_industry_policy_analysis_easy_easy072","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":20176,"cost_usd":-0.043665,"elapsed_time":76.71,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_303_enterprise_industry_policy_analysis_easy_easy073","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":191954,"cost_usd":0.069959,"elapsed_time":94.08,"gpr":1.0,"tpe":0.5905,"ee":0.375},{"task_id":"task_304_enterprise_industry_policy_analysis_easy_easy074","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":53932,"cost_usd":0.024031,"elapsed_time":36.77,"gpr":0.6,"tpe":0.81,"ee":0.6},{"task_id":"task_305_enterprise_industry_policy_analysis_easy_easy075","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":36423,"cost_usd":0.008578,"elapsed_time":47.7,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_306_enterprise_industry_policy_analysis_easy_easy076","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":19001,"cost_usd":0.006458,"elapsed_time":33.94,"gpr":0.0,"tpe":0.0,"ee":1.5},{"task_id":"task_307_enterprise_industry_policy_analysis_medium_medium001","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":857507,"cost_usd":0.224259,"elapsed_time":360.8,"gpr":1.0,"tpe":0.8136,"ee":0.4545},{"task_id":"task_308_enterprise_industry_policy_analysis_medium_medium002","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":10784,"cost_usd":-0.028413,"elapsed_time":84.7,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_309_enterprise_industry_policy_analysis_medium_medium003","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":183024,"cost_usd":0.072562,"elapsed_time":117.43,"gpr":0.4,"tpe":0.729,"ee":0.3846},{"task_id":"task_310_enterprise_industry_policy_analysis_medium_medium004","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":503812,"cost_usd":0.188906,"elapsed_time":146.78,"gpr":0.4,"tpe":1.0,"ee":0.4167},{"task_id":"task_311_enterprise_industry_policy_analysis_medium_medium005","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":38547,"cost_usd":0.017112,"elapsed_time":26.08,"gpr":0.0,"tpe":0.0,"ee":1.25},{"task_id":"task_312_enterprise_industry_policy_analysis_medium_medium006","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":22793,"cost_usd":0.004355,"elapsed_time":39.03,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_313_enterprise_industry_policy_analysis_medium_medium007","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2703702,"cost_usd":0.91229,"elapsed_time":882.01,"gpr":0.4,"tpe":0.9,"ee":0.1667},{"task_id":"task_314_enterprise_industry_policy_analysis_medium_medium008","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":61554,"cost_usd":0.025524,"elapsed_time":46.86,"gpr":0.0,"tpe":0.0,"ee":0.8333},{"task_id":"task_315_enterprise_industry_policy_analysis_medium_medium009","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":687956,"cost_usd":0.268872,"elapsed_time":200.48,"gpr":1.0,"tpe":0.8912,"ee":0.5556},{"task_id":"task_316_enterprise_industry_policy_analysis_medium_medium010","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":34223,"cost_usd":0.011454,"elapsed_time":37.54,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_317_enterprise_industry_policy_analysis_medium_medium011","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1420919,"cost_usd":0.508947,"elapsed_time":1053.99,"gpr":1.0,"tpe":0.2123,"ee":0.1613},{"task_id":"task_318_enterprise_industry_policy_analysis_medium_medium012","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":148677,"cost_usd":0.046078,"elapsed_time":121.5,"gpr":0.4,"tpe":0.7002,"ee":0.3333},{"task_id":"task_319_enterprise_industry_policy_analysis_medium_medium013","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":19334,"cost_usd":0.008858,"elapsed_time":19.84,"gpr":0.0,"tpe":0.0,"ee":2.5},{"task_id":"task_320_enterprise_industry_policy_analysis_medium_medium014","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":49321,"cost_usd":0.021918,"elapsed_time":47.78,"gpr":0.4,"tpe":1.0,"ee":1.0},{"task_id":"task_321_enterprise_industry_policy_analysis_medium_medium015","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":50212,"cost_usd":0.016484,"elapsed_time":60.65,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_322_enterprise_industry_policy_analysis_medium_medium016","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":392466,"cost_usd":0.101497,"elapsed_time":331.59,"gpr":1.0,"tpe":0.55,"ee":0.2381},{"task_id":"task_323_enterprise_industry_policy_analysis_medium_medium017","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":350752,"cost_usd":-0.051196,"elapsed_time":317.74,"gpr":1.0,"tpe":0.4906,"ee":0.3125},{"task_id":"task_324_enterprise_industry_policy_analysis_medium_medium018","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":927701,"cost_usd":0.290323,"elapsed_time":397.37,"gpr":0.8,"tpe":0.5156,"ee":0.2174},{"task_id":"task_325_enterprise_industry_policy_analysis_medium_medium019","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":364184,"cost_usd":0.136241,"elapsed_time":238.45,"gpr":1.0,"tpe":0.3938,"ee":0.3125},{"task_id":"task_326_enterprise_industry_policy_analysis_medium_medium020","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":22436,"cost_usd":0.002475,"elapsed_time":43.97,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_327_enterprise_industry_policy_analysis_medium_medium021","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1018196,"cost_usd":0.364641,"elapsed_time":302.91,"gpr":1.0,"tpe":0.5326,"ee":0.2},{"task_id":"task_328_enterprise_industry_policy_analysis_medium_medium022","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":214957,"cost_usd":0.08719,"elapsed_time":126.02,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_329_enterprise_industry_policy_analysis_medium_medium023","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":2884834,"cost_usd":0.863319,"elapsed_time":691.93,"gpr":1.0,"tpe":0.0542,"ee":0.1429},{"task_id":"task_330_enterprise_industry_policy_analysis_medium_medium024","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":5739337,"cost_usd":1.762249,"elapsed_time":1207.45,"gpr":0.8,"tpe":0.1562,"ee":0.1471},{"task_id":"task_331_enterprise_industry_policy_analysis_medium_medium025","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1167943,"cost_usd":0.307367,"elapsed_time":412.29,"gpr":1.0,"tpe":0.3628,"ee":0.1562},{"task_id":"task_332_enterprise_industry_policy_analysis_medium_medium026","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":428942,"cost_usd":-0.010584,"elapsed_time":194.93,"gpr":1.0,"tpe":0.4289,"ee":0.2381},{"task_id":"task_333_enterprise_industry_policy_analysis_medium_medium027","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1636207,"cost_usd":0.555803,"elapsed_time":472.97,"gpr":0.6,"tpe":0.8261,"ee":0.3571},{"task_id":"task_334_enterprise_industry_policy_analysis_medium_medium028","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":93557,"cost_usd":-0.028602,"elapsed_time":138.66,"gpr":1.0,"tpe":0.2015,"ee":0.1923},{"task_id":"task_335_enterprise_industry_policy_analysis_medium_medium029","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":291186,"cost_usd":0.061943,"elapsed_time":295.21,"gpr":1.0,"tpe":0.1094,"ee":0.1786},{"task_id":"task_336_enterprise_industry_policy_analysis_medium_medium030","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":444776,"cost_usd":0.147949,"elapsed_time":263.56,"gpr":0.6,"tpe":0.6086,"ee":0.2083},{"task_id":"task_337_enterprise_industry_policy_analysis_medium_medium031","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":936417,"cost_usd":0.38998,"elapsed_time":337.76,"gpr":1.0,"tpe":0.2683,"ee":0.2353},{"task_id":"task_338_enterprise_industry_policy_analysis_medium_medium032","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1113336,"cost_usd":0.338462,"elapsed_time":485.89,"gpr":0.75,"tpe":0.4441,"ee":0.1081},{"task_id":"task_339_enterprise_industry_policy_analysis_medium_medium033","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1037153,"cost_usd":0.341483,"elapsed_time":409.3,"gpr":1.0,"tpe":0.6337,"ee":0.3636},{"task_id":"task_340_enterprise_industry_policy_analysis_medium_medium034","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1399894,"cost_usd":0.434541,"elapsed_time":467.06,"gpr":0.5,"tpe":0.1074,"ee":0.0952},{"task_id":"task_341_enterprise_industry_policy_analysis_medium_medium035","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":203884,"cost_usd":-0.030355,"elapsed_time":221.37,"gpr":1.0,"tpe":0.8573,"ee":0.5714},{"task_id":"task_342_enterprise_industry_policy_analysis_medium_medium036","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1310725,"cost_usd":0.462523,"elapsed_time":413.38,"gpr":1.0,"tpe":0.2428,"ee":0.2105},{"task_id":"task_343_enterprise_industry_policy_analysis_medium_medium037","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":3941842,"cost_usd":1.327332,"elapsed_time":1009.81,"gpr":0.5,"tpe":0.81,"ee":0.1481},{"task_id":"task_344_enterprise_industry_policy_analysis_medium_medium038","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":234423,"cost_usd":0.100764,"elapsed_time":117.75,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_345_enterprise_industry_policy_analysis_medium_medium039","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1503760,"cost_usd":0.526582,"elapsed_time":458.4,"gpr":0.75,"tpe":0.3696,"ee":0.1905},{"task_id":"task_346_enterprise_industry_policy_analysis_medium_medium040","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":628270,"cost_usd":0.14031,"elapsed_time":328.29,"gpr":1.0,"tpe":0.0796,"ee":0.1111},{"task_id":"task_347_enterprise_industry_policy_analysis_medium_medium041","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":302463,"cost_usd":0.101475,"elapsed_time":203.01,"gpr":0.25,"tpe":0.4305,"ee":0.2105},{"task_id":"task_348_enterprise_industry_policy_analysis_medium_medium042","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1124952,"cost_usd":0.363249,"elapsed_time":435.44,"gpr":0.75,"tpe":0.4563,"ee":0.125},{"task_id":"task_349_enterprise_industry_policy_analysis_medium_medium043","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":850428,"cost_usd":0.269139,"elapsed_time":290.04,"gpr":1.0,"tpe":0.286,"ee":0.1333},{"task_id":"task_350_enterprise_industry_policy_analysis_medium_medium044","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2561,"cost_usd":-0.128413,"elapsed_time":172.81,"gpr":0.5,"tpe":0.7952,"ee":0.4444},{"task_id":"task_351_enterprise_industry_policy_analysis_medium_medium045","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":141353,"cost_usd":-0.08273,"elapsed_time":144.03,"gpr":0.5,"tpe":0.5905,"ee":0.4444},{"task_id":"task_352_enterprise_industry_policy_analysis_medium_medium046","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":116113,"cost_usd":-0.037015,"elapsed_time":150.54,"gpr":0.25,"tpe":0.5314,"ee":0.4},{"task_id":"task_353_enterprise_industry_policy_analysis_medium_medium047","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":689748,"cost_usd":0.193439,"elapsed_time":332.98,"gpr":1.0,"tpe":0.4305,"ee":0.1081},{"task_id":"task_354_enterprise_industry_policy_analysis_medium_medium048","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":161667,"cost_usd":0.046247,"elapsed_time":121.95,"gpr":1.0,"tpe":0.5477,"ee":0.3636},{"task_id":"task_355_enterprise_industry_policy_analysis_medium_medium049","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":208005,"cost_usd":0.065391,"elapsed_time":121.21,"gpr":1.0,"tpe":0.4238,"ee":0.2353},{"task_id":"task_356_enterprise_industry_policy_analysis_medium_medium050","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2557717,"cost_usd":0.818344,"elapsed_time":567.65,"gpr":0.5,"tpe":0.733,"ee":0.1905},{"task_id":"task_357_enterprise_industry_policy_analysis_medium_medium051","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2598947,"cost_usd":0.894535,"elapsed_time":697.38,"gpr":0.5,"tpe":0.2336,"ee":0.1081},{"task_id":"task_358_enterprise_industry_policy_analysis_medium_medium052","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1208065,"cost_usd":0.453357,"elapsed_time":318.57,"gpr":0.75,"tpe":0.7887,"ee":0.4444},{"task_id":"task_359_enterprise_industry_policy_analysis_medium_medium053","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":972583,"cost_usd":0.387625,"elapsed_time":272.48,"gpr":0.75,"tpe":0.5314,"ee":0.3636},{"task_id":"task_360_enterprise_industry_policy_analysis_medium_medium054","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2884353,"cost_usd":1.012064,"elapsed_time":847.47,"gpr":0.75,"tpe":0.6865,"ee":0.16},{"task_id":"task_361_enterprise_industry_policy_analysis_medium_medium055","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":98757,"cost_usd":-0.037451,"elapsed_time":73.74,"gpr":1.0,"tpe":0.8145,"ee":0.5714},{"task_id":"task_362_enterprise_industry_policy_analysis_medium_medium056","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2140353,"cost_usd":0.723476,"elapsed_time":688.28,"gpr":0.0,"tpe":0.0,"ee":0.1},{"task_id":"task_363_enterprise_industry_policy_analysis_medium_medium057","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":903037,"cost_usd":0.390532,"elapsed_time":262.72,"gpr":0.0,"tpe":0.0,"ee":0.5714},{"task_id":"task_364_enterprise_industry_policy_analysis_medium_medium058","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":3233179,"cost_usd":0.976771,"elapsed_time":687.91,"gpr":0.25,"tpe":0.4783,"ee":0.093},{"task_id":"task_365_enterprise_industry_policy_analysis_medium_medium059","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":19773,"cost_usd":0.007124,"elapsed_time":26.33,"gpr":0.0,"tpe":0.0,"ee":1.3333},{"task_id":"task_366_enterprise_industry_policy_analysis_medium_medium060","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":19218,"cost_usd":0.00876,"elapsed_time":19.58,"gpr":0.0,"tpe":0.0,"ee":2.0},{"task_id":"task_367_enterprise_industry_policy_analysis_medium_medium061","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2259717,"cost_usd":0.691835,"elapsed_time":421.67,"gpr":0.25,"tpe":0.3874,"ee":0.1538},{"task_id":"task_368_enterprise_industry_policy_analysis_medium_medium062","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":407933,"cost_usd":0.139239,"elapsed_time":193.62,"gpr":1.0,"tpe":0.7721,"ee":0.25},{"task_id":"task_369_enterprise_industry_policy_analysis_medium_medium063","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":162111,"cost_usd":0.05944,"elapsed_time":89.43,"gpr":0.0,"tpe":0.0,"ee":0.5714},{"task_id":"task_370_enterprise_industry_policy_analysis_medium_medium064","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":104629,"cost_usd":-0.001529,"elapsed_time":143.63,"gpr":0.25,"tpe":0.5905,"ee":0.2},{"task_id":"task_371_enterprise_industry_policy_analysis_medium_medium065","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":2959337,"cost_usd":0.767164,"elapsed_time":517.34,"gpr":0.0,"tpe":0.0,"ee":0.1538},{"task_id":"task_372_enterprise_industry_policy_analysis_medium_medium066","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":1526324,"cost_usd":0.452507,"elapsed_time":353.57,"gpr":1.0,"tpe":0.2643,"ee":0.16},{"task_id":"task_373_hypothesis_verification_hard_hard001","category":"hypothesis_verification","score":0.0,"tokens":9983,"cost_usd":0.002771,"elapsed_time":20.07,"gpr":0.0,"tpe":0.0,"ee":3.5},{"task_id":"task_374_hypothesis_verification_hard_hard002","category":"hypothesis_verification","score":1.0,"tokens":387053,"cost_usd":0.133936,"elapsed_time":225.97,"gpr":1.0,"tpe":0.2824,"ee":0.25},{"task_id":"task_375_hypothesis_verification_hard_hard003","category":"hypothesis_verification","score":0.0,"tokens":7649847,"cost_usd":2.169337,"elapsed_time":1207.42,"gpr":0.0,"tpe":0.0,"ee":0.1096},{"task_id":"task_376_hypothesis_verification_hard_hard004","category":"hypothesis_verification","score":0.0,"tokens":213349,"cost_usd":0.090511,"elapsed_time":150.92,"gpr":0.0,"tpe":0.0,"ee":2.3333},{"task_id":"task_377_hypothesis_verification_hard_hard005","category":"hypothesis_verification","score":0.0,"tokens":4764816,"cost_usd":1.296111,"elapsed_time":926.58,"gpr":0.0,"tpe":0.0,"ee":0.2045},{"task_id":"task_378_hypothesis_verification_hard_hard006","category":"hypothesis_verification","score":0.0,"tokens":75905,"cost_usd":0.026156,"elapsed_time":41.76,"gpr":0.7273,"tpe":1.0,"ee":1.1667},{"task_id":"task_379_hypothesis_verification_hard_hard007","category":"hypothesis_verification","score":0.0,"tokens":381144,"cost_usd":0.162691,"elapsed_time":130.33,"gpr":0.0,"tpe":0.0,"ee":1.8},{"task_id":"task_380_hypothesis_verification_hard_hard008","category":"hypothesis_verification","score":0.0,"tokens":4339568,"cost_usd":1.287252,"elapsed_time":914.53,"gpr":0.0,"tpe":0.0,"ee":0.1556},{"task_id":"task_381_hypothesis_verification_hard_hard009","category":"hypothesis_verification","score":0.0,"tokens":1095618,"cost_usd":0.333689,"elapsed_time":432.83,"gpr":0.0,"tpe":0.0,"ee":0.2143},{"task_id":"task_382_hypothesis_verification_hard_hard010","category":"hypothesis_verification","score":0.0,"tokens":1637676,"cost_usd":0.571827,"elapsed_time":684.14,"gpr":0.0,"tpe":0.0,"ee":0.5385},{"task_id":"task_383_hypothesis_verification_hard_hard011","category":"hypothesis_verification","score":0.0,"tokens":1094268,"cost_usd":0.430267,"elapsed_time":464.53,"gpr":0.0,"tpe":0.0,"ee":0.3889},{"task_id":"task_384_hypothesis_verification_hard_hard012","category":"hypothesis_verification","score":0.0,"tokens":2643613,"cost_usd":0.883843,"elapsed_time":791.58,"gpr":0.4545,"tpe":0.8085,"ee":0.1951},{"task_id":"task_385_hypothesis_verification_hard_hard013","category":"hypothesis_verification","score":0.0,"tokens":2204445,"cost_usd":0.735757,"elapsed_time":1207.89,"gpr":0.5333,"tpe":0.422,"ee":0.125},{"task_id":"task_386_hypothesis_verification_hard_hard014","category":"hypothesis_verification","score":0.0,"tokens":1023162,"cost_usd":0.407811,"elapsed_time":221.06,"gpr":0.0,"tpe":0.0,"ee":0.8571},{"task_id":"task_387_hypothesis_verification_hard_hard015","category":"hypothesis_verification","score":0.0,"tokens":378255,"cost_usd":0.166976,"elapsed_time":142.32,"gpr":0.0,"tpe":0.0,"ee":1.4},{"task_id":"task_388_hypothesis_verification_medium_medium001","category":"hypothesis_verification","score":0.0,"tokens":1200739,"cost_usd":0.523475,"elapsed_time":288.56,"gpr":0.0,"tpe":0.0,"ee":0.7778},{"task_id":"task_389_hypothesis_verification_medium_medium002","category":"hypothesis_verification","score":0.0,"tokens":205050,"cost_usd":0.08115,"elapsed_time":97.6,"gpr":0.0,"tpe":0.0,"ee":0.6364},{"task_id":"task_390_hypothesis_verification_medium_medium003","category":"hypothesis_verification","score":0.0,"tokens":1081644,"cost_usd":0.329951,"elapsed_time":405.58,"gpr":0.0,"tpe":0.0,"ee":0.1622},{"task_id":"task_391_hypothesis_verification_medium_medium004","category":"hypothesis_verification","score":0.0,"tokens":59833,"cost_usd":0.026789,"elapsed_time":42.83,"gpr":0.0,"tpe":0.0,"ee":1.1667},{"task_id":"task_392_hypothesis_verification_medium_medium005","category":"hypothesis_verification","score":0.0,"tokens":5155684,"cost_usd":1.553022,"elapsed_time":877.11,"gpr":0.0,"tpe":0.0,"ee":0.2258},{"task_id":"task_393_hypothesis_verification_medium_medium006","category":"hypothesis_verification","score":0.0,"tokens":3131870,"cost_usd":1.009456,"elapsed_time":695.93,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_394_hypothesis_verification_medium_medium007","category":"hypothesis_verification","score":0.0,"tokens":2605774,"cost_usd":0.950208,"elapsed_time":1021.92,"gpr":0.1111,"tpe":0.2824,"ee":0.2917},{"task_id":"task_395_hypothesis_verification_medium_medium008","category":"hypothesis_verification","score":0.0,"tokens":591140,"cost_usd":0.204629,"elapsed_time":347.0,"gpr":0.0,"tpe":0.0,"ee":0.4211},{"task_id":"task_396_hypothesis_verification_medium_medium009","category":"hypothesis_verification","score":0.0,"tokens":19372,"cost_usd":0.008873,"elapsed_time":21.84,"gpr":0.0,"tpe":0.0,"ee":3.0},{"task_id":"task_397_hypothesis_verification_medium_medium010","category":"hypothesis_verification","score":0.0,"tokens":4988521,"cost_usd":1.603536,"elapsed_time":1207.53,"gpr":0.1429,"tpe":0.0646,"ee":0.1944},{"task_id":"task_398_hypothesis_verification_medium_medium011","category":"hypothesis_verification","score":0.0,"tokens":260056,"cost_usd":0.114794,"elapsed_time":149.14,"gpr":0.0,"tpe":0.0,"ee":1.2},{"task_id":"task_399_hypothesis_verification_medium_medium012","category":"hypothesis_verification","score":0.0,"tokens":479403,"cost_usd":0.171374,"elapsed_time":190.66,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_400_hypothesis_verification_medium_medium013","category":"hypothesis_verification","score":0.0,"tokens":4135152,"cost_usd":1.270597,"elapsed_time":1207.38,"gpr":0.0,"tpe":0.0,"ee":0.1667},{"task_id":"task_401_hypothesis_verification_medium_medium014","category":"hypothesis_verification","score":0.0,"tokens":926489,"cost_usd":0.327417,"elapsed_time":267.8,"gpr":0.2,"tpe":1.0,"ee":0.4706},{"task_id":"task_402_industry_planning_hard_hard001","category":"industry_planning","score":0.0,"tokens":1612077,"cost_usd":0.621856,"elapsed_time":375.5,"gpr":0.4444,"tpe":1.0,"ee":0.7},{"task_id":"task_403_industry_planning_hard_hard002","category":"industry_planning","score":0.0,"tokens":296592,"cost_usd":0.13088,"elapsed_time":239.12,"gpr":0.0833,"tpe":1.0,"ee":2.3333},{"task_id":"task_404_industry_planning_hard_hard003","category":"industry_planning","score":0.5,"tokens":1069027,"cost_usd":0.40765,"elapsed_time":212.43,"gpr":0.0,"tpe":0.0,"ee":0.5714},{"task_id":"task_405_industry_planning_hard_hard004","category":"industry_planning","score":0.0,"tokens":1803951,"cost_usd":0.695118,"elapsed_time":370.11,"gpr":0.1667,"tpe":0.729,"ee":0.4615},{"task_id":"task_406_industry_planning_hard_hard005","category":"industry_planning","score":0.0,"tokens":29747,"cost_usd":0.011816,"elapsed_time":32.33,"gpr":0.0,"tpe":0.0,"ee":1.5},{"task_id":"task_407_industry_planning_hard_hard006","category":"industry_planning","score":0.0,"tokens":99638,"cost_usd":0.037021,"elapsed_time":96.76,"gpr":0.0,"tpe":0.0,"ee":0.7},{"task_id":"task_408_industry_planning_hard_hard007","category":"industry_planning","score":0.5,"tokens":4295001,"cost_usd":1.440563,"elapsed_time":947.14,"gpr":0.2,"tpe":0.1051,"ee":0.1765},{"task_id":"task_409_industry_planning_hard_hard008","category":"industry_planning","score":0.0,"tokens":774099,"cost_usd":0.328828,"elapsed_time":585.41,"gpr":0.5,"tpe":0.7963,"ee":0.1935},{"task_id":"task_410_industry_planning_hard_hard009","category":"industry_planning","score":0.0,"tokens":71352,"cost_usd":0.031062,"elapsed_time":247.48,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_411_industry_planning_hard_hard010","category":"industry_planning","score":0.0,"tokens":342373,"cost_usd":0.106319,"elapsed_time":209.51,"gpr":0.1111,"tpe":0.6561,"ee":0.5},{"task_id":"task_412_industry_planning_hard_hard011","category":"industry_planning","score":1.0,"tokens":3653259,"cost_usd":1.253106,"elapsed_time":789.68,"gpr":1.0,"tpe":0.1007,"ee":0.1935},{"task_id":"task_413_industry_planning_hard_hard012","category":"industry_planning","score":0.0,"tokens":1009311,"cost_usd":0.358623,"elapsed_time":269.43,"gpr":0.0,"tpe":0.0,"ee":0.8571},{"task_id":"task_414_industry_planning_hard_hard013","category":"industry_planning","score":0.0,"tokens":608497,"cost_usd":0.264564,"elapsed_time":194.61,"gpr":0.0,"tpe":0.0,"ee":1.4},{"task_id":"task_415_industry_planning_hard_hard014","category":"industry_planning","score":0.0,"tokens":3462596,"cost_usd":1.163311,"elapsed_time":1020.44,"gpr":0.125,"tpe":0.5314,"ee":0.1923},{"task_id":"task_416_industry_planning_medium_medium001","category":"industry_planning","score":0.0,"tokens":943663,"cost_usd":0.407574,"elapsed_time":265.68,"gpr":0.0,"tpe":0.0,"ee":0.8889},{"task_id":"task_417_industry_planning_medium_medium002","category":"industry_planning","score":0.0,"tokens":861436,"cost_usd":0.316025,"elapsed_time":299.51,"gpr":0.0,"tpe":0.0,"ee":0.6667},{"task_id":"task_418_industry_planning_medium_medium003","category":"industry_planning","score":0.0,"tokens":3340446,"cost_usd":1.141594,"elapsed_time":675.13,"gpr":0.0,"tpe":0.0,"ee":0.381},{"task_id":"task_419_industry_planning_medium_medium004","category":"industry_planning","score":1.0,"tokens":657587,"cost_usd":0.136745,"elapsed_time":469.7,"gpr":1.0,"tpe":0.3863,"ee":0.3182},{"task_id":"task_420_industry_planning_medium_medium005","category":"industry_planning","score":0.0,"tokens":2951137,"cost_usd":1.025465,"elapsed_time":887.72,"gpr":0.1111,"tpe":0.0886,"ee":0.2},{"task_id":"task_421_industry_planning_medium_medium006","category":"industry_planning","score":0.0,"tokens":304662,"cost_usd":0.107603,"elapsed_time":147.76,"gpr":0.0,"tpe":0.0,"ee":0.6154},{"task_id":"task_422_industry_planning_medium_medium007","category":"industry_planning","score":0.0,"tokens":31486,"cost_usd":0.010754,"elapsed_time":44.06,"gpr":0.0,"tpe":0.0,"ee":2.0},{"task_id":"task_423_industry_planning_medium_medium008","category":"industry_planning","score":0.0,"tokens":45280,"cost_usd":0.020162,"elapsed_time":150.33,"gpr":0.0,"tpe":0.0,"ee":2.0},{"task_id":"task_424_industry_planning_medium_medium009","category":"industry_planning","score":0.0,"tokens":4296924,"cost_usd":1.669645,"elapsed_time":1207.53,"gpr":0.1,"tpe":0.9,"ee":0.2963},{"task_id":"task_425_industry_planning_medium_medium010","category":"industry_planning","score":0.0,"tokens":446508,"cost_usd":0.197208,"elapsed_time":207.65,"gpr":0.0,"tpe":0.0,"ee":2.0},{"task_id":"task_426_industry_planning_medium_medium011","category":"industry_planning","score":1.0,"tokens":431219,"cost_usd":0.165435,"elapsed_time":338.22,"gpr":1.0,"tpe":0.5049,"ee":0.5333},{"task_id":"task_427_industry_planning_medium_medium012","category":"industry_planning","score":0.0,"tokens":1515505,"cost_usd":0.483165,"elapsed_time":581.71,"gpr":0.0,"tpe":0.0,"ee":0.1818},{"task_id":"task_428_industry_planning_medium_medium013","category":"industry_planning","score":0.0,"tokens":687165,"cost_usd":0.236502,"elapsed_time":315.74,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_429_industry_planning_medium_medium014","category":"industry_planning","score":1.0,"tokens":2252952,"cost_usd":0.721907,"elapsed_time":622.18,"gpr":0.3333,"tpe":0.2824,"ee":0.36},{"task_id":"task_430_international_comparison_hard_hard001","category":"international_comparison","score":0.0,"tokens":6226893,"cost_usd":2.475154,"elapsed_time":1225.82,"gpr":0.1,"tpe":0.1853,"ee":0},{"task_id":"task_431_international_comparison_hard_hard002","category":"international_comparison","score":0.0,"tokens":2394921,"cost_usd":0.808972,"elapsed_time":391.13,"gpr":0.0,"tpe":0.0,"ee":0},{"task_id":"task_432_international_comparison_hard_hard003","category":"international_comparison","score":0.0,"tokens":468101,"cost_usd":0.184071,"elapsed_time":166.36,"gpr":0.1111,"tpe":1.0,"ee":0.5},{"task_id":"task_433_international_comparison_hard_hard004","category":"international_comparison","score":0.0,"tokens":803537,"cost_usd":0.301507,"elapsed_time":191.37,"gpr":0.0909,"tpe":1.0,"ee":0},{"task_id":"task_434_international_comparison_hard_hard005","category":"international_comparison","score":1.0,"tokens":4809225,"cost_usd":1.362548,"elapsed_time":714.76,"gpr":0,"tpe":0.0,"ee":0.0833},{"task_id":"task_435_international_comparison_hard_hard006","category":"international_comparison","score":0.0,"tokens":2575956,"cost_usd":0.755136,"elapsed_time":561.24,"gpr":0.0,"tpe":0.0,"ee":0},{"task_id":"task_436_international_comparison_hard_hard007","category":"international_comparison","score":0.25,"tokens":5189878,"cost_usd":1.540052,"elapsed_time":689.24,"gpr":0.1818,"tpe":0.3079,"ee":0},{"task_id":"task_437_international_comparison_hard_hard008","category":"international_comparison","score":0.25,"tokens":1891146,"cost_usd":0.557081,"elapsed_time":329.66,"gpr":0.1667,"tpe":0.0148,"ee":0},{"task_id":"task_438_international_comparison_hard_hard009","category":"international_comparison","score":0.75,"tokens":2488041,"cost_usd":0.746995,"elapsed_time":490.88,"gpr":1.0,"tpe":0.0161,"ee":0},{"task_id":"task_439_international_comparison_hard_hard010","category":"international_comparison","score":1.0,"tokens":1558814,"cost_usd":0.471384,"elapsed_time":459.55,"gpr":1.0,"tpe":0.3449,"ee":0.2857},{"task_id":"task_440_international_comparison_hard_hard011","category":"international_comparison","score":0.5,"tokens":1695087,"cost_usd":0.666575,"elapsed_time":499.02,"gpr":0.4,"tpe":0.4135,"ee":0.375},{"task_id":"task_441_international_comparison_hard_hard012","category":"international_comparison","score":0.0,"tokens":2474279,"cost_usd":0.841747,"elapsed_time":656.48,"gpr":0.1429,"tpe":0.2059,"ee":0.2857},{"task_id":"task_442_international_comparison_hard_hard013","category":"international_comparison","score":0.5,"tokens":1033327,"cost_usd":0.380139,"elapsed_time":331.22,"gpr":0.6,"tpe":0.7707,"ee":0},{"task_id":"task_443_international_comparison_hard_hard014","category":"international_comparison","score":0.25,"tokens":2515399,"cost_usd":0.76353,"elapsed_time":384.38,"gpr":0.0,"tpe":0.0,"ee":0.2609},{"task_id":"task_444_international_comparison_medium_medium001","category":"international_comparison","score":0.0,"tokens":6049373,"cost_usd":2.489367,"elapsed_time":1228.57,"gpr":0.4,"tpe":0.3265,"ee":0},{"task_id":"task_445_international_comparison_medium_medium002","category":"international_comparison","score":0.0,"tokens":1369892,"cost_usd":0.427518,"elapsed_time":289.18,"gpr":0.3333,"tpe":0.2759,"ee":0},{"task_id":"task_446_international_comparison_medium_medium003","category":"international_comparison","score":0.5,"tokens":1450546,"cost_usd":0.464698,"elapsed_time":298.64,"gpr":0.3333,"tpe":0.1741,"ee":0},{"task_id":"task_447_international_comparison_medium_medium004","category":"international_comparison","score":0.0,"tokens":5735759,"cost_usd":2.053026,"elapsed_time":1228.09,"gpr":0.1667,"tpe":0.0718,"ee":0},{"task_id":"task_448_international_comparison_medium_medium005","category":"international_comparison","score":0.0,"tokens":5364095,"cost_usd":2.021091,"elapsed_time":1227.8,"gpr":0.1667,"tpe":0.0523,"ee":0},{"task_id":"task_449_international_comparison_medium_medium006","category":"international_comparison","score":0.0,"tokens":3157197,"cost_usd":1.008623,"elapsed_time":758.88,"gpr":0.1667,"tpe":0.0886,"ee":0},{"task_id":"task_450_international_comparison_medium_medium007","category":"international_comparison","score":0.0,"tokens":3713421,"cost_usd":1.035425,"elapsed_time":588.49,"gpr":0.5,"tpe":0.0166,"ee":0},{"task_id":"task_451_international_comparison_medium_medium008","category":"international_comparison","score":1.0,"tokens":724522,"cost_usd":0.236961,"elapsed_time":226.22,"gpr":1.0,"tpe":0.4466,"ee":0.2},{"task_id":"task_452_international_comparison_medium_medium009","category":"international_comparison","score":0.0,"tokens":1732601,"cost_usd":0.532377,"elapsed_time":394.68,"gpr":0.2857,"tpe":0.4135,"ee":0},{"task_id":"task_453_international_comparison_medium_medium010","category":"international_comparison","score":0.0,"tokens":4901254,"cost_usd":1.854942,"elapsed_time":938.58,"gpr":0.2,"tpe":0.9,"ee":0},{"task_id":"task_454_international_comparison_medium_medium011","category":"international_comparison","score":1.0,"tokens":725587,"cost_usd":0.18515,"elapsed_time":279.64,"gpr":1.0,"tpe":0.5839,"ee":0.3636},{"task_id":"task_455_international_comparison_medium_medium012","category":"international_comparison","score":1.0,"tokens":368553,"cost_usd":0.13831,"elapsed_time":115.79,"gpr":1.0,"tpe":0.4579,"ee":0.3571},{"task_id":"task_456_international_comparison_medium_medium013","category":"international_comparison","score":0.33,"tokens":281522,"cost_usd":0.098968,"elapsed_time":111.74,"gpr":0.5,"tpe":0.5314,"ee":0.3636},{"task_id":"task_457_international_comparison_medium_medium014","category":"international_comparison","score":0.0,"tokens":5843270,"cost_usd":1.651584,"elapsed_time":1163.04,"gpr":0.5,"tpe":0.6096,"ee":0},{"task_id":"task_458_international_comparison_medium_medium015","category":"international_comparison","score":0.0,"tokens":1209254,"cost_usd":0.412668,"elapsed_time":258.44,"gpr":0.0,"tpe":0.0,"ee":0},{"task_id":"task_459_international_comparison_medium_medium016","category":"international_comparison","score":0.0,"tokens":3453961,"cost_usd":1.273667,"elapsed_time":816.78,"gpr":0.3333,"tpe":0.4096,"ee":0.1316},{"task_id":"task_460_international_comparison_medium_medium017","category":"international_comparison","score":0.0,"tokens":1238672,"cost_usd":0.430985,"elapsed_time":304.53,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_461_international_comparison_medium_medium018","category":"international_comparison","score":1.0,"tokens":419354,"cost_usd":0.150356,"elapsed_time":189.47,"gpr":1.0,"tpe":0.451,"ee":0.3333},{"task_id":"task_462_international_comparison_medium_medium019","category":"international_comparison","score":0.5,"tokens":1398895,"cost_usd":0.449701,"elapsed_time":262.77,"gpr":0.5,"tpe":0.557,"ee":0.3333},{"task_id":"task_463_international_comparison_medium_medium020","category":"international_comparison","score":0.0,"tokens":4137120,"cost_usd":1.450119,"elapsed_time":948.1,"gpr":0,"tpe":0.0,"ee":0},{"task_id":"task_464_international_comparison_medium_medium021","category":"international_comparison","score":0.0,"tokens":684775,"cost_usd":0.232373,"elapsed_time":323.14,"gpr":0.1429,"tpe":0.2824,"ee":0.2273},{"task_id":"task_465_international_comparison_medium_medium022","category":"international_comparison","score":0.0,"tokens":5239815,"cost_usd":2.053951,"elapsed_time":1222.63,"gpr":0,"tpe":0.0,"ee":0},{"task_id":"task_466_international_comparison_medium_medium023","category":"international_comparison","score":0.0,"tokens":3642081,"cost_usd":1.319965,"elapsed_time":965.14,"gpr":0.1667,"tpe":1.0,"ee":0.1},{"task_id":"task_467_international_comparison_medium_medium024","category":"international_comparison","score":0.333,"tokens":925759,"cost_usd":0.315296,"elapsed_time":363.29,"gpr":0.3333,"tpe":0.5613,"ee":0.2273},{"task_id":"task_468_international_comparison_medium_medium025","category":"international_comparison","score":0.0,"tokens":1622867,"cost_usd":0.568089,"elapsed_time":327.89,"gpr":0.1667,"tpe":1.0,"ee":0},{"task_id":"task_469_risk_assessment_hard_hard001","category":"risk_assessment","score":0.0,"tokens":4492886,"cost_usd":1.375231,"elapsed_time":1007.85,"gpr":0.2,"tpe":0.3487,"ee":0.2188},{"task_id":"task_470_risk_assessment_hard_hard002","category":"risk_assessment","score":0.0,"tokens":1305074,"cost_usd":0.437935,"elapsed_time":796.12,"gpr":0.0,"tpe":0.0,"ee":0.4118},{"task_id":"task_471_risk_assessment_hard_hard003","category":"risk_assessment","score":0.0,"tokens":242540,"cost_usd":0.10702,"elapsed_time":168.25,"gpr":0.0909,"tpe":1.0,"ee":2.3333},{"task_id":"task_472_risk_assessment_hard_hard004","category":"risk_assessment","score":0.0,"tokens":466382,"cost_usd":0.180109,"elapsed_time":439.78,"gpr":0.4,"tpe":0.8145,"ee":0.4375},{"task_id":"task_473_risk_assessment_hard_hard005","category":"risk_assessment","score":1.0,"tokens":322778,"cost_usd":0.123936,"elapsed_time":283.79,"gpr":1.0,"tpe":0.3138,"ee":0.3889},{"task_id":"task_474_risk_assessment_hard_hard006","category":"risk_assessment","score":0.0,"tokens":45725,"cost_usd":0.020501,"elapsed_time":105.67,"gpr":0.0,"tpe":0.0,"ee":1.75},{"task_id":"task_475_risk_assessment_hard_hard007","category":"risk_assessment","score":1.0,"tokens":2351564,"cost_usd":0.852913,"elapsed_time":540.55,"gpr":1.0,"tpe":0.5371,"ee":0.4667},{"task_id":"task_476_risk_assessment_hard_hard008","category":"risk_assessment","score":0.0,"tokens":238768,"cost_usd":0.098196,"elapsed_time":126.3,"gpr":0.0,"tpe":0.0,"ee":0.4375},{"task_id":"task_477_risk_assessment_hard_hard009","category":"risk_assessment","score":0.0,"tokens":19441,"cost_usd":0.00873,"elapsed_time":22.96,"gpr":0.0,"tpe":0.0,"ee":3.5},{"task_id":"task_478_risk_assessment_hard_hard010","category":"risk_assessment","score":0.0,"tokens":1727912,"cost_usd":0.66826,"elapsed_time":342.54,"gpr":0.5,"tpe":1.0,"ee":0.8},{"task_id":"task_479_risk_assessment_hard_hard011","category":"risk_assessment","score":0.0,"tokens":829660,"cost_usd":0.281498,"elapsed_time":140.76,"gpr":0.0,"tpe":0.0,"ee":1.1667},{"task_id":"task_480_risk_assessment_hard_hard012","category":"risk_assessment","score":1.0,"tokens":1750237,"cost_usd":0.582208,"elapsed_time":865.46,"gpr":1.0,"tpe":0.0375,"ee":0.1556},{"task_id":"task_481_risk_assessment_hard_hard013","category":"risk_assessment","score":0.0,"tokens":1289250,"cost_usd":0.498986,"elapsed_time":235.24,"gpr":0.0,"tpe":0.0,"ee":0.9},{"task_id":"task_482_risk_assessment_medium_medium001","category":"risk_assessment","score":0.0,"tokens":1989337,"cost_usd":0.667956,"elapsed_time":474.77,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_483_risk_assessment_medium_medium002","category":"risk_assessment","score":0.0,"tokens":228059,"cost_usd":0.082925,"elapsed_time":142.9,"gpr":0.0,"tpe":0.0,"ee":0.4615},{"task_id":"task_484_risk_assessment_medium_medium003","category":"risk_assessment","score":0.0,"tokens":1112453,"cost_usd":0.362987,"elapsed_time":423.11,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_485_risk_assessment_medium_medium004","category":"risk_assessment","score":0.0,"tokens":4687781,"cost_usd":1.429916,"elapsed_time":779.88,"gpr":0.1,"tpe":1.0,"ee":0.1852},{"task_id":"task_486_risk_assessment_medium_medium005","category":"risk_assessment","score":0.0,"tokens":682876,"cost_usd":0.245769,"elapsed_time":283.21,"gpr":0.25,"tpe":0.2288,"ee":0.2381},{"task_id":"task_487_risk_assessment_medium_medium006","category":"risk_assessment","score":0.5,"tokens":2052196,"cost_usd":0.699603,"elapsed_time":398.84,"gpr":0.1818,"tpe":0.5314,"ee":0.3333},{"task_id":"task_488_risk_assessment_medium_medium007","category":"risk_assessment","score":0.0,"tokens":1219175,"cost_usd":0.390368,"elapsed_time":408.4,"gpr":0.0,"tpe":0.0,"ee":0.1463},{"task_id":"task_489_risk_assessment_medium_medium008","category":"risk_assessment","score":0.333,"tokens":375970,"cost_usd":0.141512,"elapsed_time":225.98,"gpr":0.0,"tpe":0.0,"ee":0.4615},{"task_id":"task_490_risk_assessment_medium_medium009","category":"risk_assessment","score":0.0,"tokens":2446333,"cost_usd":0.771092,"elapsed_time":729.09,"gpr":0.0833,"tpe":0.3487,"ee":0.1667},{"task_id":"task_491_risk_assessment_medium_medium010","category":"risk_assessment","score":0.0,"tokens":180144,"cost_usd":0.070876,"elapsed_time":84.93,"gpr":0.3,"tpe":1.0,"ee":0.8571},{"task_id":"task_492_risk_assessment_medium_medium011","category":"risk_assessment","score":0.0,"tokens":587211,"cost_usd":0.255153,"elapsed_time":220.2,"gpr":0.0,"tpe":0.0,"ee":1.2}]},{"model":"gpt-5.4-2026-03-05","acc":0.2338,"ee":0.4631,"gpr":0.1845,"tpe":0.3306,"efficiency":{"total_tokens":206857854,"total_cost_usd":428.21,"total_requests":5610},"category_scores":{"comprehensive_decision":{"acc":0.3124,"ee":0.459,"gpr":0.1784,"tpe":0.3142,"count":70},"enterprise_industry_analysis":{"acc":0.2478,"ee":0.4233,"gpr":0.1565,"tpe":0.2312,"count":226},"enterprise_industry_policy_analysis":{"acc":0.2895,"ee":0.5589,"gpr":0.3111,"tpe":0.5086,"count":76},"hypothesis_verification":{"acc":0.0459,"ee":0.4,"gpr":0.1289,"tpe":0.3402,"count":29},"industry_planning":{"acc":0.125,"ee":0.5687,"gpr":0.165,"tpe":0.4215,"count":28},"international_comparison":{"acc":0.1624,"ee":0.2303,"gpr":0.1557,"tpe":0.4132,"count":39},"risk_assessment":{"acc":0.1667,"ee":0.601,"gpr":0.2483,"tpe":0.4504,"count":24}},"tasks":[{"task_id":"task_001_comprehensive_decision_easy_easy001","category":"comprehensive_decision","score":0.0,"tokens":423219,"cost_usd":0.629286,"elapsed_time":96.76,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_002_comprehensive_decision_easy_easy002","category":"comprehensive_decision","score":0.2,"tokens":312262,"cost_usd":0.793493,"elapsed_time":55.78,"gpr":0.1667,"tpe":0.6561,"ee":0.2857},{"task_id":"task_003_comprehensive_decision_easy_easy003","category":"comprehensive_decision","score":0.0,"tokens":389571,"cost_usd":0.969913,"elapsed_time":87.3,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_004_comprehensive_decision_easy_easy004","category":"comprehensive_decision","score":0.0,"tokens":79306,"cost_usd":0.156971,"elapsed_time":45.82,"gpr":0.5,"tpe":0.7047,"ee":0.2857},{"task_id":"task_005_comprehensive_decision_easy_easy005","category":"comprehensive_decision","score":0.0,"tokens":118859,"cost_usd":0.291185,"elapsed_time":60.5,"gpr":0.3333,"tpe":0.9,"ee":0.5},{"task_id":"task_006_comprehensive_decision_easy_easy006","category":"comprehensive_decision","score":0.0,"tokens":196481,"cost_usd":0.43459,"elapsed_time":76.19,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_007_comprehensive_decision_hard_hard001","category":"comprehensive_decision","score":0.0,"tokens":1705606,"cost_usd":4.222931,"elapsed_time":195.34,"gpr":0.125,"tpe":0.5314,"ee":0.4118},{"task_id":"task_008_comprehensive_decision_hard_hard002","category":"comprehensive_decision","score":0.0,"tokens":210225,"cost_usd":0.475154,"elapsed_time":103.91,"gpr":0.8,"tpe":0.81,"ee":0.7},{"task_id":"task_009_comprehensive_decision_hard_hard003","category":"comprehensive_decision","score":0.0,"tokens":2305627,"cost_usd":5.384283,"elapsed_time":293.86,"gpr":0.2857,"tpe":0.2288,"ee":0.2917},{"task_id":"task_010_comprehensive_decision_hard_hard004","category":"comprehensive_decision","score":0.0,"tokens":194258,"cost_usd":0.358618,"elapsed_time":105.42,"gpr":0.125,"tpe":0.729,"ee":0.6364},{"task_id":"task_011_comprehensive_decision_hard_hard005","category":"comprehensive_decision","score":0.0,"tokens":1912418,"cost_usd":3.592138,"elapsed_time":334.1,"gpr":0.5714,"tpe":0.7552,"ee":0.3043},{"task_id":"task_012_comprehensive_decision_hard_hard006","category":"comprehensive_decision","score":0.0,"tokens":821541,"cost_usd":1.457677,"elapsed_time":270.91,"gpr":0.3333,"tpe":0.1216,"ee":0.25},{"task_id":"task_013_comprehensive_decision_hard_hard007","category":"comprehensive_decision","score":0.0,"tokens":619496,"cost_usd":1.506236,"elapsed_time":91.43,"gpr":0.0,"tpe":0.0,"ee":0.875},{"task_id":"task_014_comprehensive_decision_hard_hard008","category":"comprehensive_decision","score":0.0,"tokens":722052,"cost_usd":1.332629,"elapsed_time":223.09,"gpr":0.5,"tpe":0.5542,"ee":0.2917},{"task_id":"task_015_comprehensive_decision_hard_hard009","category":"comprehensive_decision","score":0.0,"tokens":356920,"cost_usd":0.707645,"elapsed_time":149.0,"gpr":0.0,"tpe":0.0,"ee":0.6154},{"task_id":"task_016_comprehensive_decision_hard_hard010","category":"comprehensive_decision","score":0.0,"tokens":272387,"cost_usd":0.567121,"elapsed_time":95.62,"gpr":0.1429,"tpe":1.0,"ee":0.7},{"task_id":"task_017_comprehensive_decision_hard_hard011","category":"comprehensive_decision","score":0.0,"tokens":274310,"cost_usd":0.465155,"elapsed_time":120.43,"gpr":0.5556,"tpe":0.7499,"ee":0.4667},{"task_id":"task_018_comprehensive_decision_hard_hard012","category":"comprehensive_decision","score":0.0,"tokens":1338206,"cost_usd":2.367881,"elapsed_time":299.36,"gpr":0.0,"tpe":0.0,"ee":0.4118},{"task_id":"task_019_comprehensive_decision_hard_hard013","category":"comprehensive_decision","score":0.0,"tokens":977972,"cost_usd":2.127223,"elapsed_time":164.76,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_020_comprehensive_decision_hard_hard014","category":"comprehensive_decision","score":0.0,"tokens":117825,"cost_usd":0.232439,"elapsed_time":61.57,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_021_comprehensive_decision_hard_hard015","category":"comprehensive_decision","score":1.0,"tokens":364130,"cost_usd":0.811932,"elapsed_time":150.22,"gpr":1.0,"tpe":0.5118,"ee":0.5},{"task_id":"task_022_comprehensive_decision_hard_hard016","category":"comprehensive_decision","score":0.0,"tokens":162307,"cost_usd":0.363508,"elapsed_time":61.26,"gpr":0.125,"tpe":1.0,"ee":0.8571},{"task_id":"task_023_comprehensive_decision_hard_hard017","category":"comprehensive_decision","score":0.0,"tokens":401943,"cost_usd":0.946472,"elapsed_time":139.57,"gpr":0.0,"tpe":0.0,"ee":0.8571},{"task_id":"task_024_comprehensive_decision_hard_hard018","category":"comprehensive_decision","score":0.0,"tokens":300070,"cost_usd":0.667307,"elapsed_time":64.73,"gpr":0.2857,"tpe":1.0,"ee":0.875},{"task_id":"task_025_comprehensive_decision_hard_hard019","category":"comprehensive_decision","score":0.6667,"tokens":198796,"cost_usd":0.353321,"elapsed_time":106.74,"gpr":0.7143,"tpe":0.5905,"ee":0.5},{"task_id":"task_026_comprehensive_decision_medium_medium001","category":"comprehensive_decision","score":0.0,"tokens":339342,"cost_usd":0.592531,"elapsed_time":216.03,"gpr":0.25,"tpe":0.5905,"ee":0.2273},{"task_id":"task_027_comprehensive_decision_medium_medium002","category":"comprehensive_decision","score":1.0,"tokens":140133,"cost_usd":0.23267,"elapsed_time":88.7,"gpr":1.0,"tpe":0.89,"ee":0.625},{"task_id":"task_028_comprehensive_decision_medium_medium003","category":"comprehensive_decision","score":0.0,"tokens":770802,"cost_usd":1.56625,"elapsed_time":159.55,"gpr":0.0,"tpe":0.0,"ee":0.3571},{"task_id":"task_029_comprehensive_decision_medium_medium004","category":"comprehensive_decision","score":0.0,"tokens":335529,"cost_usd":0.619175,"elapsed_time":206.11,"gpr":0.125,"tpe":0.3487,"ee":0.25},{"task_id":"task_030_comprehensive_decision_medium_medium005","category":"comprehensive_decision","score":1.0,"tokens":123422,"cost_usd":0.262873,"elapsed_time":109.0,"gpr":1.0,"tpe":0.9,"ee":0.625},{"task_id":"task_031_comprehensive_decision_medium_medium006","category":"comprehensive_decision","score":1.0,"tokens":164264,"cost_usd":0.277118,"elapsed_time":67.72,"gpr":1.0,"tpe":0.6597,"ee":0.4},{"task_id":"task_032_comprehensive_decision_medium_medium007","category":"comprehensive_decision","score":0.0,"tokens":698468,"cost_usd":1.241462,"elapsed_time":217.27,"gpr":0.0,"tpe":0.0,"ee":0.2174},{"task_id":"task_033_comprehensive_decision_medium_medium008","category":"comprehensive_decision","score":1.0,"tokens":1251799,"cost_usd":2.662784,"elapsed_time":233.01,"gpr":1.0,"tpe":0.3371,"ee":0.3125},{"task_id":"task_034_comprehensive_decision_medium_medium009","category":"comprehensive_decision","score":1.0,"tokens":251261,"cost_usd":0.464994,"elapsed_time":145.41,"gpr":1.0,"tpe":0.5013,"ee":0.3571},{"task_id":"task_035_comprehensive_decision_medium_medium010","category":"comprehensive_decision","score":0.0,"tokens":92098,"cost_usd":0.103529,"elapsed_time":49.11,"gpr":1.0,"tpe":0.81,"ee":0.625},{"task_id":"task_036_comprehensive_decision_medium_medium011","category":"comprehensive_decision","score":0.0,"tokens":487463,"cost_usd":0.908042,"elapsed_time":116.89,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_037_comprehensive_decision_medium_medium012","category":"comprehensive_decision","score":0.0,"tokens":353216,"cost_usd":0.627019,"elapsed_time":183.28,"gpr":0.25,"tpe":0.5314,"ee":0.2941},{"task_id":"task_038_comprehensive_decision_medium_medium013","category":"comprehensive_decision","score":0.5,"tokens":178745,"cost_usd":0.360325,"elapsed_time":96.27,"gpr":0.4,"tpe":0.6302,"ee":0.4167},{"task_id":"task_039_comprehensive_decision_medium_medium014","category":"comprehensive_decision","score":0.5,"tokens":435821,"cost_usd":0.848085,"elapsed_time":165.2,"gpr":0.5,"tpe":0.6561,"ee":0.3125},{"task_id":"task_040_comprehensive_decision_medium_medium015","category":"comprehensive_decision","score":1.0,"tokens":409847,"cost_usd":0.951812,"elapsed_time":171.99,"gpr":1.0,"tpe":0.5561,"ee":0.3571},{"task_id":"task_041_comprehensive_decision_medium_medium016","category":"comprehensive_decision","score":1.0,"tokens":167849,"cost_usd":0.320076,"elapsed_time":113.69,"gpr":1.0,"tpe":0.5905,"ee":0.5},{"task_id":"task_042_comprehensive_decision_medium_medium017","category":"comprehensive_decision","score":1.0,"tokens":149569,"cost_usd":0.216694,"elapsed_time":97.97,"gpr":1.0,"tpe":0.5314,"ee":0.4545},{"task_id":"task_043_comprehensive_decision_medium_medium018","category":"comprehensive_decision","score":1.0,"tokens":361508,"cost_usd":0.830946,"elapsed_time":89.01,"gpr":1.0,"tpe":0.6561,"ee":0.5556},{"task_id":"task_044_comprehensive_decision_medium_medium019","category":"comprehensive_decision","score":1.0,"tokens":275160,"cost_usd":0.441419,"elapsed_time":114.91,"gpr":1.0,"tpe":0.3874,"ee":0.3571},{"task_id":"task_045_comprehensive_decision_medium_medium020","category":"comprehensive_decision","score":0.0,"tokens":285528,"cost_usd":0.549021,"elapsed_time":113.75,"gpr":0.0,"tpe":0.0,"ee":0.4167},{"task_id":"task_046_comprehensive_decision_medium_medium021","category":"comprehensive_decision","score":0.0,"tokens":226724,"cost_usd":0.388934,"elapsed_time":110.95,"gpr":0.0,"tpe":0.0,"ee":0.5556},{"task_id":"task_047_comprehensive_decision_medium_medium022","category":"comprehensive_decision","score":0.0,"tokens":684714,"cost_usd":1.265652,"elapsed_time":105.07,"gpr":0.0,"tpe":0.0,"ee":0.5556},{"task_id":"task_048_comprehensive_decision_medium_medium023","category":"comprehensive_decision","score":1.0,"tokens":838476,"cost_usd":1.797643,"elapsed_time":121.7,"gpr":1.0,"tpe":0.5477,"ee":0.3636},{"task_id":"task_049_comprehensive_decision_medium_medium024","category":"comprehensive_decision","score":0.0,"tokens":1352242,"cost_usd":3.115375,"elapsed_time":212.08,"gpr":0.0,"tpe":0.0,"ee":0.4167},{"task_id":"task_050_comprehensive_decision_medium_medium025","category":"comprehensive_decision","score":0.0,"tokens":794638,"cost_usd":1.67534,"elapsed_time":140.4,"gpr":0.0,"tpe":0.0,"ee":0.3571},{"task_id":"task_051_comprehensive_decision_medium_medium026","category":"comprehensive_decision","score":1.0,"tokens":279681,"cost_usd":0.656768,"elapsed_time":66.79,"gpr":1.0,"tpe":0.729,"ee":0.625},{"task_id":"task_052_comprehensive_decision_medium_medium027","category":"comprehensive_decision","score":1.0,"tokens":185983,"cost_usd":0.340534,"elapsed_time":76.43,"gpr":1.0,"tpe":0.5905,"ee":0.5},{"task_id":"task_053_comprehensive_decision_medium_medium028","category":"comprehensive_decision","score":0.0,"tokens":251990,"cost_usd":0.453476,"elapsed_time":146.25,"gpr":0.0,"tpe":0.0,"ee":0.4167},{"task_id":"task_054_comprehensive_decision_medium_medium029","category":"comprehensive_decision","score":0.0,"tokens":156336,"cost_usd":0.334258,"elapsed_time":54.34,"gpr":0.5,"tpe":1.0,"ee":0.7143},{"task_id":"task_055_comprehensive_decision_medium_medium030","category":"comprehensive_decision","score":0.0,"tokens":1165943,"cost_usd":2.419591,"elapsed_time":142.47,"gpr":0.0,"tpe":0.0,"ee":0.3571},{"task_id":"task_056_comprehensive_decision_medium_medium031","category":"comprehensive_decision","score":1.0,"tokens":198258,"cost_usd":0.350869,"elapsed_time":98.73,"gpr":1.0,"tpe":0.5968,"ee":0.3636},{"task_id":"task_057_comprehensive_decision_medium_medium032","category":"comprehensive_decision","score":1.0,"tokens":298700,"cost_usd":0.555188,"elapsed_time":100.84,"gpr":1.0,"tpe":0.6518,"ee":0.4545},{"task_id":"task_058_comprehensive_decision_medium_medium033","category":"comprehensive_decision","score":1.0,"tokens":171995,"cost_usd":0.300066,"elapsed_time":81.27,"gpr":1.0,"tpe":0.7968,"ee":0.5556},{"task_id":"task_059_comprehensive_decision_medium_medium034","category":"comprehensive_decision","score":0.0,"tokens":992389,"cost_usd":2.224797,"elapsed_time":144.22,"gpr":0.0,"tpe":0.0,"ee":0.4545},{"task_id":"task_060_comprehensive_decision_medium_medium035","category":"comprehensive_decision","score":0.0,"tokens":675383,"cost_usd":1.63335,"elapsed_time":105.73,"gpr":0.0,"tpe":0.0,"ee":0.4167},{"task_id":"task_061_comprehensive_decision_medium_medium036","category":"comprehensive_decision","score":1.0,"tokens":260628,"cost_usd":0.548153,"elapsed_time":111.86,"gpr":1.0,"tpe":0.5511,"ee":0.4545},{"task_id":"task_062_comprehensive_decision_medium_medium037","category":"comprehensive_decision","score":1.0,"tokens":160789,"cost_usd":0.26851,"elapsed_time":80.03,"gpr":1.0,"tpe":0.729,"ee":0.5556},{"task_id":"task_063_comprehensive_decision_medium_medium038","category":"comprehensive_decision","score":0.0,"tokens":240733,"cost_usd":0.564225,"elapsed_time":69.49,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_064_comprehensive_decision_medium_medium039","category":"comprehensive_decision","score":0.0,"tokens":141557,"cost_usd":0.279362,"elapsed_time":151.09,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_065_comprehensive_decision_medium_medium040","category":"comprehensive_decision","score":0.0,"tokens":210912,"cost_usd":0.354143,"elapsed_time":140.42,"gpr":0.3333,"tpe":0.81,"ee":0.4167},{"task_id":"task_066_comprehensive_decision_medium_medium041","category":"comprehensive_decision","score":0.0,"tokens":272206,"cost_usd":0.596152,"elapsed_time":127.16,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_067_comprehensive_decision_medium_medium042","category":"comprehensive_decision","score":0.0,"tokens":407659,"cost_usd":0.784991,"elapsed_time":157.25,"gpr":0.0,"tpe":0.0,"ee":0.4167},{"task_id":"task_068_comprehensive_decision_medium_medium043","category":"comprehensive_decision","score":0.0,"tokens":92633,"cost_usd":0.168939,"elapsed_time":87.77,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_069_comprehensive_decision_medium_medium044","category":"comprehensive_decision","score":0.0,"tokens":528398,"cost_usd":0.861223,"elapsed_time":80.29,"gpr":0.0,"tpe":0.0,"ee":0.7143},{"task_id":"task_070_comprehensive_decision_medium_medium045","category":"comprehensive_decision","score":1.0,"tokens":365326,"cost_usd":0.518239,"elapsed_time":147.33,"gpr":1.0,"tpe":0.2478,"ee":0.2632},{"task_id":"task_071_enterprise_industry_analysis_easy_easy001","category":"enterprise_industry_analysis","score":0.0,"tokens":1307082,"cost_usd":3.086589,"elapsed_time":186.51,"gpr":0.25,"tpe":0.6561,"ee":0.2143},{"task_id":"task_072_enterprise_industry_analysis_easy_easy002","category":"enterprise_industry_analysis","score":0.0,"tokens":276785,"cost_usd":0.458948,"elapsed_time":126.29,"gpr":0.5,"tpe":0.3874,"ee":0.2308},{"task_id":"task_073_enterprise_industry_analysis_easy_easy003","category":"enterprise_industry_analysis","score":0.0,"tokens":162654,"cost_usd":0.264632,"elapsed_time":116.78,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_074_enterprise_industry_analysis_easy_easy004","category":"enterprise_industry_analysis","score":0.0,"tokens":106277,"cost_usd":0.177025,"elapsed_time":45.08,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_075_enterprise_industry_analysis_easy_easy005","category":"enterprise_industry_analysis","score":0.0,"tokens":701440,"cost_usd":1.641702,"elapsed_time":121.3,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_076_enterprise_industry_analysis_easy_easy006","category":"enterprise_industry_analysis","score":0.0,"tokens":90383,"cost_usd":0.13033,"elapsed_time":72.62,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_077_enterprise_industry_analysis_easy_easy007","category":"enterprise_industry_analysis","score":0.0,"tokens":143735,"cost_usd":0.343195,"elapsed_time":62.08,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_078_enterprise_industry_analysis_easy_easy008","category":"enterprise_industry_analysis","score":0.0,"tokens":1025137,"cost_usd":1.998776,"elapsed_time":131.45,"gpr":0.0,"tpe":0.0,"ee":0.1875},{"task_id":"task_079_enterprise_industry_analysis_easy_easy009","category":"enterprise_industry_analysis","score":0.0,"tokens":118971,"cost_usd":0.275944,"elapsed_time":56.44,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_080_enterprise_industry_analysis_easy_easy010","category":"enterprise_industry_analysis","score":0.0,"tokens":354055,"cost_usd":0.842231,"elapsed_time":72.66,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_081_enterprise_industry_analysis_easy_easy011","category":"enterprise_industry_analysis","score":0.0,"tokens":848412,"cost_usd":1.470692,"elapsed_time":113.65,"gpr":0.0,"tpe":0.0,"ee":0.2143},{"task_id":"task_082_enterprise_industry_analysis_easy_easy012","category":"enterprise_industry_analysis","score":0.0,"tokens":489222,"cost_usd":1.043162,"elapsed_time":98.75,"gpr":0.0,"tpe":0.0,"ee":0.1765},{"task_id":"task_083_enterprise_industry_analysis_easy_easy013","category":"enterprise_industry_analysis","score":1.0,"tokens":163152,"cost_usd":0.27783,"elapsed_time":78.16,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_084_enterprise_industry_analysis_easy_easy014","category":"enterprise_industry_analysis","score":0.0,"tokens":73036,"cost_usd":0.114855,"elapsed_time":46.55,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_085_enterprise_industry_analysis_easy_easy015","category":"enterprise_industry_analysis","score":0.0,"tokens":162866,"cost_usd":0.167792,"elapsed_time":61.72,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_086_enterprise_industry_analysis_easy_easy016","category":"enterprise_industry_analysis","score":0.0,"tokens":626390,"cost_usd":1.47906,"elapsed_time":110.72,"gpr":0.3333,"tpe":0.5905,"ee":0.2308},{"task_id":"task_087_enterprise_industry_analysis_easy_easy017","category":"enterprise_industry_analysis","score":0.0,"tokens":265149,"cost_usd":0.565409,"elapsed_time":137.52,"gpr":0.3333,"tpe":0.6561,"ee":0.2143},{"task_id":"task_088_enterprise_industry_analysis_easy_easy018","category":"enterprise_industry_analysis","score":0.0,"tokens":82709,"cost_usd":0.110649,"elapsed_time":53.32,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_089_enterprise_industry_analysis_easy_easy019","category":"enterprise_industry_analysis","score":0.0,"tokens":132469,"cost_usd":0.240435,"elapsed_time":70.1,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_090_enterprise_industry_analysis_easy_easy020","category":"enterprise_industry_analysis","score":0.0,"tokens":100001,"cost_usd":0.151555,"elapsed_time":76.05,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_091_enterprise_industry_analysis_easy_easy021","category":"enterprise_industry_analysis","score":0.0,"tokens":198745,"cost_usd":0.262961,"elapsed_time":80.75,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_092_enterprise_industry_analysis_easy_easy022","category":"enterprise_industry_analysis","score":0.0,"tokens":84099,"cost_usd":0.100612,"elapsed_time":49.2,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_093_enterprise_industry_analysis_easy_easy023","category":"enterprise_industry_analysis","score":1.0,"tokens":270634,"cost_usd":0.509753,"elapsed_time":124.76,"gpr":1.0,"tpe":0.3138,"ee":0.2143},{"task_id":"task_094_enterprise_industry_analysis_easy_easy024","category":"enterprise_industry_analysis","score":0.0,"tokens":103808,"cost_usd":0.201005,"elapsed_time":47.59,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_095_enterprise_industry_analysis_easy_easy025","category":"enterprise_industry_analysis","score":0.0,"tokens":150439,"cost_usd":0.281174,"elapsed_time":100.05,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_096_enterprise_industry_analysis_easy_easy026","category":"enterprise_industry_analysis","score":0.0,"tokens":50223,"cost_usd":0.069058,"elapsed_time":51.52,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_097_enterprise_industry_analysis_easy_easy027","category":"enterprise_industry_analysis","score":0.0,"tokens":301025,"cost_usd":0.545928,"elapsed_time":110.03,"gpr":0.0,"tpe":0.0,"ee":0.1667},{"task_id":"task_098_enterprise_industry_analysis_easy_easy028","category":"enterprise_industry_analysis","score":0.0,"tokens":89347,"cost_usd":0.186438,"elapsed_time":43.08,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_099_enterprise_industry_analysis_easy_easy029","category":"enterprise_industry_analysis","score":0.0,"tokens":130122,"cost_usd":0.257785,"elapsed_time":46.02,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_100_enterprise_industry_analysis_easy_easy030","category":"enterprise_industry_analysis","score":0.0,"tokens":60965,"cost_usd":0.072231,"elapsed_time":45.84,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_101_enterprise_industry_analysis_easy_easy031","category":"enterprise_industry_analysis","score":0.0,"tokens":326489,"cost_usd":0.463217,"elapsed_time":162.08,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_102_enterprise_industry_analysis_easy_easy032","category":"enterprise_industry_analysis","score":0.0,"tokens":218068,"cost_usd":0.341025,"elapsed_time":130.21,"gpr":0.3333,"tpe":0.4305,"ee":0.2308},{"task_id":"task_103_enterprise_industry_analysis_easy_easy033","category":"enterprise_industry_analysis","score":0.0,"tokens":222767,"cost_usd":0.241693,"elapsed_time":151.0,"gpr":0.0,"tpe":0.0,"ee":0.1875},{"task_id":"task_104_enterprise_industry_analysis_easy_easy034","category":"enterprise_industry_analysis","score":0.0,"tokens":255695,"cost_usd":0.534382,"elapsed_time":145.47,"gpr":0.0,"tpe":0.0,"ee":0.1875},{"task_id":"task_105_enterprise_industry_analysis_easy_easy035","category":"enterprise_industry_analysis","score":0.0,"tokens":50951,"cost_usd":0.11535,"elapsed_time":56.96,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_106_enterprise_industry_analysis_easy_easy036","category":"enterprise_industry_analysis","score":0.0,"tokens":114782,"cost_usd":0.147722,"elapsed_time":74.56,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_107_enterprise_industry_analysis_easy_easy037","category":"enterprise_industry_analysis","score":0.0,"tokens":72286,"cost_usd":0.103663,"elapsed_time":46.64,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_108_enterprise_industry_analysis_easy_easy038","category":"enterprise_industry_analysis","score":0.0,"tokens":76706,"cost_usd":0.122906,"elapsed_time":42.66,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_109_enterprise_industry_analysis_easy_easy039","category":"enterprise_industry_analysis","score":1.0,"tokens":174354,"cost_usd":0.321168,"elapsed_time":80.32,"gpr":1.0,"tpe":0.4305,"ee":0.2727},{"task_id":"task_110_enterprise_industry_analysis_easy_easy040","category":"enterprise_industry_analysis","score":0.0,"tokens":50641,"cost_usd":0.091088,"elapsed_time":35.1,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_111_enterprise_industry_analysis_easy_easy041","category":"enterprise_industry_analysis","score":0.0,"tokens":69496,"cost_usd":0.127732,"elapsed_time":60.83,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_112_enterprise_industry_analysis_easy_easy042","category":"enterprise_industry_analysis","score":0.0,"tokens":112064,"cost_usd":0.19557,"elapsed_time":73.2,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_113_enterprise_industry_analysis_easy_easy043","category":"enterprise_industry_analysis","score":0.0,"tokens":69606,"cost_usd":0.104078,"elapsed_time":41.51,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_114_enterprise_industry_analysis_easy_easy044","category":"enterprise_industry_analysis","score":0.0,"tokens":50442,"cost_usd":0.068489,"elapsed_time":33.95,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_115_enterprise_industry_analysis_easy_easy045","category":"enterprise_industry_analysis","score":0.0,"tokens":50523,"cost_usd":0.101099,"elapsed_time":36.02,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_116_enterprise_industry_analysis_easy_easy046","category":"enterprise_industry_analysis","score":0.0,"tokens":126418,"cost_usd":0.255211,"elapsed_time":71.75,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_117_enterprise_industry_analysis_easy_easy047","category":"enterprise_industry_analysis","score":0.0,"tokens":50480,"cost_usd":0.090123,"elapsed_time":35.23,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_118_enterprise_industry_analysis_easy_easy048","category":"enterprise_industry_analysis","score":0.0,"tokens":103625,"cost_usd":0.179175,"elapsed_time":58.06,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_119_enterprise_industry_analysis_easy_easy049","category":"enterprise_industry_analysis","score":0.0,"tokens":437847,"cost_usd":0.928406,"elapsed_time":122.78,"gpr":0.0,"tpe":0.0,"ee":0.2308},{"task_id":"task_120_enterprise_industry_analysis_easy_easy050","category":"enterprise_industry_analysis","score":0.0,"tokens":51440,"cost_usd":0.079062,"elapsed_time":40.14,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_121_enterprise_industry_analysis_easy_easy051","category":"enterprise_industry_analysis","score":0.0,"tokens":378770,"cost_usd":0.873308,"elapsed_time":85.62,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_122_enterprise_industry_analysis_easy_easy052","category":"enterprise_industry_analysis","score":0.0,"tokens":63985,"cost_usd":0.147572,"elapsed_time":56.03,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_123_enterprise_industry_analysis_easy_easy053","category":"enterprise_industry_analysis","score":0.0,"tokens":62999,"cost_usd":0.122208,"elapsed_time":40.02,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_124_enterprise_industry_analysis_easy_easy054","category":"enterprise_industry_analysis","score":0.0,"tokens":63992,"cost_usd":0.126191,"elapsed_time":43.53,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_125_enterprise_industry_analysis_easy_easy055","category":"enterprise_industry_analysis","score":1.0,"tokens":169934,"cost_usd":0.313689,"elapsed_time":75.49,"gpr":1.0,"tpe":0.5314,"ee":0.3333},{"task_id":"task_126_enterprise_industry_analysis_easy_easy056","category":"enterprise_industry_analysis","score":0.0,"tokens":91846,"cost_usd":0.168035,"elapsed_time":60.42,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_127_enterprise_industry_analysis_easy_easy057","category":"enterprise_industry_analysis","score":0.0,"tokens":118401,"cost_usd":0.232743,"elapsed_time":65.01,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_128_enterprise_industry_analysis_easy_easy058","category":"enterprise_industry_analysis","score":0.0,"tokens":59446,"cost_usd":0.102545,"elapsed_time":51.72,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_129_enterprise_industry_analysis_easy_easy059","category":"enterprise_industry_analysis","score":0.0,"tokens":349863,"cost_usd":0.487297,"elapsed_time":45.91,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_130_enterprise_industry_analysis_easy_easy060","category":"enterprise_industry_analysis","score":0.0,"tokens":242405,"cost_usd":0.570973,"elapsed_time":84.89,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_131_enterprise_industry_analysis_easy_easy061","category":"enterprise_industry_analysis","score":1.0,"tokens":524881,"cost_usd":1.183169,"elapsed_time":165.31,"gpr":1.0,"tpe":0.3874,"ee":0.25},{"task_id":"task_132_enterprise_industry_analysis_easy_easy062","category":"enterprise_industry_analysis","score":0.0,"tokens":115518,"cost_usd":0.228082,"elapsed_time":80.52,"gpr":0.6667,"tpe":0.7695,"ee":0.375},{"task_id":"task_133_enterprise_industry_analysis_easy_easy063","category":"enterprise_industry_analysis","score":0.0,"tokens":202404,"cost_usd":0.420375,"elapsed_time":57.3,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_134_enterprise_industry_analysis_easy_easy064","category":"enterprise_industry_analysis","score":1.0,"tokens":86028,"cost_usd":0.150471,"elapsed_time":61.25,"gpr":1.0,"tpe":0.6561,"ee":0.4286},{"task_id":"task_135_enterprise_industry_analysis_easy_easy065","category":"enterprise_industry_analysis","score":1.0,"tokens":79554,"cost_usd":0.130775,"elapsed_time":41.08,"gpr":1.0,"tpe":0.81,"ee":0.5},{"task_id":"task_136_enterprise_industry_analysis_easy_easy066","category":"enterprise_industry_analysis","score":1.0,"tokens":66934,"cost_usd":0.111119,"elapsed_time":44.41,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_137_enterprise_industry_analysis_easy_easy067","category":"enterprise_industry_analysis","score":0.0,"tokens":100121,"cost_usd":0.191384,"elapsed_time":62.25,"gpr":0.6667,"tpe":0.81,"ee":0.4286},{"task_id":"task_138_enterprise_industry_analysis_easy_easy068","category":"enterprise_industry_analysis","score":0.0,"tokens":236873,"cost_usd":0.528737,"elapsed_time":88.61,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_139_enterprise_industry_analysis_easy_easy069","category":"enterprise_industry_analysis","score":0.0,"tokens":176496,"cost_usd":0.36168,"elapsed_time":75.49,"gpr":1.0,"tpe":0.5314,"ee":0.3},{"task_id":"task_140_enterprise_industry_analysis_easy_easy070","category":"enterprise_industry_analysis","score":0.0,"tokens":110442,"cost_usd":0.186287,"elapsed_time":43.98,"gpr":0.6667,"tpe":0.9,"ee":0.5},{"task_id":"task_141_enterprise_industry_analysis_easy_easy071","category":"enterprise_industry_analysis","score":1.0,"tokens":210496,"cost_usd":0.465799,"elapsed_time":66.24,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_142_enterprise_industry_analysis_easy_easy072","category":"enterprise_industry_analysis","score":1.0,"tokens":205210,"cost_usd":0.390325,"elapsed_time":73.19,"gpr":1.0,"tpe":0.6994,"ee":0.3},{"task_id":"task_143_enterprise_industry_analysis_easy_easy073","category":"enterprise_industry_analysis","score":0.0,"tokens":142884,"cost_usd":0.311561,"elapsed_time":59.89,"gpr":0.3333,"tpe":0.729,"ee":0.4286},{"task_id":"task_144_enterprise_industry_analysis_easy_easy074","category":"enterprise_industry_analysis","score":1.0,"tokens":128030,"cost_usd":0.266526,"elapsed_time":55.54,"gpr":1.0,"tpe":0.5708,"ee":0.3333},{"task_id":"task_145_enterprise_industry_analysis_easy_easy075","category":"enterprise_industry_analysis","score":1.0,"tokens":125505,"cost_usd":0.159588,"elapsed_time":64.48,"gpr":1.0,"tpe":0.4624,"ee":0.2727},{"task_id":"task_146_enterprise_industry_analysis_easy_easy076","category":"enterprise_industry_analysis","score":1.0,"tokens":106228,"cost_usd":0.179371,"elapsed_time":58.04,"gpr":1.0,"tpe":0.5708,"ee":0.3333},{"task_id":"task_147_enterprise_industry_analysis_easy_easy077","category":"enterprise_industry_analysis","score":1.0,"tokens":86856,"cost_usd":0.162704,"elapsed_time":52.18,"gpr":1.0,"tpe":0.783,"ee":0.5},{"task_id":"task_148_enterprise_industry_analysis_easy_easy078","category":"enterprise_industry_analysis","score":1.0,"tokens":86637,"cost_usd":0.160338,"elapsed_time":112.31,"gpr":1.0,"tpe":0.6585,"ee":0.375},{"task_id":"task_149_enterprise_industry_analysis_easy_easy079","category":"enterprise_industry_analysis","score":1.0,"tokens":62092,"cost_usd":0.098586,"elapsed_time":50.76,"gpr":1.0,"tpe":0.783,"ee":0.5},{"task_id":"task_150_enterprise_industry_analysis_easy_easy080","category":"enterprise_industry_analysis","score":1.0,"tokens":73421,"cost_usd":0.090789,"elapsed_time":44.9,"gpr":1.0,"tpe":0.7047,"ee":0.4286},{"task_id":"task_151_enterprise_industry_analysis_easy_easy081","category":"enterprise_industry_analysis","score":1.0,"tokens":116530,"cost_usd":0.190957,"elapsed_time":60.79,"gpr":1.0,"tpe":0.6585,"ee":0.375},{"task_id":"task_152_enterprise_industry_analysis_easy_easy082","category":"enterprise_industry_analysis","score":1.0,"tokens":109041,"cost_usd":0.22004,"elapsed_time":54.84,"gpr":1.0,"tpe":0.783,"ee":0.5},{"task_id":"task_153_enterprise_industry_analysis_easy_easy083","category":"enterprise_industry_analysis","score":1.0,"tokens":68432,"cost_usd":0.099854,"elapsed_time":44.86,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_154_enterprise_industry_analysis_easy_easy084","category":"enterprise_industry_analysis","score":1.0,"tokens":136952,"cost_usd":0.283388,"elapsed_time":70.95,"gpr":1.0,"tpe":0.6804,"ee":0.4286},{"task_id":"task_155_enterprise_industry_analysis_easy_easy085","category":"enterprise_industry_analysis","score":0.0,"tokens":300098,"cost_usd":0.671474,"elapsed_time":91.03,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_156_enterprise_industry_analysis_easy_easy086","category":"enterprise_industry_analysis","score":1.0,"tokens":107199,"cost_usd":0.172039,"elapsed_time":68.38,"gpr":1.0,"tpe":0.5708,"ee":0.3333},{"task_id":"task_157_enterprise_industry_analysis_easy_easy088","category":"enterprise_industry_analysis","score":0.0,"tokens":49689,"cost_usd":0.065768,"elapsed_time":40.32,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_158_enterprise_industry_analysis_easy_easy090","category":"enterprise_industry_analysis","score":0.0,"tokens":114635,"cost_usd":0.204807,"elapsed_time":60.22,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_159_enterprise_industry_analysis_easy_easy091","category":"enterprise_industry_analysis","score":0.0,"tokens":81982,"cost_usd":0.163584,"elapsed_time":63.54,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_160_enterprise_industry_analysis_easy_easy092","category":"enterprise_industry_analysis","score":0.0,"tokens":61363,"cost_usd":0.095465,"elapsed_time":47.5,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_161_enterprise_industry_analysis_easy_easy093","category":"enterprise_industry_analysis","score":0.0,"tokens":58828,"cost_usd":0.076059,"elapsed_time":101.75,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_162_enterprise_industry_analysis_easy_easy094","category":"enterprise_industry_analysis","score":0.0,"tokens":71507,"cost_usd":0.114249,"elapsed_time":50.57,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_163_enterprise_industry_analysis_easy_easy095","category":"enterprise_industry_analysis","score":0.0,"tokens":51818,"cost_usd":0.099218,"elapsed_time":58.39,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_164_enterprise_industry_analysis_easy_easy096","category":"enterprise_industry_analysis","score":0.0,"tokens":59518,"cost_usd":0.067664,"elapsed_time":41.24,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_165_enterprise_industry_analysis_easy_easy097","category":"enterprise_industry_analysis","score":0.0,"tokens":28115,"cost_usd":0.049423,"elapsed_time":22.53,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_166_enterprise_industry_analysis_easy_easy098","category":"enterprise_industry_analysis","score":0.0,"tokens":39841,"cost_usd":0.074981,"elapsed_time":33.4,"gpr":0.0,"tpe":0.0,"ee":0.75},{"task_id":"task_167_enterprise_industry_analysis_easy_easy099","category":"enterprise_industry_analysis","score":0.0,"tokens":262112,"cost_usd":0.607653,"elapsed_time":119.69,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_168_enterprise_industry_analysis_easy_easy100","category":"enterprise_industry_analysis","score":0.0,"tokens":188411,"cost_usd":0.344262,"elapsed_time":67.07,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_169_enterprise_industry_analysis_easy_easy101","category":"enterprise_industry_analysis","score":0.0,"tokens":90664,"cost_usd":0.135844,"elapsed_time":73.24,"gpr":0.0,"tpe":0.0,"ee":0.375},{"task_id":"task_170_enterprise_industry_analysis_easy_easy102","category":"enterprise_industry_analysis","score":0.0,"tokens":108080,"cost_usd":0.184515,"elapsed_time":61.09,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_171_enterprise_industry_analysis_easy_easy103","category":"enterprise_industry_analysis","score":0.0,"tokens":62152,"cost_usd":0.086344,"elapsed_time":42.46,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_172_enterprise_industry_analysis_easy_easy104","category":"enterprise_industry_analysis","score":0.0,"tokens":58720,"cost_usd":0.089601,"elapsed_time":37.25,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_173_enterprise_industry_analysis_easy_easy105","category":"enterprise_industry_analysis","score":0.0,"tokens":51899,"cost_usd":0.084265,"elapsed_time":40.32,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_174_enterprise_industry_analysis_easy_easy106","category":"enterprise_industry_analysis","score":0.0,"tokens":63591,"cost_usd":0.095409,"elapsed_time":47.0,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_175_enterprise_industry_analysis_easy_easy107","category":"enterprise_industry_analysis","score":0.0,"tokens":59044,"cost_usd":0.08888,"elapsed_time":37.8,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_176_enterprise_industry_analysis_easy_easy108","category":"enterprise_industry_analysis","score":0.0,"tokens":62726,"cost_usd":0.078113,"elapsed_time":43.45,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_177_enterprise_industry_analysis_easy_easy109","category":"enterprise_industry_analysis","score":0.0,"tokens":164437,"cost_usd":0.352755,"elapsed_time":44.71,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_178_enterprise_industry_analysis_easy_easy110","category":"enterprise_industry_analysis","score":0.0,"tokens":51635,"cost_usd":0.084854,"elapsed_time":61.04,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_179_enterprise_industry_analysis_easy_easy111","category":"enterprise_industry_analysis","score":0.0,"tokens":51812,"cost_usd":0.081294,"elapsed_time":36.69,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_180_enterprise_industry_analysis_easy_easy112","category":"enterprise_industry_analysis","score":0.0,"tokens":48792,"cost_usd":0.084491,"elapsed_time":33.67,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_181_enterprise_industry_analysis_easy_easy113","category":"enterprise_industry_analysis","score":0.0,"tokens":84698,"cost_usd":0.176498,"elapsed_time":55.62,"gpr":0.0,"tpe":0.0,"ee":0.4286},{"task_id":"task_182_enterprise_industry_analysis_easy_easy114","category":"enterprise_industry_analysis","score":0.0,"tokens":63236,"cost_usd":0.128516,"elapsed_time":55.4,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_183_enterprise_industry_analysis_easy_easy115","category":"enterprise_industry_analysis","score":0.0,"tokens":79945,"cost_usd":0.153467,"elapsed_time":42.94,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_184_enterprise_industry_analysis_easy_easy116","category":"enterprise_industry_analysis","score":0.0,"tokens":263881,"cost_usd":0.582108,"elapsed_time":67.07,"gpr":0.0,"tpe":0.0,"ee":0.3},{"task_id":"task_185_enterprise_industry_analysis_easy_easy117","category":"enterprise_industry_analysis","score":0.0,"tokens":59454,"cost_usd":0.079715,"elapsed_time":36.97,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_186_enterprise_industry_analysis_medium_medium001","category":"enterprise_industry_analysis","score":1.0,"tokens":1146913,"cost_usd":2.269264,"elapsed_time":206.55,"gpr":1.0,"tpe":0.2612,"ee":0.2353},{"task_id":"task_187_enterprise_industry_analysis_medium_medium002","category":"enterprise_industry_analysis","score":0.0,"tokens":1371909,"cost_usd":2.684709,"elapsed_time":257.21,"gpr":0.0,"tpe":0.0,"ee":0.1739},{"task_id":"task_188_enterprise_industry_analysis_medium_medium003","category":"enterprise_industry_analysis","score":0.0,"tokens":341618,"cost_usd":0.681285,"elapsed_time":214.61,"gpr":0.0,"tpe":0.0,"ee":0.2},{"task_id":"task_189_enterprise_industry_analysis_medium_medium004","category":"enterprise_industry_analysis","score":0.0,"tokens":378319,"cost_usd":0.791579,"elapsed_time":177.58,"gpr":0.0,"tpe":0.0,"ee":0.2222},{"task_id":"task_190_enterprise_industry_analysis_medium_medium005","category":"enterprise_industry_analysis","score":0.0,"tokens":127730,"cost_usd":0.189426,"elapsed_time":97.58,"gpr":0.0,"tpe":0.0,"ee":0.3636},{"task_id":"task_191_enterprise_industry_analysis_medium_medium006","category":"enterprise_industry_analysis","score":0.0,"tokens":314305,"cost_usd":0.60463,"elapsed_time":109.49,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_192_enterprise_industry_analysis_medium_medium007","category":"enterprise_industry_analysis","score":0.0,"tokens":315094,"cost_usd":0.572367,"elapsed_time":126.91,"gpr":0.0,"tpe":0.0,"ee":0.25},{"task_id":"task_193_enterprise_industry_analysis_medium_medium008","category":"enterprise_industry_analysis","score":1.0,"tokens":847784,"cost_usd":2.048253,"elapsed_time":122.46,"gpr":1.0,"tpe":0.5139,"ee":0.3636},{"task_id":"task_194_enterprise_industry_analysis_medium_medium009","category":"enterprise_industry_analysis","score":0.0,"tokens":177623,"cost_usd":0.290231,"elapsed_time":87.62,"gpr":0.2,"tpe":0.5905,"ee":0.2857},{"task_id":"task_195_enterprise_industry_analysis_medium_medium010","category":"enterprise_industry_analysis","score":0.0,"tokens":147741,"cost_usd":0.25286,"elapsed_time":85.78,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_196_enterprise_industry_analysis_medium_medium011","category":"enterprise_industry_analysis","score":0.0,"tokens":191858,"cost_usd":0.379564,"elapsed_time":65.9,"gpr":0.0,"tpe":0.0,"ee":0.3636},{"task_id":"task_197_enterprise_industry_analysis_medium_medium012","category":"enterprise_industry_analysis","score":0.0,"tokens":233725,"cost_usd":0.434389,"elapsed_time":147.55,"gpr":0.0,"tpe":0.0,"ee":0.2353},{"task_id":"task_198_enterprise_industry_analysis_medium_medium013","category":"enterprise_industry_analysis","score":0.0,"tokens":96681,"cost_usd":0.131973,"elapsed_time":56.35,"gpr":0.0,"tpe":0.0,"ee":0.4444},{"task_id":"task_199_enterprise_industry_analysis_medium_medium014","category":"enterprise_industry_analysis","score":1.0,"tokens":208652,"cost_usd":0.43855,"elapsed_time":68.96,"gpr":1.0,"tpe":0.7468,"ee":0.5556},{"task_id":"task_200_enterprise_industry_analysis_medium_medium015","category":"enterprise_industry_analysis","score":0.0,"tokens":234828,"cost_usd":0.409206,"elapsed_time":105.54,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_201_enterprise_industry_analysis_medium_medium016","category":"enterprise_industry_analysis","score":0.0,"tokens":1197147,"cost_usd":2.591285,"elapsed_time":183.93,"gpr":0.2,"tpe":0.6561,"ee":0.2941},{"task_id":"task_202_enterprise_industry_analysis_medium_medium017","category":"enterprise_industry_analysis","score":0.0,"tokens":1364888,"cost_usd":2.629278,"elapsed_time":217.7,"gpr":0.0,"tpe":0.0,"ee":0.2632},{"task_id":"task_203_enterprise_industry_analysis_medium_medium018","category":"enterprise_industry_analysis","score":0.0,"tokens":439287,"cost_usd":1.052916,"elapsed_time":147.9,"gpr":0.6,"tpe":0.9333,"ee":0.7143},{"task_id":"task_204_enterprise_industry_analysis_medium_medium019","category":"enterprise_industry_analysis","score":0.0,"tokens":161079,"cost_usd":0.257902,"elapsed_time":98.42,"gpr":0.4,"tpe":0.6561,"ee":0.5},{"task_id":"task_205_enterprise_industry_analysis_medium_medium020","category":"enterprise_industry_analysis","score":0.0,"tokens":151501,"cost_usd":0.316442,"elapsed_time":66.67,"gpr":0.4,"tpe":0.729,"ee":0.5556},{"task_id":"task_206_enterprise_industry_analysis_medium_medium021","category":"enterprise_industry_analysis","score":0.0,"tokens":251524,"cost_usd":0.430757,"elapsed_time":86.41,"gpr":0.2,"tpe":0.6561,"ee":0.5},{"task_id":"task_207_enterprise_industry_analysis_medium_medium022","category":"enterprise_industry_analysis","score":0.0,"tokens":177340,"cost_usd":0.292776,"elapsed_time":92.12,"gpr":0.4,"tpe":0.9,"ee":0.4545},{"task_id":"task_208_enterprise_industry_analysis_medium_medium023","category":"enterprise_industry_analysis","score":0.0,"tokens":226679,"cost_usd":0.371479,"elapsed_time":111.86,"gpr":0.2,"tpe":0.5905,"ee":0.3846},{"task_id":"task_209_enterprise_industry_analysis_medium_medium024","category":"enterprise_industry_analysis","score":0.0,"tokens":338691,"cost_usd":0.632192,"elapsed_time":126.53,"gpr":0.0,"tpe":0.0,"ee":0.2941},{"task_id":"task_210_enterprise_industry_analysis_medium_medium025","category":"enterprise_industry_analysis","score":0.0,"tokens":137194,"cost_usd":0.237456,"elapsed_time":68.57,"gpr":0.0,"tpe":0.0,"ee":0.4545},{"task_id":"task_211_enterprise_industry_analysis_medium_medium026","category":"enterprise_industry_analysis","score":0.0,"tokens":293869,"cost_usd":0.647044,"elapsed_time":77.92,"gpr":0.4,"tpe":0.729,"ee":0.5},{"task_id":"task_212_enterprise_industry_analysis_medium_medium027","category":"enterprise_industry_analysis","score":0.0,"tokens":81531,"cost_usd":0.148013,"elapsed_time":83.88,"gpr":0.0,"tpe":0.0,"ee":0.8333},{"task_id":"task_213_enterprise_industry_analysis_medium_medium028","category":"enterprise_industry_analysis","score":1.0,"tokens":198930,"cost_usd":0.335811,"elapsed_time":104.62,"gpr":1.0,"tpe":0.5208,"ee":0.4167},{"task_id":"task_214_enterprise_industry_analysis_medium_medium029","category":"enterprise_industry_analysis","score":0.0,"tokens":105864,"cost_usd":0.194334,"elapsed_time":60.08,"gpr":0.4,"tpe":0.9,"ee":0.7143},{"task_id":"task_215_enterprise_industry_analysis_medium_medium030","category":"enterprise_industry_analysis","score":1.0,"tokens":283322,"cost_usd":0.634694,"elapsed_time":94.31,"gpr":1.0,"tpe":0.4783,"ee":0.4167},{"task_id":"task_216_enterprise_industry_analysis_medium_medium031","category":"enterprise_industry_analysis","score":0.0,"tokens":204640,"cost_usd":0.427543,"elapsed_time":116.05,"gpr":0.8,"tpe":1.0,"ee":0.3846},{"task_id":"task_217_enterprise_industry_analysis_medium_medium032","category":"enterprise_industry_analysis","score":0.0,"tokens":144680,"cost_usd":0.276596,"elapsed_time":89.16,"gpr":0.8,"tpe":0.6561,"ee":0.5},{"task_id":"task_218_enterprise_industry_analysis_medium_medium033","category":"enterprise_industry_analysis","score":0.0,"tokens":330024,"cost_usd":0.604429,"elapsed_time":132.21,"gpr":0.4,"tpe":0.2542,"ee":0.2632},{"task_id":"task_219_enterprise_industry_analysis_medium_medium034","category":"enterprise_industry_analysis","score":1.0,"tokens":120590,"cost_usd":0.212763,"elapsed_time":74.38,"gpr":1.0,"tpe":0.729,"ee":0.625},{"task_id":"task_220_enterprise_industry_analysis_medium_medium035","category":"enterprise_industry_analysis","score":0.0,"tokens":155913,"cost_usd":0.296615,"elapsed_time":70.4,"gpr":0.8,"tpe":0.6925,"ee":0.5},{"task_id":"task_221_enterprise_industry_analysis_medium_medium036","category":"enterprise_industry_analysis","score":0.0,"tokens":163988,"cost_usd":0.316014,"elapsed_time":71.26,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_222_enterprise_industry_analysis_medium_medium037","category":"enterprise_industry_analysis","score":1.0,"tokens":143732,"cost_usd":0.258478,"elapsed_time":64.19,"gpr":1.0,"tpe":0.8536,"ee":0.625},{"task_id":"task_223_enterprise_industry_analysis_medium_medium038","category":"enterprise_industry_analysis","score":0.0,"tokens":288150,"cost_usd":0.467285,"elapsed_time":189.09,"gpr":0.0,"tpe":0.0,"ee":0.2941},{"task_id":"task_224_enterprise_industry_analysis_medium_medium039","category":"enterprise_industry_analysis","score":1.0,"tokens":196384,"cost_usd":0.368957,"elapsed_time":78.91,"gpr":1.0,"tpe":0.6789,"ee":0.4545},{"task_id":"task_225_enterprise_industry_analysis_medium_medium040","category":"enterprise_industry_analysis","score":0.0,"tokens":152642,"cost_usd":0.316838,"elapsed_time":102.32,"gpr":0.2,"tpe":0.6561,"ee":0.5},{"task_id":"task_226_enterprise_industry_analysis_medium_medium041","category":"enterprise_industry_analysis","score":0.0,"tokens":125361,"cost_usd":0.247374,"elapsed_time":112.07,"gpr":0.8,"tpe":0.8575,"ee":0.625},{"task_id":"task_227_enterprise_industry_analysis_medium_medium042","category":"enterprise_industry_analysis","score":0.0,"tokens":341388,"cost_usd":0.620319,"elapsed_time":81.61,"gpr":0.8,"tpe":1.0,"ee":0.7143},{"task_id":"task_228_enterprise_industry_analysis_medium_medium043","category":"enterprise_industry_analysis","score":0.0,"tokens":164125,"cost_usd":0.322724,"elapsed_time":54.68,"gpr":0.8,"tpe":0.8373,"ee":0.5556},{"task_id":"task_229_enterprise_industry_analysis_medium_medium044","category":"enterprise_industry_analysis","score":0.0,"tokens":160470,"cost_usd":0.341151,"elapsed_time":76.78,"gpr":0.4,"tpe":0.729,"ee":0.5556},{"task_id":"task_230_enterprise_industry_analysis_medium_medium045","category":"enterprise_industry_analysis","score":0.0,"tokens":257763,"cost_usd":0.534242,"elapsed_time":107.97,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_231_enterprise_industry_analysis_medium_medium046","category":"enterprise_industry_analysis","score":0.0,"tokens":215779,"cost_usd":0.451568,"elapsed_time":122.31,"gpr":0.0,"tpe":0.0,"ee":0.4545},{"task_id":"task_232_enterprise_industry_analysis_medium_medium047","category":"enterprise_industry_analysis","score":1.0,"tokens":206028,"cost_usd":0.350048,"elapsed_time":114.82,"gpr":1.0,"tpe":0.4219,"ee":0.3571},{"task_id":"task_233_enterprise_industry_analysis_medium_medium048","category":"enterprise_industry_analysis","score":0.0,"tokens":157615,"cost_usd":0.286245,"elapsed_time":152.88,"gpr":0.8,"tpe":0.729,"ee":0.5},{"task_id":"task_234_enterprise_industry_analysis_medium_medium049","category":"enterprise_industry_analysis","score":0.0,"tokens":178418,"cost_usd":0.312647,"elapsed_time":92.79,"gpr":0.6,"tpe":0.6561,"ee":0.4545},{"task_id":"task_235_enterprise_industry_analysis_medium_medium050","category":"enterprise_industry_analysis","score":0.0,"tokens":299025,"cost_usd":0.527479,"elapsed_time":104.98,"gpr":0.6,"tpe":0.6563,"ee":0.3846},{"task_id":"task_236_enterprise_industry_analysis_medium_medium051","category":"enterprise_industry_analysis","score":0.0,"tokens":197265,"cost_usd":0.313143,"elapsed_time":98.19,"gpr":0.8,"tpe":0.612,"ee":0.4167},{"task_id":"task_237_enterprise_industry_analysis_medium_medium052","category":"enterprise_industry_analysis","score":1.0,"tokens":264356,"cost_usd":0.55277,"elapsed_time":105.18,"gpr":1.0,"tpe":0.5413,"ee":0.3846},{"task_id":"task_238_enterprise_industry_analysis_medium_medium053","category":"enterprise_industry_analysis","score":0.0,"tokens":135120,"cost_usd":0.261163,"elapsed_time":58.97,"gpr":0.2,"tpe":0.9,"ee":0.625},{"task_id":"task_239_enterprise_industry_analysis_medium_medium054","category":"enterprise_industry_analysis","score":1.0,"tokens":206670,"cost_usd":0.376727,"elapsed_time":76.62,"gpr":1.0,"tpe":0.6986,"ee":0.5},{"task_id":"task_240_enterprise_industry_analysis_medium_medium055","category":"enterprise_industry_analysis","score":0.0,"tokens":590268,"cost_usd":1.385352,"elapsed_time":122.94,"gpr":0.4,"tpe":0.81,"ee":0.4167},{"task_id":"task_241_enterprise_industry_analysis_medium_medium056","category":"enterprise_industry_analysis","score":0.0,"tokens":455596,"cost_usd":1.047957,"elapsed_time":117.73,"gpr":0.0,"tpe":0.0,"ee":0.3846},{"task_id":"task_242_enterprise_industry_analysis_medium_medium057","category":"enterprise_industry_analysis","score":0.0,"tokens":544170,"cost_usd":1.303178,"elapsed_time":118.9,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_243_enterprise_industry_analysis_medium_medium058","category":"enterprise_industry_analysis","score":0.0,"tokens":170313,"cost_usd":0.290844,"elapsed_time":61.28,"gpr":0.2,"tpe":0.729,"ee":0.5},{"task_id":"task_244_enterprise_industry_analysis_medium_medium059","category":"enterprise_industry_analysis","score":0.0,"tokens":148164,"cost_usd":0.265388,"elapsed_time":61.08,"gpr":0.8,"tpe":0.8645,"ee":0.5556},{"task_id":"task_245_enterprise_industry_analysis_medium_medium060","category":"enterprise_industry_analysis","score":0.0,"tokens":601091,"cost_usd":1.402743,"elapsed_time":97.21,"gpr":0.0,"tpe":0.0,"ee":0.4545},{"task_id":"task_246_enterprise_industry_analysis_medium_medium061","category":"enterprise_industry_analysis","score":1.0,"tokens":280029,"cost_usd":0.590758,"elapsed_time":106.42,"gpr":1.0,"tpe":0.4996,"ee":0.4167},{"task_id":"task_247_enterprise_industry_analysis_medium_medium062","category":"enterprise_industry_analysis","score":0.0,"tokens":410672,"cost_usd":0.670057,"elapsed_time":92.28,"gpr":0.0,"tpe":0.0,"ee":0.4167},{"task_id":"task_248_enterprise_industry_analysis_medium_medium063","category":"enterprise_industry_analysis","score":0.0,"tokens":186541,"cost_usd":0.378785,"elapsed_time":80.43,"gpr":0.5,"tpe":0.561,"ee":0.3636},{"task_id":"task_249_enterprise_industry_analysis_medium_medium064","category":"enterprise_industry_analysis","score":0.0,"tokens":191859,"cost_usd":0.338656,"elapsed_time":101.64,"gpr":0.0,"tpe":0.0,"ee":0.2857},{"task_id":"task_250_enterprise_industry_analysis_medium_medium065","category":"enterprise_industry_analysis","score":1.0,"tokens":502316,"cost_usd":1.215007,"elapsed_time":147.83,"gpr":1.0,"tpe":0.6929,"ee":0.4444},{"task_id":"task_251_enterprise_industry_analysis_medium_medium066","category":"enterprise_industry_analysis","score":1.0,"tokens":63500,"cost_usd":0.123435,"elapsed_time":41.49,"gpr":1.0,"tpe":0.81,"ee":0.6667},{"task_id":"task_252_enterprise_industry_analysis_medium_medium067","category":"enterprise_industry_analysis","score":1.0,"tokens":531468,"cost_usd":1.100354,"elapsed_time":129.52,"gpr":1.0,"tpe":0.6115,"ee":0.2857},{"task_id":"task_253_enterprise_industry_analysis_medium_medium068","category":"enterprise_industry_analysis","score":1.0,"tokens":168583,"cost_usd":0.320843,"elapsed_time":96.91,"gpr":1.0,"tpe":0.6679,"ee":0.4},{"task_id":"task_254_enterprise_industry_analysis_medium_medium069","category":"enterprise_industry_analysis","score":1.0,"tokens":426871,"cost_usd":1.05621,"elapsed_time":78.97,"gpr":1.0,"tpe":0.7898,"ee":0.5714},{"task_id":"task_255_enterprise_industry_analysis_medium_medium070","category":"enterprise_industry_analysis","score":0.0,"tokens":196345,"cost_usd":0.288961,"elapsed_time":60.79,"gpr":0.6667,"tpe":0.8145,"ee":0.5},{"task_id":"task_256_enterprise_industry_analysis_medium_medium071","category":"enterprise_industry_analysis","score":0.0,"tokens":123248,"cost_usd":0.258681,"elapsed_time":59.88,"gpr":0.5,"tpe":0.6925,"ee":0.5},{"task_id":"task_257_enterprise_industry_analysis_medium_medium072","category":"enterprise_industry_analysis","score":1.0,"tokens":117447,"cost_usd":0.20429,"elapsed_time":99.95,"gpr":1.0,"tpe":0.5462,"ee":0.4},{"task_id":"task_258_enterprise_industry_analysis_medium_medium073","category":"enterprise_industry_analysis","score":1.0,"tokens":109443,"cost_usd":0.171904,"elapsed_time":56.93,"gpr":1.0,"tpe":0.7968,"ee":0.5714},{"task_id":"task_259_enterprise_industry_analysis_medium_medium074","category":"enterprise_industry_analysis","score":0.0,"tokens":402979,"cost_usd":0.97248,"elapsed_time":52.18,"gpr":0.5,"tpe":1.0,"ee":0.8},{"task_id":"task_260_enterprise_industry_analysis_medium_medium075","category":"enterprise_industry_analysis","score":1.0,"tokens":116937,"cost_usd":0.171245,"elapsed_time":61.61,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_261_enterprise_industry_analysis_medium_medium076","category":"enterprise_industry_analysis","score":0.0,"tokens":698829,"cost_usd":1.704849,"elapsed_time":114.82,"gpr":0.25,"tpe":0.5314,"ee":0.3636},{"task_id":"task_262_enterprise_industry_analysis_medium_medium077","category":"enterprise_industry_analysis","score":1.0,"tokens":193049,"cost_usd":0.397484,"elapsed_time":53.59,"gpr":1.0,"tpe":0.7275,"ee":0.4444},{"task_id":"task_263_enterprise_industry_analysis_medium_medium078","category":"enterprise_industry_analysis","score":0.0,"tokens":217675,"cost_usd":0.403191,"elapsed_time":98.76,"gpr":0.0,"tpe":0.0,"ee":0.3077},{"task_id":"task_264_enterprise_industry_analysis_medium_medium079","category":"enterprise_industry_analysis","score":1.0,"tokens":925259,"cost_usd":2.247167,"elapsed_time":133.54,"gpr":1.0,"tpe":0.4137,"ee":0.2857},{"task_id":"task_265_enterprise_industry_analysis_medium_medium080","category":"enterprise_industry_analysis","score":0.0,"tokens":387708,"cost_usd":0.595248,"elapsed_time":100.1,"gpr":0.0,"tpe":0.0,"ee":0.3636},{"task_id":"task_266_enterprise_industry_analysis_medium_medium081","category":"enterprise_industry_analysis","score":1.0,"tokens":152703,"cost_usd":0.334303,"elapsed_time":69.87,"gpr":1.0,"tpe":0.8325,"ee":0.5714},{"task_id":"task_267_enterprise_industry_analysis_medium_medium082","category":"enterprise_industry_analysis","score":0.0,"tokens":95609,"cost_usd":0.178542,"elapsed_time":56.33,"gpr":0.0,"tpe":0.0,"ee":0.6667},{"task_id":"task_268_enterprise_industry_analysis_medium_medium083","category":"enterprise_industry_analysis","score":0.0,"tokens":357268,"cost_usd":0.864243,"elapsed_time":89.61,"gpr":0.0,"tpe":0.0,"ee":0.6667},{"task_id":"task_269_enterprise_industry_analysis_medium_medium084","category":"enterprise_industry_analysis","score":0.0,"tokens":110006,"cost_usd":0.202112,"elapsed_time":47.17,"gpr":0.25,"tpe":1.0,"ee":0.5714},{"task_id":"task_270_enterprise_industry_analysis_medium_medium085","category":"enterprise_industry_analysis","score":1.0,"tokens":161182,"cost_usd":0.246888,"elapsed_time":62.11,"gpr":1.0,"tpe":0.6441,"ee":0.3636},{"task_id":"task_271_enterprise_industry_analysis_medium_medium086","category":"enterprise_industry_analysis","score":1.0,"tokens":176236,"cost_usd":0.336769,"elapsed_time":70.24,"gpr":1.0,"tpe":0.5672,"ee":0.3636},{"task_id":"task_272_enterprise_industry_analysis_medium_medium087","category":"enterprise_industry_analysis","score":0.0,"tokens":169428,"cost_usd":0.220758,"elapsed_time":59.55,"gpr":0.5,"tpe":0.7695,"ee":0.4},{"task_id":"task_273_enterprise_industry_analysis_medium_medium088","category":"enterprise_industry_analysis","score":1.0,"tokens":372127,"cost_usd":0.780392,"elapsed_time":113.44,"gpr":1.0,"tpe":0.3691,"ee":0.2857},{"task_id":"task_274_enterprise_industry_analysis_medium_medium089","category":"enterprise_industry_analysis","score":1.0,"tokens":248803,"cost_usd":0.529706,"elapsed_time":93.7,"gpr":1.0,"tpe":0.3885,"ee":0.2857},{"task_id":"task_275_enterprise_industry_analysis_medium_medium090","category":"enterprise_industry_analysis","score":0.0,"tokens":380049,"cost_usd":0.742253,"elapsed_time":73.38,"gpr":0.5,"tpe":0.733,"ee":0.4444},{"task_id":"task_276_enterprise_industry_analysis_medium_medium091","category":"enterprise_industry_analysis","score":1.0,"tokens":191798,"cost_usd":0.423522,"elapsed_time":82.47,"gpr":1.0,"tpe":0.8829,"ee":0.4},{"task_id":"task_277_enterprise_industry_analysis_medium_medium092","category":"enterprise_industry_analysis","score":0.0,"tokens":322563,"cost_usd":0.602866,"elapsed_time":100.12,"gpr":0.5,"tpe":0.481,"ee":0.3077},{"task_id":"task_278_enterprise_industry_analysis_medium_medium093","category":"enterprise_industry_analysis","score":1.0,"tokens":634196,"cost_usd":1.548172,"elapsed_time":102.59,"gpr":1.0,"tpe":0.8848,"ee":0.5714},{"task_id":"task_279_enterprise_industry_analysis_medium_medium094","category":"enterprise_industry_analysis","score":0.0,"tokens":281780,"cost_usd":0.54164,"elapsed_time":99.99,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_280_enterprise_industry_analysis_medium_medium095","category":"enterprise_industry_analysis","score":0.0,"tokens":390549,"cost_usd":0.923764,"elapsed_time":88.46,"gpr":0.0,"tpe":0.0,"ee":0.5714},{"task_id":"task_281_enterprise_industry_analysis_medium_medium096","category":"enterprise_industry_analysis","score":1.0,"tokens":118109,"cost_usd":0.231324,"elapsed_time":50.78,"gpr":1.0,"tpe":0.7108,"ee":0.5},{"task_id":"task_282_enterprise_industry_analysis_medium_medium097","category":"enterprise_industry_analysis","score":1.0,"tokens":121816,"cost_usd":0.275726,"elapsed_time":53.1,"gpr":1.0,"tpe":0.6925,"ee":0.5},{"task_id":"task_283_enterprise_industry_analysis_medium_medium098","category":"enterprise_industry_analysis","score":0.0,"tokens":188384,"cost_usd":0.359939,"elapsed_time":71.88,"gpr":0.25,"tpe":0.5314,"ee":0.3636},{"task_id":"task_284_enterprise_industry_analysis_medium_medium099","category":"enterprise_industry_analysis","score":0.0,"tokens":260906,"cost_usd":0.547563,"elapsed_time":92.55,"gpr":0.0,"tpe":0.0,"ee":0.3077},{"task_id":"task_285_enterprise_industry_analysis_medium_medium100","category":"enterprise_industry_analysis","score":0.0,"tokens":394764,"cost_usd":0.945661,"elapsed_time":92.65,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_286_enterprise_industry_analysis_medium_medium101","category":"enterprise_industry_analysis","score":1.0,"tokens":89163,"cost_usd":0.139002,"elapsed_time":57.2,"gpr":1.0,"tpe":0.8645,"ee":0.5},{"task_id":"task_287_enterprise_industry_analysis_medium_medium102","category":"enterprise_industry_analysis","score":0.0,"tokens":108556,"cost_usd":0.221998,"elapsed_time":55.24,"gpr":0.25,"tpe":0.9,"ee":0.5714},{"task_id":"task_288_enterprise_industry_analysis_medium_medium103","category":"enterprise_industry_analysis","score":0.0,"tokens":82719,"cost_usd":0.147101,"elapsed_time":54.61,"gpr":0.5,"tpe":0.905,"ee":0.5714},{"task_id":"task_289_enterprise_industry_analysis_medium_medium104","category":"enterprise_industry_analysis","score":0.0,"tokens":182764,"cost_usd":0.333471,"elapsed_time":68.51,"gpr":0.75,"tpe":0.496,"ee":0.3333},{"task_id":"task_290_enterprise_industry_analysis_medium_medium105","category":"enterprise_industry_analysis","score":0.0,"tokens":277807,"cost_usd":0.542964,"elapsed_time":155.62,"gpr":0.5,"tpe":0.733,"ee":0.4},{"task_id":"task_291_enterprise_industry_analysis_medium_medium106","category":"enterprise_industry_analysis","score":0.0,"tokens":430095,"cost_usd":1.044761,"elapsed_time":73.54,"gpr":0.5,"tpe":1.0,"ee":0.6667},{"task_id":"task_292_enterprise_industry_analysis_medium_medium107","category":"enterprise_industry_analysis","score":0.0,"tokens":145326,"cost_usd":0.278777,"elapsed_time":60.68,"gpr":0.75,"tpe":0.8187,"ee":0.4444},{"task_id":"task_293_enterprise_industry_analysis_medium_medium108","category":"enterprise_industry_analysis","score":1.0,"tokens":103642,"cost_usd":0.23394,"elapsed_time":59.36,"gpr":1.0,"tpe":0.88,"ee":0.6667},{"task_id":"task_294_enterprise_industry_analysis_medium_medium109","category":"enterprise_industry_analysis","score":1.0,"tokens":109322,"cost_usd":0.198845,"elapsed_time":58.81,"gpr":1.0,"tpe":0.7535,"ee":0.5},{"task_id":"task_295_enterprise_industry_analysis_medium_medium110","category":"enterprise_industry_analysis","score":0.0,"tokens":687221,"cost_usd":1.179629,"elapsed_time":100.07,"gpr":0.0,"tpe":0.0,"ee":0.3636},{"task_id":"task_296_enterprise_industry_analysis_medium_medium111","category":"enterprise_industry_analysis","score":0.0,"tokens":111632,"cost_usd":0.223891,"elapsed_time":46.39,"gpr":0.25,"tpe":0.9,"ee":0.5714},{"task_id":"task_297_enterprise_industry_policy_analysis_easy_easy067","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":445089,"cost_usd":1.059325,"elapsed_time":50.61,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_298_enterprise_industry_policy_analysis_easy_easy068","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":54184,"cost_usd":0.097792,"elapsed_time":36.14,"gpr":1.0,"tpe":0.81,"ee":0.6},{"task_id":"task_299_enterprise_industry_policy_analysis_easy_easy069","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":128640,"cost_usd":0.314819,"elapsed_time":35.3,"gpr":0.0,"tpe":0.0,"ee":1.0},{"task_id":"task_300_enterprise_industry_policy_analysis_easy_easy070","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":64051,"cost_usd":0.142028,"elapsed_time":23.18,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_301_enterprise_industry_policy_analysis_easy_easy071","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":538293,"cost_usd":1.073066,"elapsed_time":77.43,"gpr":1.0,"tpe":0.4783,"ee":0.3},{"task_id":"task_302_enterprise_industry_policy_analysis_easy_easy072","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":59223,"cost_usd":0.099384,"elapsed_time":42.85,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_303_enterprise_industry_policy_analysis_easy_easy073","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":48068,"cost_usd":0.09119,"elapsed_time":24.98,"gpr":1.0,"tpe":1.0,"ee":1.0},{"task_id":"task_304_enterprise_industry_policy_analysis_easy_easy074","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":50493,"cost_usd":0.116193,"elapsed_time":37.62,"gpr":0.0,"tpe":0.0,"ee":0.75},{"task_id":"task_305_enterprise_industry_policy_analysis_easy_easy075","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":100867,"cost_usd":0.171746,"elapsed_time":43.35,"gpr":1.0,"tpe":0.729,"ee":0.5},{"task_id":"task_306_enterprise_industry_policy_analysis_easy_easy076","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":122168,"cost_usd":0.272196,"elapsed_time":32.88,"gpr":1.0,"tpe":0.9,"ee":0.75},{"task_id":"task_307_enterprise_industry_policy_analysis_medium_medium001","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":184574,"cost_usd":0.369838,"elapsed_time":102.74,"gpr":0.2,"tpe":0.6561,"ee":0.4545},{"task_id":"task_308_enterprise_industry_policy_analysis_medium_medium002","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":97030,"cost_usd":0.150006,"elapsed_time":114.7,"gpr":1.0,"tpe":0.962,"ee":0.625},{"task_id":"task_309_enterprise_industry_policy_analysis_medium_medium003","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":102959,"cost_usd":0.175698,"elapsed_time":92.39,"gpr":0.8,"tpe":0.7695,"ee":0.5556},{"task_id":"task_310_enterprise_industry_policy_analysis_medium_medium004","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":159194,"cost_usd":0.343132,"elapsed_time":52.18,"gpr":0.2,"tpe":1.0,"ee":0.7143},{"task_id":"task_311_enterprise_industry_policy_analysis_medium_medium005","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":293328,"cost_usd":0.699237,"elapsed_time":97.07,"gpr":0.0,"tpe":0.0,"ee":0.7143},{"task_id":"task_312_enterprise_industry_policy_analysis_medium_medium006","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":201801,"cost_usd":0.412399,"elapsed_time":100.08,"gpr":0.0,"tpe":0.0,"ee":0.4167},{"task_id":"task_313_enterprise_industry_policy_analysis_medium_medium007","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":148373,"cost_usd":0.233976,"elapsed_time":62.88,"gpr":0.6,"tpe":0.729,"ee":0.5556},{"task_id":"task_314_enterprise_industry_policy_analysis_medium_medium008","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":610145,"cost_usd":1.037621,"elapsed_time":107.71,"gpr":0.6,"tpe":0.6145,"ee":0.4167},{"task_id":"task_315_enterprise_industry_policy_analysis_medium_medium009","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":299679,"cost_usd":0.580977,"elapsed_time":99.97,"gpr":1.0,"tpe":0.7937,"ee":0.5556},{"task_id":"task_316_enterprise_industry_policy_analysis_medium_medium010","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":384928,"cost_usd":0.907497,"elapsed_time":119.58,"gpr":0.2,"tpe":0.729,"ee":0.4545},{"task_id":"task_317_enterprise_industry_policy_analysis_medium_medium011","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":161341,"cost_usd":0.307368,"elapsed_time":76.04,"gpr":1.0,"tpe":0.7425,"ee":0.4545},{"task_id":"task_318_enterprise_industry_policy_analysis_medium_medium012","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":253433,"cost_usd":0.571922,"elapsed_time":85.93,"gpr":0.0,"tpe":0.0,"ee":0.7143},{"task_id":"task_319_enterprise_industry_policy_analysis_medium_medium013","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":162304,"cost_usd":0.281138,"elapsed_time":97.13,"gpr":1.0,"tpe":0.6738,"ee":0.5},{"task_id":"task_320_enterprise_industry_policy_analysis_medium_medium014","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":97655,"cost_usd":0.182131,"elapsed_time":50.75,"gpr":1.0,"tpe":0.98,"ee":0.8333},{"task_id":"task_321_enterprise_industry_policy_analysis_medium_medium015","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":149961,"cost_usd":0.31467,"elapsed_time":70.32,"gpr":0.4,"tpe":0.9,"ee":0.7143},{"task_id":"task_322_enterprise_industry_policy_analysis_medium_medium016","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":529572,"cost_usd":1.215164,"elapsed_time":98.47,"gpr":1.0,"tpe":0.7913,"ee":0.4167},{"task_id":"task_323_enterprise_industry_policy_analysis_medium_medium017","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":162185,"cost_usd":0.355919,"elapsed_time":86.15,"gpr":0.8,"tpe":0.914,"ee":0.5},{"task_id":"task_324_enterprise_industry_policy_analysis_medium_medium018","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":169327,"cost_usd":0.326319,"elapsed_time":91.81,"gpr":0.8,"tpe":0.5303,"ee":0.3571},{"task_id":"task_325_enterprise_industry_policy_analysis_medium_medium019","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":160664,"cost_usd":0.355456,"elapsed_time":64.17,"gpr":0.2,"tpe":1.0,"ee":0.7143},{"task_id":"task_326_enterprise_industry_policy_analysis_medium_medium020","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":183205,"cost_usd":0.319706,"elapsed_time":80.5,"gpr":0.4,"tpe":1.0,"ee":0.4545},{"task_id":"task_327_enterprise_industry_policy_analysis_medium_medium021","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":312454,"cost_usd":0.741015,"elapsed_time":75.55,"gpr":0.4,"tpe":1.0,"ee":0.7143},{"task_id":"task_328_enterprise_industry_policy_analysis_medium_medium022","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":292352,"cost_usd":0.62148,"elapsed_time":124.0,"gpr":0.0,"tpe":0.0,"ee":0.4167},{"task_id":"task_329_enterprise_industry_policy_analysis_medium_medium023","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":305884,"cost_usd":0.603531,"elapsed_time":116.38,"gpr":1.0,"tpe":0.4496,"ee":0.3846},{"task_id":"task_330_enterprise_industry_policy_analysis_medium_medium024","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":145886,"cost_usd":0.275393,"elapsed_time":82.46,"gpr":0.4,"tpe":0.733,"ee":0.5},{"task_id":"task_331_enterprise_industry_policy_analysis_medium_medium025","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":608238,"cost_usd":1.375822,"elapsed_time":130.79,"gpr":0.4,"tpe":0.6937,"ee":0.3125},{"task_id":"task_332_enterprise_industry_policy_analysis_medium_medium026","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":528417,"cost_usd":1.114811,"elapsed_time":101.99,"gpr":1.0,"tpe":1.0,"ee":0.3571},{"task_id":"task_333_enterprise_industry_policy_analysis_medium_medium027","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":209147,"cost_usd":0.417103,"elapsed_time":91.86,"gpr":0.8,"tpe":0.4783,"ee":0.3846},{"task_id":"task_334_enterprise_industry_policy_analysis_medium_medium028","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":1783937,"cost_usd":4.388034,"elapsed_time":218.52,"gpr":0.8,"tpe":0.6036,"ee":0.3571},{"task_id":"task_335_enterprise_industry_policy_analysis_medium_medium029","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":451178,"cost_usd":1.048391,"elapsed_time":112.69,"gpr":0.2,"tpe":0.729,"ee":0.3571},{"task_id":"task_336_enterprise_industry_policy_analysis_medium_medium030","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":128472,"cost_usd":0.280336,"elapsed_time":54.8,"gpr":0.6,"tpe":1.0,"ee":0.8333},{"task_id":"task_337_enterprise_industry_policy_analysis_medium_medium031","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":285214,"cost_usd":0.477424,"elapsed_time":79.79,"gpr":1.0,"tpe":0.7657,"ee":0.4},{"task_id":"task_338_enterprise_industry_policy_analysis_medium_medium032","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":317679,"cost_usd":0.737951,"elapsed_time":119.88,"gpr":0.5,"tpe":1.0,"ee":0.8},{"task_id":"task_339_enterprise_industry_policy_analysis_medium_medium033","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":240340,"cost_usd":0.531379,"elapsed_time":76.2,"gpr":1.0,"tpe":0.6383,"ee":0.4},{"task_id":"task_340_enterprise_industry_policy_analysis_medium_medium034","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":673066,"cost_usd":1.600656,"elapsed_time":106.18,"gpr":0.0,"tpe":0.0,"ee":0.3333},{"task_id":"task_341_enterprise_industry_policy_analysis_medium_medium035","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":152845,"cost_usd":0.313269,"elapsed_time":59.96,"gpr":1.0,"tpe":0.5905,"ee":0.4444},{"task_id":"task_342_enterprise_industry_policy_analysis_medium_medium036","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":92347,"cost_usd":0.179664,"elapsed_time":60.07,"gpr":0.0,"tpe":0.0,"ee":0.5714},{"task_id":"task_343_enterprise_industry_policy_analysis_medium_medium037","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":121806,"cost_usd":0.27007,"elapsed_time":48.58,"gpr":1.0,"tpe":0.975,"ee":0.8},{"task_id":"task_344_enterprise_industry_policy_analysis_medium_medium038","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":463825,"cost_usd":1.085965,"elapsed_time":113.58,"gpr":1.0,"tpe":0.4997,"ee":0.3077},{"task_id":"task_345_enterprise_industry_policy_analysis_medium_medium039","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":681629,"cost_usd":1.559783,"elapsed_time":118.99,"gpr":0.5,"tpe":0.7391,"ee":0.3333},{"task_id":"task_346_enterprise_industry_policy_analysis_medium_medium040","category":"enterprise_industry_policy_analysis","score":1.0,"tokens":190645,"cost_usd":0.460892,"elapsed_time":55.37,"gpr":1.0,"tpe":1.0,"ee":0.6667},{"task_id":"task_347_enterprise_industry_policy_analysis_medium_medium041","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":179123,"cost_usd":0.387409,"elapsed_time":66.04,"gpr":0.0,"tpe":0.0,"ee":0.5714},{"task_id":"task_348_enterprise_industry_policy_analysis_medium_medium042","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":190916,"cost_usd":0.426135,"elapsed_time":67.67,"gpr":0.0,"tpe":0.0,"ee":0.4444},{"task_id":"task_349_enterprise_industry_policy_analysis_medium_medium043","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":136017,"cost_usd":0.234846,"elapsed_time":78.76,"gpr":0.25,"tpe":0.5905,"ee":0.4},{"task_id":"task_350_enterprise_industry_policy_analysis_medium_medium044","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":470670,"cost_usd":1.134367,"elapsed_time":83.21,"gpr":0.5,"tpe":1.0,"ee":0.4444},{"task_id":"task_351_enterprise_industry_policy_analysis_medium_medium045","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":553078,"cost_usd":1.186955,"elapsed_time":101.52,"gpr":0.5,"tpe":1.0,"ee":0.3077},{"task_id":"task_352_enterprise_industry_policy_analysis_medium_medium046","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":164946,"cost_usd":0.350026,"elapsed_time":83.95,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_353_enterprise_industry_policy_analysis_medium_medium047","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":114497,"cost_usd":0.223132,"elapsed_time":54.27,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_354_enterprise_industry_policy_analysis_medium_medium048","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":161453,"cost_usd":0.299855,"elapsed_time":72.85,"gpr":0.25,"tpe":0.5905,"ee":0.3636},{"task_id":"task_355_enterprise_industry_policy_analysis_medium_medium049","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":248610,"cost_usd":0.478095,"elapsed_time":68.93,"gpr":0.75,"tpe":0.843,"ee":0.4444},{"task_id":"task_356_enterprise_industry_policy_analysis_medium_medium050","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":311084,"cost_usd":0.714325,"elapsed_time":65.81,"gpr":0.25,"tpe":0.9,"ee":0.5714},{"task_id":"task_357_enterprise_industry_policy_analysis_medium_medium051","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":117154,"cost_usd":0.221917,"elapsed_time":56.74,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_358_enterprise_industry_policy_analysis_medium_medium052","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":245249,"cost_usd":0.454919,"elapsed_time":68.79,"gpr":0.0,"tpe":0.0,"ee":0.3636},{"task_id":"task_359_enterprise_industry_policy_analysis_medium_medium053","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":249926,"cost_usd":0.595235,"elapsed_time":59.19,"gpr":0.0,"tpe":0.0,"ee":0.5714},{"task_id":"task_360_enterprise_industry_policy_analysis_medium_medium054","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":188256,"cost_usd":0.366911,"elapsed_time":120.23,"gpr":0.5,"tpe":0.7452,"ee":0.3333},{"task_id":"task_361_enterprise_industry_policy_analysis_medium_medium055","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":116957,"cost_usd":0.226338,"elapsed_time":100.35,"gpr":0.0,"tpe":0.0,"ee":0.4444},{"task_id":"task_362_enterprise_industry_policy_analysis_medium_medium056","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":127577,"cost_usd":0.224977,"elapsed_time":77.81,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_363_enterprise_industry_policy_analysis_medium_medium057","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":213317,"cost_usd":0.402253,"elapsed_time":55.27,"gpr":0.0,"tpe":0.0,"ee":0.5714},{"task_id":"task_364_enterprise_industry_policy_analysis_medium_medium058","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":152699,"cost_usd":0.327396,"elapsed_time":49.56,"gpr":0.75,"tpe":1.0,"ee":0.5714},{"task_id":"task_365_enterprise_industry_policy_analysis_medium_medium059","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":154671,"cost_usd":0.230482,"elapsed_time":41.43,"gpr":0.0,"tpe":0.0,"ee":0.8},{"task_id":"task_366_enterprise_industry_policy_analysis_medium_medium060","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":404515,"cost_usd":0.881449,"elapsed_time":77.83,"gpr":0.5,"tpe":1.0,"ee":0.4},{"task_id":"task_367_enterprise_industry_policy_analysis_medium_medium061","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":302347,"cost_usd":0.580047,"elapsed_time":77.38,"gpr":0.0,"tpe":0.0,"ee":0.4444},{"task_id":"task_368_enterprise_industry_policy_analysis_medium_medium062","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":447226,"cost_usd":1.06789,"elapsed_time":156.82,"gpr":1.0,"tpe":0.975,"ee":0.4},{"task_id":"task_369_enterprise_industry_policy_analysis_medium_medium063","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":113831,"cost_usd":0.15273,"elapsed_time":52.62,"gpr":0.75,"tpe":1.0,"ee":0.5},{"task_id":"task_370_enterprise_industry_policy_analysis_medium_medium064","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":139262,"cost_usd":0.295644,"elapsed_time":51.48,"gpr":0.5,"tpe":1.0,"ee":0.5},{"task_id":"task_371_enterprise_industry_policy_analysis_medium_medium065","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":258687,"cost_usd":0.42272,"elapsed_time":86.1,"gpr":0.0,"tpe":0.0,"ee":0.6667},{"task_id":"task_372_enterprise_industry_policy_analysis_medium_medium066","category":"enterprise_industry_policy_analysis","score":0.0,"tokens":224220,"cost_usd":0.504148,"elapsed_time":84.59,"gpr":0.5,"tpe":1.0,"ee":0.5714},{"task_id":"task_373_hypothesis_verification_hard_hard001","category":"hypothesis_verification","score":0.0,"tokens":2045581,"cost_usd":3.788032,"elapsed_time":294.24,"gpr":0.2222,"tpe":1.0,"ee":0.4118},{"task_id":"task_374_hypothesis_verification_hard_hard002","category":"hypothesis_verification","score":0.33,"tokens":781649,"cost_usd":1.703465,"elapsed_time":169.55,"gpr":0.375,"tpe":0.9333,"ee":0.5556},{"task_id":"task_375_hypothesis_verification_hard_hard003","category":"hypothesis_verification","score":0.0,"tokens":686597,"cost_usd":1.577031,"elapsed_time":277.18,"gpr":0.1111,"tpe":0.4783,"ee":0.381},{"task_id":"task_376_hypothesis_verification_hard_hard004","category":"hypothesis_verification","score":0.0,"tokens":545304,"cost_usd":0.953989,"elapsed_time":158.95,"gpr":0.0,"tpe":0.0,"ee":0.3889},{"task_id":"task_377_hypothesis_verification_hard_hard005","category":"hypothesis_verification","score":0.0,"tokens":268137,"cost_usd":0.614389,"elapsed_time":85.8,"gpr":0.2308,"tpe":1.0,"ee":1.125},{"task_id":"task_378_hypothesis_verification_hard_hard006","category":"hypothesis_verification","score":0.0,"tokens":512746,"cost_usd":1.241434,"elapsed_time":132.67,"gpr":0.0,"tpe":0.0,"ee":0.7},{"task_id":"task_379_hypothesis_verification_hard_hard007","category":"hypothesis_verification","score":0.0,"tokens":2490458,"cost_usd":4.450758,"elapsed_time":374.34,"gpr":0.0,"tpe":0.0,"ee":0.2647},{"task_id":"task_380_hypothesis_verification_hard_hard008","category":"hypothesis_verification","score":0.0,"tokens":338581,"cost_usd":0.743556,"elapsed_time":170.9,"gpr":0.5556,"tpe":0.6182,"ee":0.4375},{"task_id":"task_381_hypothesis_verification_hard_hard009","category":"hypothesis_verification","score":0.0,"tokens":307559,"cost_usd":0.603981,"elapsed_time":137.15,"gpr":0.0,"tpe":0.0,"ee":0.3158},{"task_id":"task_382_hypothesis_verification_hard_hard010","category":"hypothesis_verification","score":0.0,"tokens":515247,"cost_usd":1.031065,"elapsed_time":185.86,"gpr":0.2,"tpe":0.9333,"ee":0.6364},{"task_id":"task_383_hypothesis_verification_hard_hard011","category":"hypothesis_verification","score":0.0,"tokens":438772,"cost_usd":0.862679,"elapsed_time":151.93,"gpr":0.0,"tpe":0.0,"ee":0.3684},{"task_id":"task_384_hypothesis_verification_hard_hard012","category":"hypothesis_verification","score":0.0,"tokens":721191,"cost_usd":1.6534,"elapsed_time":264.13,"gpr":0.0,"tpe":0.0,"ee":0.4211},{"task_id":"task_385_hypothesis_verification_hard_hard013","category":"hypothesis_verification","score":0.0,"tokens":1179735,"cost_usd":2.743877,"elapsed_time":271.3,"gpr":0.0,"tpe":0.0,"ee":0.3182},{"task_id":"task_386_hypothesis_verification_hard_hard014","category":"hypothesis_verification","score":0.0,"tokens":854434,"cost_usd":1.866677,"elapsed_time":220.56,"gpr":0.5,"tpe":0.528,"ee":0.2727},{"task_id":"task_387_hypothesis_verification_hard_hard015","category":"hypothesis_verification","score":0.0,"tokens":468434,"cost_usd":0.852853,"elapsed_time":199.75,"gpr":0.3636,"tpe":0.8116,"ee":0.3889},{"task_id":"task_388_hypothesis_verification_medium_medium001","category":"hypothesis_verification","score":0.0,"tokens":400473,"cost_usd":0.657243,"elapsed_time":284.76,"gpr":0.0,"tpe":0.0,"ee":0.2917},{"task_id":"task_389_hypothesis_verification_medium_medium002","category":"hypothesis_verification","score":0.0,"tokens":1014711,"cost_usd":1.731039,"elapsed_time":177.2,"gpr":0.0,"tpe":0.0,"ee":0.4118},{"task_id":"task_390_hypothesis_verification_medium_medium003","category":"hypothesis_verification","score":0.0,"tokens":315080,"cost_usd":0.650373,"elapsed_time":121.17,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_391_hypothesis_verification_medium_medium004","category":"hypothesis_verification","score":0.0,"tokens":249193,"cost_usd":0.551391,"elapsed_time":84.75,"gpr":0.2,"tpe":1.0,"ee":0.6364},{"task_id":"task_392_hypothesis_verification_medium_medium005","category":"hypothesis_verification","score":0.0,"tokens":538044,"cost_usd":1.28787,"elapsed_time":126.05,"gpr":0.4,"tpe":0.6925,"ee":0.5833},{"task_id":"task_393_hypothesis_verification_medium_medium006","category":"hypothesis_verification","score":0.0,"tokens":156246,"cost_usd":0.338808,"elapsed_time":80.2,"gpr":0.0,"tpe":0.0,"ee":0.6},{"task_id":"task_394_hypothesis_verification_medium_medium007","category":"hypothesis_verification","score":0.0,"tokens":288408,"cost_usd":0.668616,"elapsed_time":95.47,"gpr":0.0,"tpe":0.0,"ee":0.7778},{"task_id":"task_395_hypothesis_verification_medium_medium008","category":"hypothesis_verification","score":0.0,"tokens":232305,"cost_usd":0.538453,"elapsed_time":99.34,"gpr":0.0,"tpe":0.0,"ee":0.7273},{"task_id":"task_396_hypothesis_verification_medium_medium009","category":"hypothesis_verification","score":0.0,"tokens":265784,"cost_usd":0.577647,"elapsed_time":80.8,"gpr":0.25,"tpe":0.5314,"ee":0.4286},{"task_id":"task_397_hypothesis_verification_medium_medium010","category":"hypothesis_verification","score":0.0,"tokens":489930,"cost_usd":1.006034,"elapsed_time":146.7,"gpr":0.0,"tpe":0.0,"ee":0.7},{"task_id":"task_398_hypothesis_verification_medium_medium011","category":"hypothesis_verification","score":1.0,"tokens":262268,"cost_usd":0.447927,"elapsed_time":108.71,"gpr":1.0,"tpe":0.5916,"ee":0.4},{"task_id":"task_399_hypothesis_verification_medium_medium012","category":"hypothesis_verification","score":0.0,"tokens":200817,"cost_usd":0.429427,"elapsed_time":101.44,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_400_hypothesis_verification_medium_medium013","category":"hypothesis_verification","score":0.0,"tokens":235260,"cost_usd":0.505142,"elapsed_time":76.28,"gpr":0.0,"tpe":0.0,"ee":0.875},{"task_id":"task_401_hypothesis_verification_medium_medium014","category":"hypothesis_verification","score":0.0,"tokens":614592,"cost_usd":1.495274,"elapsed_time":93.99,"gpr":0.2,"tpe":1.0,"ee":0.8889},{"task_id":"task_402_industry_planning_hard_hard001","category":"industry_planning","score":1.0,"tokens":800508,"cost_usd":1.210857,"elapsed_time":139.94,"gpr":1.0,"tpe":0.8833,"ee":0.7778},{"task_id":"task_403_industry_planning_hard_hard002","category":"industry_planning","score":0.0,"tokens":1671729,"cost_usd":4.108486,"elapsed_time":221.15,"gpr":0.4167,"tpe":0.6561,"ee":0.4667},{"task_id":"task_404_industry_planning_hard_hard003","category":"industry_planning","score":0.0,"tokens":652787,"cost_usd":1.59528,"elapsed_time":197.95,"gpr":0.0909,"tpe":0.9,"ee":0.6667},{"task_id":"task_405_industry_planning_hard_hard004","category":"industry_planning","score":0.0,"tokens":206352,"cost_usd":0.498566,"elapsed_time":114.62,"gpr":0.25,"tpe":0.81,"ee":0.6667},{"task_id":"task_406_industry_planning_hard_hard005","category":"industry_planning","score":0.0,"tokens":417674,"cost_usd":0.953985,"elapsed_time":129.7,"gpr":0.1111,"tpe":0.729,"ee":0.4},{"task_id":"task_407_industry_planning_hard_hard006","category":"industry_planning","score":0.0,"tokens":996612,"cost_usd":2.178641,"elapsed_time":159.8,"gpr":0.4167,"tpe":1.0,"ee":0.5833},{"task_id":"task_408_industry_planning_hard_hard007","category":"industry_planning","score":0.0,"tokens":254347,"cost_usd":0.518651,"elapsed_time":124.86,"gpr":0.0,"tpe":0.0,"ee":0.4615},{"task_id":"task_409_industry_planning_hard_hard008","category":"industry_planning","score":0.0,"tokens":1299280,"cost_usd":3.028462,"elapsed_time":219.67,"gpr":0.125,"tpe":0.729,"ee":0.4286},{"task_id":"task_410_industry_planning_hard_hard009","category":"industry_planning","score":0.0,"tokens":405424,"cost_usd":0.932708,"elapsed_time":148.81,"gpr":0.7,"tpe":0.818,"ee":0.6},{"task_id":"task_411_industry_planning_hard_hard010","category":"industry_planning","score":0.0,"tokens":580464,"cost_usd":1.178694,"elapsed_time":246.48,"gpr":0.3333,"tpe":1.0,"ee":0.375},{"task_id":"task_412_industry_planning_hard_hard011","category":"industry_planning","score":1.0,"tokens":381529,"cost_usd":0.876014,"elapsed_time":115.8,"gpr":1.0,"tpe":0.7898,"ee":0.4615},{"task_id":"task_413_industry_planning_hard_hard012","category":"industry_planning","score":0.0,"tokens":528476,"cost_usd":1.219241,"elapsed_time":149.25,"gpr":0.0,"tpe":0.0,"ee":0.3158},{"task_id":"task_414_industry_planning_hard_hard013","category":"industry_planning","score":0.0,"tokens":595455,"cost_usd":1.360389,"elapsed_time":187.68,"gpr":0.0,"tpe":0.0,"ee":0.5385},{"task_id":"task_415_industry_planning_hard_hard014","category":"industry_planning","score":0.0,"tokens":540484,"cost_usd":1.060661,"elapsed_time":202.32,"gpr":0.0,"tpe":0.0,"ee":0.2941},{"task_id":"task_416_industry_planning_medium_medium001","category":"industry_planning","score":0.5,"tokens":264321,"cost_usd":0.484634,"elapsed_time":166.94,"gpr":0.1667,"tpe":1.0,"ee":0.5333},{"task_id":"task_417_industry_planning_medium_medium002","category":"industry_planning","score":0.0,"tokens":618838,"cost_usd":1.079781,"elapsed_time":217.56,"gpr":0.0,"tpe":0.0,"ee":0.3636},{"task_id":"task_418_industry_planning_medium_medium003","category":"industry_planning","score":0.0,"tokens":617447,"cost_usd":1.259161,"elapsed_time":209.72,"gpr":0.1429,"tpe":0.3487,"ee":0.4211},{"task_id":"task_419_industry_planning_medium_medium004","category":"industry_planning","score":1.0,"tokens":368968,"cost_usd":0.729125,"elapsed_time":144.99,"gpr":1.0,"tpe":0.5895,"ee":0.4667},{"task_id":"task_420_industry_planning_medium_medium005","category":"industry_planning","score":0.0,"tokens":394165,"cost_usd":0.746756,"elapsed_time":158.66,"gpr":0.2222,"tpe":0.4329,"ee":0.4118},{"task_id":"task_421_industry_planning_medium_medium006","category":"industry_planning","score":0.0,"tokens":750359,"cost_usd":1.462224,"elapsed_time":199.52,"gpr":0.0,"tpe":0.0,"ee":0.7273},{"task_id":"task_422_industry_planning_medium_medium007","category":"industry_planning","score":0.0,"tokens":396055,"cost_usd":0.745131,"elapsed_time":216.85,"gpr":0.0,"tpe":0.0,"ee":0.3478},{"task_id":"task_423_industry_planning_medium_medium008","category":"industry_planning","score":0.0,"tokens":1031285,"cost_usd":2.13901,"elapsed_time":205.24,"gpr":0.0,"tpe":0.0,"ee":0.5333},{"task_id":"task_424_industry_planning_medium_medium009","category":"industry_planning","score":0.0,"tokens":319012,"cost_usd":0.690546,"elapsed_time":131.23,"gpr":0.3,"tpe":0.6561,"ee":0.5333},{"task_id":"task_425_industry_planning_medium_medium010","category":"industry_planning","score":0.0,"tokens":322920,"cost_usd":0.630673,"elapsed_time":111.75,"gpr":0.6,"tpe":0.6475,"ee":0.5333},{"task_id":"task_426_industry_planning_medium_medium011","category":"industry_planning","score":0.0,"tokens":411369,"cost_usd":0.908223,"elapsed_time":226.5,"gpr":0.25,"tpe":0.81,"ee":0.4444},{"task_id":"task_427_industry_planning_medium_medium012","category":"industry_planning","score":0.0,"tokens":1444156,"cost_usd":2.808926,"elapsed_time":187.39,"gpr":0.0,"tpe":0.0,"ee":0.5333},{"task_id":"task_428_industry_planning_medium_medium013","category":"industry_planning","score":0.0,"tokens":662238,"cost_usd":1.505197,"elapsed_time":174.14,"gpr":0.0,"tpe":0.0,"ee":0.6667},{"task_id":"task_429_industry_planning_medium_medium014","category":"industry_planning","score":0.0,"tokens":248605,"cost_usd":0.489725,"elapsed_time":82.82,"gpr":0.0,"tpe":0.0,"ee":0.75},{"task_id":"task_430_international_comparison_hard_hard001","category":"international_comparison","score":0.0,"tokens":489841,"cost_usd":1.201989,"elapsed_time":98.68,"gpr":0.0,"tpe":0.0,"ee":0.75},{"task_id":"task_431_international_comparison_hard_hard002","category":"international_comparison","score":0.0,"tokens":1388746,"cost_usd":2.551312,"elapsed_time":516.15,"gpr":0.0,"tpe":0.0,"ee":0.1765},{"task_id":"task_432_international_comparison_hard_hard003","category":"international_comparison","score":0.0,"tokens":560234,"cost_usd":1.216819,"elapsed_time":162.02,"gpr":0.2222,"tpe":0.6569,"ee":0.3158},{"task_id":"task_433_international_comparison_hard_hard004","category":"international_comparison","score":0.0,"tokens":313905,"cost_usd":0.780776,"elapsed_time":165.31,"gpr":0.0,"tpe":0.0,"ee":0.5455},{"task_id":"task_434_international_comparison_hard_hard005","category":"international_comparison","score":0.0,"tokens":1111812,"cost_usd":2.7513,"elapsed_time":273.79,"gpr":0.0,"tpe":0.0,"ee":0.5455},{"task_id":"task_435_international_comparison_hard_hard006","category":"international_comparison","score":0.25,"tokens":503823,"cost_usd":1.201536,"elapsed_time":131.53,"gpr":0.1667,"tpe":0.729,"ee":0.7},{"task_id":"task_436_international_comparison_hard_hard007","category":"international_comparison","score":0.25,"tokens":441109,"cost_usd":0.98503,"elapsed_time":273.04,"gpr":0.3636,"tpe":0.5117,"ee":0.4375},{"task_id":"task_437_international_comparison_hard_hard008","category":"international_comparison","score":0.0,"tokens":3244675,"cost_usd":6.61794,"elapsed_time":683.15,"gpr":0.0,"tpe":0.0,"ee":0.1277},{"task_id":"task_438_international_comparison_hard_hard009","category":"international_comparison","score":0.5,"tokens":3261771,"cost_usd":7.592342,"elapsed_time":311.43,"gpr":0.5,"tpe":0.7608,"ee":0.3043},{"task_id":"task_439_international_comparison_hard_hard010","category":"international_comparison","score":0.25,"tokens":384060,"cost_usd":0.815678,"elapsed_time":95.66,"gpr":0.1667,"tpe":0.6561,"ee":0.5455},{"task_id":"task_440_international_comparison_hard_hard011","category":"international_comparison","score":0.25,"tokens":1294480,"cost_usd":3.210764,"elapsed_time":240.98,"gpr":0.2,"tpe":0.729,"ee":0.5},{"task_id":"task_441_international_comparison_hard_hard012","category":"international_comparison","score":0.0,"tokens":374695,"cost_usd":0.691494,"elapsed_time":162.48,"gpr":0.0,"tpe":0.0,"ee":0.6667},{"task_id":"task_442_international_comparison_hard_hard013","category":"international_comparison","score":0.25,"tokens":319005,"cost_usd":0.584331,"elapsed_time":194.59,"gpr":0.2,"tpe":1.0,"ee":0.4615},{"task_id":"task_443_international_comparison_hard_hard014","category":"international_comparison","score":0.25,"tokens":567474,"cost_usd":1.338066,"elapsed_time":155.3,"gpr":0.0,"tpe":0.0,"ee":0.2727},{"task_id":"task_444_international_comparison_medium_medium001","category":"international_comparison","score":1.0,"tokens":1125956,"cost_usd":2.643864,"elapsed_time":337.67,"gpr":1.0,"tpe":0.4347,"ee":0.25},{"task_id":"task_445_international_comparison_medium_medium002","category":"international_comparison","score":0.0,"tokens":379923,"cost_usd":0.95678,"elapsed_time":147.91,"gpr":0.1667,"tpe":0.81,"ee":0.5},{"task_id":"task_446_international_comparison_medium_medium003","category":"international_comparison","score":0.0,"tokens":740075,"cost_usd":1.863026,"elapsed_time":151.96,"gpr":0.0,"tpe":0.0,"ee":0.2857},{"task_id":"task_447_international_comparison_medium_medium004","category":"international_comparison","score":0.0,"tokens":150121,"cost_usd":0.322423,"elapsed_time":68.57,"gpr":0.6667,"tpe":0.7603,"ee":0.4444},{"task_id":"task_448_international_comparison_medium_medium005","category":"international_comparison","score":0.0,"tokens":1129286,"cost_usd":2.836409,"elapsed_time":158.39,"gpr":0.0,"tpe":0.0,"ee":0.2667},{"task_id":"task_449_international_comparison_medium_medium006","category":"international_comparison","score":0.0,"tokens":732558,"cost_usd":1.828821,"elapsed_time":133.46,"gpr":0.1667,"tpe":1.0,"ee":0.3125},{"task_id":"task_450_international_comparison_medium_medium007","category":"international_comparison","score":0.0,"tokens":479136,"cost_usd":1.109122,"elapsed_time":185.05,"gpr":0.0,"tpe":0.0,"ee":0.2105},{"task_id":"task_451_international_comparison_medium_medium008","category":"international_comparison","score":0.3333,"tokens":248237,"cost_usd":0.625874,"elapsed_time":156.7,"gpr":0.3333,"tpe":0.6561,"ee":0.5},{"task_id":"task_452_international_comparison_medium_medium009","category":"international_comparison","score":0.0,"tokens":318105,"cost_usd":0.795091,"elapsed_time":110.03,"gpr":0.1429,"tpe":0.81,"ee":0.4545},{"task_id":"task_453_international_comparison_medium_medium010","category":"international_comparison","score":0.0,"tokens":1983720,"cost_usd":4.813038,"elapsed_time":433.36,"gpr":0.0,"tpe":0.0,"ee":0.1786},{"task_id":"task_454_international_comparison_medium_medium011","category":"international_comparison","score":1.0,"tokens":628339,"cost_usd":1.495982,"elapsed_time":200.8,"gpr":1.0,"tpe":0.4692,"ee":0.2105},{"task_id":"task_455_international_comparison_medium_medium012","category":"international_comparison","score":0.6667,"tokens":565089,"cost_usd":0.983492,"elapsed_time":122.18,"gpr":0.5,"tpe":0.6189,"ee":0.3846},{"task_id":"task_456_international_comparison_medium_medium013","category":"international_comparison","score":0.0,"tokens":1707434,"cost_usd":4.209958,"elapsed_time":276.2,"gpr":0.0,"tpe":0.0,"ee":0.1818},{"task_id":"task_457_international_comparison_medium_medium014","category":"international_comparison","score":0.5,"tokens":263340,"cost_usd":0.450959,"elapsed_time":140.39,"gpr":0.375,"tpe":0.5137,"ee":0.4167},{"task_id":"task_458_international_comparison_medium_medium015","category":"international_comparison","score":0.0,"tokens":385488,"cost_usd":0.514053,"elapsed_time":153.89,"gpr":0.0,"tpe":0.0,"ee":0.2941},{"task_id":"task_459_international_comparison_medium_medium016","category":"international_comparison","score":0.0,"tokens":537945,"cost_usd":1.336182,"elapsed_time":96.35,"gpr":0.0,"tpe":0.0,"ee":0.625},{"task_id":"task_460_international_comparison_medium_medium017","category":"international_comparison","score":0.0,"tokens":2323245,"cost_usd":5.229518,"elapsed_time":355.01,"gpr":0.2,"tpe":0.6569,"ee":0.2273},{"task_id":"task_461_international_comparison_medium_medium018","category":"international_comparison","score":0.0,"tokens":562944,"cost_usd":1.178418,"elapsed_time":216.48,"gpr":0.1429,"tpe":1.0,"ee":0.2857},{"task_id":"task_462_international_comparison_medium_medium019","category":"international_comparison","score":0.5,"tokens":209450,"cost_usd":0.500611,"elapsed_time":170.95,"gpr":0.1667,"tpe":0.4305,"ee":0.3077},{"task_id":"task_463_international_comparison_medium_medium020","category":"international_comparison","score":0.0,"tokens":3271704,"cost_usd":6.935072,"elapsed_time":473.6,"gpr":0.0,"tpe":0.0,"ee":0.122},{"task_id":"task_464_international_comparison_medium_medium021","category":"international_comparison","score":0.0,"tokens":1290002,"cost_usd":3.009914,"elapsed_time":217.61,"gpr":0.1429,"tpe":0.3874,"ee":0.2381},{"task_id":"task_465_international_comparison_medium_medium022","category":"international_comparison","score":0.0,"tokens":4403261,"cost_usd":11.005465,"elapsed_time":651.18,"gpr":0.2727,"tpe":0.3185,"ee":0.1163},{"task_id":"task_466_international_comparison_medium_medium023","category":"international_comparison","score":0.0,"tokens":273296,"cost_usd":0.701953,"elapsed_time":225.96,"gpr":0.3333,"tpe":0.95,"ee":0.4545},{"task_id":"task_467_international_comparison_medium_medium024","category":"international_comparison","score":0.333,"tokens":274579,"cost_usd":0.460059,"elapsed_time":206.42,"gpr":0.1667,"tpe":0.5214,"ee":0.2941},{"task_id":"task_468_international_comparison_medium_medium025","category":"international_comparison","score":0.0,"tokens":992083,"cost_usd":2.481477,"elapsed_time":185.5,"gpr":0.1667,"tpe":0.81,"ee":0.2105},{"task_id":"task_469_risk_assessment_hard_hard001","category":"risk_assessment","score":0.0,"tokens":949483,"cost_usd":1.747996,"elapsed_time":185.11,"gpr":0.0,"tpe":0.0,"ee":0.28},{"task_id":"task_470_risk_assessment_hard_hard002","category":"risk_assessment","score":0.0,"tokens":587414,"cost_usd":1.15799,"elapsed_time":200.73,"gpr":0.2,"tpe":0.81,"ee":0.3333},{"task_id":"task_471_risk_assessment_hard_hard003","category":"risk_assessment","score":0.0,"tokens":827187,"cost_usd":1.642252,"elapsed_time":198.92,"gpr":0.3636,"tpe":0.8696,"ee":0.3889},{"task_id":"task_472_risk_assessment_hard_hard004","category":"risk_assessment","score":1.0,"tokens":137803,"cost_usd":0.248874,"elapsed_time":85.72,"gpr":1.0,"tpe":1.0,"ee":0.6364},{"task_id":"task_473_risk_assessment_hard_hard005","category":"risk_assessment","score":1.0,"tokens":393909,"cost_usd":0.851611,"elapsed_time":155.83,"gpr":1.0,"tpe":0.4305,"ee":0.4667},{"task_id":"task_474_risk_assessment_hard_hard006","category":"risk_assessment","score":0.0,"tokens":585053,"cost_usd":1.407285,"elapsed_time":142.41,"gpr":0.2,"tpe":0.5314,"ee":0.5},{"task_id":"task_475_risk_assessment_hard_hard007","category":"risk_assessment","score":1.0,"tokens":496113,"cost_usd":1.148023,"elapsed_time":84.89,"gpr":1.0,"tpe":0.8733,"ee":0.7},{"task_id":"task_476_risk_assessment_hard_hard008","category":"risk_assessment","score":0.0,"tokens":201264,"cost_usd":0.400794,"elapsed_time":130.0,"gpr":0.0,"tpe":0.0,"ee":0.5833},{"task_id":"task_477_risk_assessment_hard_hard009","category":"risk_assessment","score":0.0,"tokens":1445901,"cost_usd":3.155913,"elapsed_time":177.37,"gpr":0.0,"tpe":0.0,"ee":0.4667},{"task_id":"task_478_risk_assessment_hard_hard010","category":"risk_assessment","score":0.0,"tokens":540101,"cost_usd":1.292598,"elapsed_time":155.11,"gpr":0.0,"tpe":0.0,"ee":0.6154},{"task_id":"task_479_risk_assessment_hard_hard011","category":"risk_assessment","score":0.0,"tokens":3560717,"cost_usd":5.67971,"elapsed_time":383.63,"gpr":0.1,"tpe":1.0,"ee":0.3182},{"task_id":"task_480_risk_assessment_hard_hard012","category":"risk_assessment","score":0.0,"tokens":271866,"cost_usd":0.662807,"elapsed_time":111.12,"gpr":0.75,"tpe":0.729,"ee":0.6364},{"task_id":"task_481_risk_assessment_hard_hard013","category":"risk_assessment","score":0.0,"tokens":787768,"cost_usd":1.961293,"elapsed_time":192.63,"gpr":0.0,"tpe":0.0,"ee":1.125},{"task_id":"task_482_risk_assessment_medium_medium001","category":"risk_assessment","score":0.5,"tokens":1723945,"cost_usd":3.623255,"elapsed_time":215.57,"gpr":0.1667,"tpe":1.0,"ee":0.3333},{"task_id":"task_483_risk_assessment_medium_medium002","category":"risk_assessment","score":0.0,"tokens":349112,"cost_usd":0.773456,"elapsed_time":106.41,"gpr":0.25,"tpe":0.5314,"ee":0.4286},{"task_id":"task_484_risk_assessment_medium_medium003","category":"risk_assessment","score":0.0,"tokens":218757,"cost_usd":0.403094,"elapsed_time":95.37,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_485_risk_assessment_medium_medium004","category":"risk_assessment","score":0.0,"tokens":3910578,"cost_usd":6.147382,"elapsed_time":420.62,"gpr":0.8,"tpe":0.2111,"ee":0.1724},{"task_id":"task_486_risk_assessment_medium_medium005","category":"risk_assessment","score":0.0,"tokens":4793715,"cost_usd":8.903726,"elapsed_time":671.24,"gpr":0.375,"tpe":0.4488,"ee":0.125},{"task_id":"task_487_risk_assessment_medium_medium006","category":"risk_assessment","score":0.5,"tokens":2018106,"cost_usd":2.575192,"elapsed_time":270.3,"gpr":0.5455,"tpe":0.8884,"ee":0.25},{"task_id":"task_488_risk_assessment_medium_medium007","category":"risk_assessment","score":0.0,"tokens":1106576,"cost_usd":2.406243,"elapsed_time":210.01,"gpr":0.0,"tpe":0.0,"ee":0.5},{"task_id":"task_489_risk_assessment_medium_medium008","category":"risk_assessment","score":0.0,"tokens":480383,"cost_usd":0.81891,"elapsed_time":136.23,"gpr":0.0,"tpe":0.0,"ee":0.4},{"task_id":"task_490_risk_assessment_medium_medium009","category":"risk_assessment","score":0.0,"tokens":1661562,"cost_usd":3.100099,"elapsed_time":190.67,"gpr":0.5,"tpe":0.95,"ee":0.3529},{"task_id":"task_491_risk_assessment_medium_medium010","category":"risk_assessment","score":0.0,"tokens":392455,"cost_usd":0.869068,"elapsed_time":144.68,"gpr":0.6,"tpe":0.7439,"ee":0.4},{"task_id":"task_492_risk_assessment_medium_medium011","category":"risk_assessment","score":0.0,"tokens":708723,"cost_usd":1.32979,"elapsed_time":151.4,"gpr":0.3636,"tpe":0.7439,"ee":0.3529}]}]};
+const COLORS = ['#06b6d4','#3b82f6','#8b5cf6','#10b981','#f59e0b','#f43f5e'];
+let lang = 'en';
+let lbSort = {col:'acc', asc:false};
+const TASK_PAGE = 200;
+let taskShown = TASK_PAGE;
+
+function init() {
+  document.getElementById('pill-tasks').textContent = D.total_tasks;
+  document.getElementById('pill-cats').textContent = D.total_categories;
+  document.getElementById('pill-models').textContent = D.models.length;
+
+  renderOverview();
+  renderLeaderboard();
+  renderCategoryBars();
+  populateFilters();
+  renderTaskTable();
+  renderPricing();
+  renderFooter();
+  bindNavScroll();
+  bindTableSort();
+  bindModalKeys();
+}
+
+/* ── Pricing Modal ── */
+function fmtRate(v) { return v == null ? '<span class="na">—</span>' : '$' + Number(v).toFixed(v < 0.1 ? 4 : (v < 1 ? 3 : 2)); }
+function renderPricing() {
+  const tb = document.getElementById('pricing-body');
+  if (!tb || !D.pricing) return;
+  tb.innerHTML = '';
+  D.pricing.forEach(p => {
+    tb.innerHTML += '<tr>' +
+      '<td>' + esc(p.model) + '</td>' +
+      '<td>' + fmtRate(p.input) + '</td>' +
+      '<td>' + fmtRate(p.output) + '</td>' +
+      '<td>' + fmtRate(p.cache_read) + '</td>' +
+      '<td>' + fmtRate(p.cache_write) + '</td>' +
+    '</tr>';
+  });
+  document.getElementById('pricing-source').textContent =
+    lang === 'zh' ? D.pricing_source_zh : D.pricing_source;
+}
+function openPricingModal(e) {
+  if (e) e.stopPropagation();
+  document.getElementById('pricing-modal').classList.add('show');
+  document.body.style.overflow = 'hidden';
+}
+function closePricingModal() {
+  document.getElementById('pricing-modal').classList.remove('show');
+  document.body.style.overflow = '';
+}
+function closePricingModalOverlay(e) {
+  if (e.target.id === 'pricing-modal') closePricingModal();
+}
+function openMetricsModal(e) {
+  if (e) e.stopPropagation();
+  document.getElementById('metrics-modal').classList.add('show');
+  document.body.style.overflow = 'hidden';
+}
+function closeMetricsModal() {
+  document.getElementById('metrics-modal').classList.remove('show');
+  document.body.style.overflow = '';
+}
+function closeMetricsModalOverlay(e) {
+  if (e.target.id === 'metrics-modal') closeMetricsModal();
+}
+function bindModalKeys() {
+  document.addEventListener('keydown', e => {
+    if (e.key === 'Escape') { closePricingModal(); closeMetricsModal(); }
+  });
+}
+
+/* ── Overview Cards ── */
+function renderOverview() {
+  const c = document.getElementById('overview-cards');
+  c.innerHTML = '';
+  D.models.slice(0, 3).forEach((m, i) => {
+    const cls = i === 0 ? 'ov-card gold fade-in' : 'ov-card fade-in fade-in-' + i;
+    const rc = 'rank-' + (i+1);
+    c.innerHTML += '<div class="' + cls + '">' +
+      '<div class="ov-rank ' + rc + '">' + (i+1) + '</div>' +
+      '<div class="ov-model">' + esc(m.model) + '</div>' +
+      '<div class="ov-score"><span>' + pct(m.acc) + '</span></div>' +
+      '<div class="ov-label">Acc.</div>' +
+      '<div class="ov-stats">' +
+        stat(m.ee.toFixed(2), 'EE') +
+        stat(pct(m.gpr), 'GPR') +
+        stat(m.tpe.toFixed(2), 'TPE') +
+      '</div>' +
+      '<div class="ov-stats" style="margin-top:8px">' +
+        stat(fmtTok(m.efficiency.total_tokens), lang==='zh'?'TOKEN':'TOKENS') +
+        stat('$' + m.efficiency.total_cost_usd.toFixed(2), lang==='zh'?'费用':'COST') +
+        stat(m.efficiency.total_requests, lang==='zh'?'请求数':'REQUESTS') +
+      '</div>' +
+    '</div>';
+  });
+}
+function stat(v, k) {
+  return '<div><div class="ov-stat-val">' + v + '</div><div class="ov-stat-key">' + k + '</div></div>';
+}
+
+/* ── Leaderboard ── */
+function renderLeaderboard() {
+  const tb = document.getElementById('lb-body');
+  const models = [...D.models];
+  const col = lbSort.col;
+  models.sort((a, b) => {
+    let va = getVal(a, col), vb = getVal(b, col);
+    return lbSort.asc ? va - vb : vb - va;
+  });
+  tb.innerHTML = '';
+  models.forEach((m, i) => {
+    const rk = i + 1;
+    const rc = rk <= 3 ? ' r' + rk : '';
+    const score = m.acc;
+    tb.innerHTML += '<tr>' +
+      '<td class="rank-cell' + rc + '">' + rk + '</td>' +
+      '<td class="model-cell">' + esc(m.model) + '</td>' +
+      '<td><div class="score-bar"><div class="score-bar-bg"><div class="score-bar-fill" style="width:' + (score*100).toFixed(1) + '%"></div></div><div class="score-bar-val">' + pct(score) + '</div></div></td>' +
+      '<td class="mono">' + m.ee.toFixed(2) + '</td>' +
+      '<td class="mono">' + pct(m.gpr) + '</td>' +
+      '<td class="mono">' + m.tpe.toFixed(2) + '</td>' +
+      '<td class="mono">' + fmtTok(m.efficiency.total_tokens) + '</td>' +
+      '<td class="mono">$' + m.efficiency.total_cost_usd.toFixed(2) + '</td>' +
+      '<td class="mono">' + m.efficiency.total_requests + '</td>' +
+    '</tr>';
+  });
+}
+
+function getVal(m, col) {
+  if (col === 'acc') return m.acc;
+  if (col === 'ee') return m.ee;
+  if (col === 'gpr') return m.gpr;
+  if (col === 'tpe') return m.tpe;
+  if (col === 'total_tokens') return m.efficiency.total_tokens;
+  if (col === 'total_cost_usd') return m.efficiency.total_cost_usd;
+  if (col === 'total_requests') return m.efficiency.total_requests;
+  if (col === 'model') return m.model;
+  return 0;
+}
+
+function bindTableSort() {
+  document.querySelectorAll('#leaderboard th[data-col]').forEach(th => {
+    th.addEventListener('click', () => {
+      const col = th.dataset.col;
+      if (col === 'rank') return;
+      if (lbSort.col === col) lbSort.asc = !lbSort.asc;
+      else { lbSort.col = col; lbSort.asc = false; }
+      document.querySelectorAll('#leaderboard th').forEach(h => h.classList.remove('sorted'));
+      th.classList.add('sorted');
+      th.querySelector('.sort-arrow').textContent = lbSort.asc ? '↑' : '↓';
+      renderLeaderboard();
+    });
+  });
+}
+
+/* ── Category Bars ── */
+function renderCategoryBars() {
+  const container = document.getElementById('cat-bars');
+  container.innerHTML = '';
+  const allCats = new Set();
+  D.models.forEach(m => Object.keys(m.category_scores).forEach(c => allCats.add(c)));
+  const cats = [...allCats].sort();
+  const names = lang === 'zh' ? D.category_names_zh : D.category_names;
+
+  cats.forEach(cat => {
+    let html = '<div class="cat-group">';
+    html += '<div class="cat-group-title">' + esc(names[cat] || cat) + ' <span class="cat-badge">' + (D.models[0]?.category_scores[cat]?.count || 0) + ' ' + (lang==='zh'?'任务':'tasks') + '</span></div>';
+    D.models.forEach((m, i) => {
+      const cs = m.category_scores[cat];
+      const v = cs ? cs.acc : 0;
+      const ci = i % COLORS.length;
+      html += '<div class="bar-row">' +
+        '<div class="bar-model">' + esc(m.model) + '</div>' +
+        '<div class="bar-track"><div class="bar-fill m-color-' + ci + '" style="width:' + (v*100).toFixed(1) + '%"><span>' + pct(v) + '</span></div></div>' +
+        '<div class="bar-pct">' + pct(v) + '</div>' +
+      '</div>';
+    });
+    html += '</div>';
+    container.innerHTML += html;
+  });
+}
+
+/* ── Task Explorer ── */
+function populateFilters() {
+  const catSel = document.getElementById('filter-cat');
+  const modSel = document.getElementById('filter-model');
+  const allCats = new Set();
+  D.models.forEach(m => m.tasks.forEach(t => allCats.add(t.category)));
+  const names = lang === 'zh' ? D.category_names_zh : D.category_names;
+  [...allCats].sort().forEach(c => {
+    const o = document.createElement('option');
+    o.value = c; o.textContent = names[c] || c;
+    o.dataset.en = D.category_names[c] || c;
+    o.dataset.zh = D.category_names_zh[c] || c;
+    catSel.appendChild(o);
+  });
+  D.models.forEach(m => {
+    const o = document.createElement('option');
+    o.value = m.model; o.textContent = m.model;
+    modSel.appendChild(o);
+  });
+}
+
+function renderTaskTable() {
+  const catF = document.getElementById('filter-cat').value;
+  const modF = document.getElementById('filter-model').value;
+  const tb = document.getElementById('task-body');
+  const names = lang === 'zh' ? D.category_names_zh : D.category_names;
+  let rows = [];
+  D.models.forEach(m => {
+    if (modF !== 'all' && m.model !== modF) return;
+    m.tasks.forEach(t => {
+      if (catF !== 'all' && t.category !== catF) return;
+      rows.push({...t, model: m.model});
+    });
+  });
+  rows.sort((a, b) => b.score - a.score);
+  const total = rows.length;
+  if (taskShown > total) taskShown = Math.max(TASK_PAGE, total);
+  const shown = Math.min(taskShown, total);
+  tb.innerHTML = '';
+  const dash = '<span style="color:var(--text3)">—</span>';
+  rows.slice(0, shown).forEach(r => {
+    const isCorrect = r.score >= 1.0;
+    const eeCell = isCorrect ? r.ee.toFixed(2) : dash;
+    const gprCell = isCorrect ? dash : pct(r.gpr);
+    const tpeCell = isCorrect ? dash : r.tpe.toFixed(2);
+    tb.innerHTML += '<tr>' +
+      '<td class="mono" style="font-size:11px">' + esc(r.task_id) + '</td>' +
+      '<td style="font-size:12px">' + esc(names[r.category] || r.category) + '</td>' +
+      '<td class="model-cell" style="font-size:12px">' + esc(r.model) + '</td>' +
+      '<td class="mono">' + pct(r.score) + '</td>' +
+      '<td class="mono">' + eeCell + '</td>' +
+      '<td class="mono">' + gprCell + '</td>' +
+      '<td class="mono">' + tpeCell + '</td>' +
+      '<td class="mono">' + (r.tokens / 1e6).toFixed(1) + 'M</td>' +
+      '<td class="mono">$' + r.cost_usd.toFixed(2) + '</td>' +
+      '<td class="mono">' + r.elapsed_time.toFixed(0) + '</td>' +
+    '</tr>';
+  });
+  const wrap = document.getElementById('load-more-wrap');
+  const info = document.getElementById('load-more-info');
+  const btn = document.getElementById('load-more-btn');
+  const label = lang === 'zh'
+    ? '显示 ' + shown + ' / ' + total
+    : 'Showing ' + shown + ' of ' + total;
+  info.textContent = label;
+  if (shown < total) {
+    wrap.style.display = 'flex';
+    btn.style.display = '';
+  } else if (total > TASK_PAGE) {
+    wrap.style.display = 'flex';
+    btn.style.display = 'none';
+  } else {
+    wrap.style.display = 'none';
+  }
+}
+
+function resetTaskTable() {
+  taskShown = TASK_PAGE;
+  renderTaskTable();
+}
+
+function loadMoreTasks() {
+  taskShown += TASK_PAGE;
+  renderTaskTable();
+}
+
+/* ── Footer ── */
+function renderFooter() {
+  const f = document.getElementById('footer-info');
+  f.innerHTML =
+    '<span>' + (lang==='zh'?'生成于':'Generated') + ' ' + esc(D.generated_at) + '</span>' +
+    '<span>' + (lang==='zh'?'基准版本':'Benchmark') + ' ' + esc(D.benchmark_version) + '</span>' +
+    '<a href="https://github.com/GTMLLab/DataClaw" target="_blank">GitHub</a>';
+}
+
+/* ── Language Toggle ── */
+function toggleLang() {
+  lang = lang === 'en' ? 'zh' : 'en';
+  document.getElementById('lang-btn').textContent = lang === 'en' ? '中文' : 'EN';
+  document.querySelectorAll('[data-en]').forEach(el => {
+    el.textContent = lang === 'zh' ? el.dataset.zh : el.dataset.en;
+  });
+  // re-render dynamic sections
+  renderOverview();
+  renderCategoryBars();
+  renderPricing();
+  renderFooter();
+  // update filter options
+  const catSel = document.getElementById('filter-cat');
+  [...catSel.options].forEach(o => {
+    if (o.dataset.en) o.textContent = lang === 'zh' ? o.dataset.zh : o.dataset.en;
+  });
+  renderTaskTable();
+}
+
+/* ── Nav Scroll ── */
+function bindNavScroll() {
+  const links = document.querySelectorAll('.nav-links a');
+  const sections = ['overview','leaderboard','categories','tasks'];
+  const observer = new IntersectionObserver(entries => {
+    entries.forEach(e => {
+      if (e.isIntersecting) {
+        links.forEach(l => l.classList.remove('active'));
+        const id = e.target.id || e.target.closest('[id]')?.id;
+        const link = document.querySelector('.nav-links a[href="#' + id + '"]');
+        if (link) link.classList.add('active');
+      }
+    });
+  }, {rootMargin: '-30% 0px -60% 0px'});
+  sections.forEach(id => {
+    const el = document.getElementById(id);
+    if (el) observer.observe(el);
+  });
+}
+
+/* ── Helpers ── */
+function pct(v) { return (v * 100).toFixed(2) + '%'; }
+function fmtTok(v) { return (v / 1e6).toFixed(1) + 'M'; }
+function esc(s) { const d = document.createElement('div'); d.textContent = s; return d.innerHTML; }
+
+document.addEventListener('DOMContentLoaded', init);
+</script>
+</body>
+</html>
\ No newline at end of file
diff --git a/logo.png b/logo.png
new file mode 100644
index 0000000000000000000000000000000000000000..574b50751d3cad0d83ef4bb6a4553bca65805988
--- /dev/null
+++ b/logo.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88c366ccbd7006a31c86d9936d3d4523831d5c03dc921032b3a46c9c8a96732c
+size 428558
diff --git a/pyproject.toml b/pyproject.toml
new file mode 100644
index 0000000000000000000000000000000000000000..7555f2061f10b3b636ad57823932e09708b1f825
--- /dev/null
+++ b/pyproject.toml
@@ -0,0 +1,32 @@
+[project]
+name = "dataclaw"
+version = "0.1.0"
+description = "DataClaw — Process-oriented data-analysis benchmark for OpenClaw-style end-to-end agents"
+readme = "README.md"
+requires-python = ">=3.10"
+license = {text = "MIT"}
+authors = [{name = "The DataClaw Authors"}]
+keywords = ["benchmark", "agent", "data-analysis", "openclaw", "llm"]
+classifiers = [
+  "Development Status :: 4 - Beta",
+  "License :: OSI Approved :: MIT License",
+  "Programming Language :: Python :: 3.10",
+  "Topic :: Scientific/Engineering :: Artificial Intelligence",
+]
+dependencies = [
+  "pyyaml>=6.0.1",
+  "python-dotenv>=1.0.0",
+]
+
+[project.optional-dependencies]
+dev = [
+  "ruff>=0.6.0",
+]
+
+[project.urls]
+homepage = "https://gtmllab.github.io/DataClaw/"
+repository = "https://github.com/GTMLLab/DataClaw"
+
+[tool.ruff]
+line-length = 100
+target-version = "py310"
diff --git a/script/docker_save_image.sh b/script/docker_save_image.sh
new file mode 100644
index 0000000000000000000000000000000000000000..87d32c59262cb643a2f4709231e1596391f2e446
--- /dev/null
+++ b/script/docker_save_image.sh
@@ -0,0 +1,49 @@
+#!/usr/bin/env bash
+# Build the benchmark image and write a tarball under Images/.
+# Usage (from repo root): ./script/docker_save_image.sh
+# Optional: IMAGE_VERSION=0.2.0 ./script/docker_save_image.sh
+set -euo pipefail
+
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "${ROOT}"
+
+if [[ -n "${IMAGE_VERSION:-}" ]]; then
+  VERSION="${IMAGE_VERSION}"
+else
+  VERSION="$(
+    grep -E '^version[[:space:]]*=[[:space:]]*"' pyproject.toml \
+      | head -1 \
+      | sed -E 's/^version[[:space:]]*=[[:space:]]*"([^"]+)".*/\1/'
+  )"
+fi
+
+if [[ -z "${VERSION}" ]]; then
+  echo "Could not read version from pyproject.toml" >&2
+  exit 1
+fi
+
+TAG="${DOCKER_IMAGE_NAME:-dataclaw}:${VERSION}"
+OUT_DIR="${OUT_DIR:-${ROOT}/Images}"
+TAR_BASENAME="${TAR_BASENAME:-dataclaw_ubuntu_v${VERSION}.tar}"
+COMPRESS="${COMPRESS:-0}"
+
+mkdir -p "${OUT_DIR}"
+TAR_PATH="${OUT_DIR}/${TAR_BASENAME}"
+
+echo "Building image ${TAG}..."
+docker build -t "${TAG}" "${ROOT}"
+
+echo "Saving to ${TAR_PATH}..."
+docker save -o "${TAR_PATH}" "${TAG}"
+
+if [[ "${COMPRESS}" == "1" ]]; then
+  echo "Compressing..."
+  gzip -f "${TAR_PATH}"
+  TAR_PATH="${TAR_PATH}.gz"
+fi
+
+echo ""
+echo "Done: ${TAR_PATH}"
+echo "Publish: upload this file (e.g. Hugging Face dataset repo)."
+echo "Consumer: docker load -i $(basename "${TAR_PATH}")"
+echo "Run:      python dataclaw/eval/run_batch.py --model <model_id>"
diff --git a/script/run.sh b/script/run.sh
new file mode 100644
index 0000000000000000000000000000000000000000..0cd91dec6a1f0111ec982e013e24b75ad444b849
--- /dev/null
+++ b/script/run.sh
@@ -0,0 +1,24 @@
+#!/usr/bin/env bash
+set -euo pipefail
+cd "$(dirname "$0")/.."
+
+# --- Build the runtime image (only needed once) ---
+# docker build -t dataclaw:0.1.0 .
+
+# --- Run all tasks ---
+# python dataclaw/eval/run_batch.py --model openrouter/anthropic/claude-sonnet-4.6
+
+# --- Run all tasks with parallelism ---
+# python dataclaw/eval/run_batch.py --model openrouter/anthropic/claude-sonnet-4.6 --parallel 4
+
+# --- Run specific tasks ---
+# python dataclaw/eval/run_batch.py --model openrouter/anthropic/claude-sonnet-4.6 --suite task_001,task_002
+
+# --- Run a single task file ---
+# python dataclaw/eval/run_batch.py --task tasks/task_001_comprehensive_decision_hard_hard001.md --model openrouter/anthropic/claude-sonnet-4.6
+
+# --- Default: run all tasks sequentially ---
+python dataclaw/eval/run_batch.py \
+  --model "${DEFAULT_MODEL:?Set DEFAULT_MODEL in .env}" \
+  --parallel "${DEFAULT_PARALLEL:-1}" \
+  "$@"
diff --git a/tasks/task_001_comprehensive_decision_easy_easy001.md b/tasks/task_001_comprehensive_decision_easy_easy001.md
new file mode 100644
index 0000000000000000000000000000000000000000..7e0ea8c6dfe05ffa6dbbe979bb48174722267c3e
--- /dev/null
+++ b/tasks/task_001_comprehensive_decision_easy_easy001.md
@@ -0,0 +1,117 @@
+---
+id: task_001_comprehensive_decision_easy_easy001
+name: comprehensive_decision-easy-easy001
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/easy001.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In which province is the enterprise with the highest R&D investment ratio nationwide in 2022 located?
+
+Output guidelines:
+The answer must be "yes" or "no". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Jiangsu Province"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_002_comprehensive_decision_easy_easy002.md b/tasks/task_002_comprehensive_decision_easy_easy002.md
new file mode 100644
index 0000000000000000000000000000000000000000..8dfa4d15ccd771b4023a95e07912dc8d4bd5ec9d
--- /dev/null
+++ b/tasks/task_002_comprehensive_decision_easy_easy002.md
@@ -0,0 +1,119 @@
+---
+id: task_002_comprehensive_decision_easy_easy002
+name: comprehensive_decision-easy-easy002
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/easy002.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, nationwide, how is the chemical raw materials and chemical products manufacturing industry ranked by asset scale? Please list the top five provinces.
+
+Output guidelines:
+Answer format: [Province A, Province B, ...]. Output only province names, do not add any other explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["Shandong Province", "Zhejiang Province", "Jiangsu Province", "Shanghai", "Guangdong Province"]`
+
+Scoring rules:
+- The gold answer is a list with N=5 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_4` each as 0 or 1.
+- Return `total = (sum(part_i)) / 5` exactly.
+- If the model output is missing or cannot be parsed into 5 comparable parts, score all parts 0.
+
diff --git a/tasks/task_003_comprehensive_decision_easy_easy003.md b/tasks/task_003_comprehensive_decision_easy_easy003.md
new file mode 100644
index 0000000000000000000000000000000000000000..1f00e24676e27336984441129cc551a5d4738881
--- /dev/null
+++ b/tasks/task_003_comprehensive_decision_easy_easy003.md
@@ -0,0 +1,117 @@
+---
+id: task_003_comprehensive_decision_easy_easy003
+name: comprehensive_decision-easy-easy003
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/easy003.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in the Qilu region (Shandong Province), how many policies support Zhongbai Jinmao Chain Company in its industry?
+
+Output guidelines:
+The answer must be an exact number. Output only the number, do not add units, commas, or any textual explanation. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`2`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_004_comprehensive_decision_easy_easy004.md b/tasks/task_004_comprehensive_decision_easy_easy004.md
new file mode 100644
index 0000000000000000000000000000000000000000..c39aed84c045e181723458c22da3c816444a2000
--- /dev/null
+++ b/tasks/task_004_comprehensive_decision_easy_easy004.md
@@ -0,0 +1,119 @@
+---
+id: task_004_comprehensive_decision_easy_easy004
+name: comprehensive_decision-easy-easy004
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/easy004.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, nationwide, what is the ranking of provinces by asset size in the Information Transmission, Software and Information Technology Services industry? Please list the top five provinces.
+
+Output guidelines:
+Answer format: [Province A, Province B, ...]. Output only the province ranking, do not add any other explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["Beijing", "Zhejiang Province", "Guangdong Province", "Shanghai", "Jiangsu Province"]`
+
+Scoring rules:
+- The gold answer is a list with N=5 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_4` each as 0 or 1.
+- Return `total = (sum(part_i)) / 5` exactly.
+- If the model output is missing or cannot be parsed into 5 comparable parts, score all parts 0.
+
diff --git a/tasks/task_005_comprehensive_decision_easy_easy005.md b/tasks/task_005_comprehensive_decision_easy_easy005.md
new file mode 100644
index 0000000000000000000000000000000000000000..d991be5bef66887d7f3f2a6a7ad0e3f8e539da3e
--- /dev/null
+++ b/tasks/task_005_comprehensive_decision_easy_easy005.md
@@ -0,0 +1,119 @@
+---
+id: task_005_comprehensive_decision_easy_easy005
+name: comprehensive_decision-easy-easy005
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/easy005.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, nationwide, what is the ranking of provinces by profitability in the Real Estate industry? Please list the top five provinces.
+
+Output guidelines:
+Answer format: [Province A, Province B, ...]. Output only the province ranking, do not add any other explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["Hong Kong SAR", "Guangdong Province", "Zhejiang Province", "Beijing", "Jilin Province"]`
+
+Scoring rules:
+- The gold answer is a list with N=5 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_4` each as 0 or 1.
+- Return `total = (sum(part_i)) / 5` exactly.
+- If the model output is missing or cannot be parsed into 5 comparable parts, score all parts 0.
+
diff --git a/tasks/task_006_comprehensive_decision_easy_easy006.md b/tasks/task_006_comprehensive_decision_easy_easy006.md
new file mode 100644
index 0000000000000000000000000000000000000000..cf0c945462d7b22f524382b4b0a7c1794dc296db
--- /dev/null
+++ b/tasks/task_006_comprehensive_decision_easy_easy006.md
@@ -0,0 +1,117 @@
+---
+id: task_006_comprehensive_decision_easy_easy006
+name: comprehensive_decision-easy-easy006
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/easy006.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, what is Sichuan Province's national ranking by average R&D investment in the Information Transmission, Software and Information Technology Services industry?
+
+Output guidelines:
+The answer must be an exact number representing the ranking. Output only the number, do not add units, commas, or any textual explanation. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`10`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_007_comprehensive_decision_hard_hard001.md b/tasks/task_007_comprehensive_decision_hard_hard001.md
new file mode 100644
index 0000000000000000000000000000000000000000..73b9ec7932f27f28aebe22f86c2e961ce11ce052
--- /dev/null
+++ b/tasks/task_007_comprehensive_decision_hard_hard001.md
@@ -0,0 +1,117 @@
+---
+id: task_007_comprehensive_decision_hard_hard001
+name: comprehensive_decision-hard-hard001
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard001.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, a strategic consulting firm was commissioned by a provincial government to quantitatively rank the comprehensive attractiveness of pharmaceutical manufacturing across provinces, in order to identify priority target regions for attracting leading enterprises. The company designed a four-dimensional weighted scoring system: four original indicators—enterprise agglomeration level (weight 30%), R&D expenditure as a share of revenue (weight 30%), regional policy coverage intensity (weight 20%), and R&D human resource penetration rate (weight 20%)—were normalized (min-max) and then weighted to produce a composite score. Among these, agglomeration level is measured by the proportion of enterprises in each province to the national total in pharmaceutical manufacturing; policy intensity is measured by the ratio of relevant policy items in each province to the total number of relevant policies nationwide; and human resource penetration rate is the total number of R&D personnel in each province divided by total employees. What is the specific composite score value of the province with the highest weighted composite score after normalization across provinces?
+
+Output guidelines:
+The answer should be a numerical value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.92`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_008_comprehensive_decision_hard_hard002.md b/tasks/task_008_comprehensive_decision_hard_hard002.md
new file mode 100644
index 0000000000000000000000000000000000000000..39b944e72bf49740720b3a0f38892440eb2feae1
--- /dev/null
+++ b/tasks/task_008_comprehensive_decision_hard_hard002.md
@@ -0,0 +1,117 @@
+---
+id: task_008_comprehensive_decision_hard_hard002
+name: comprehensive_decision-hard-hard002
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard002.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, conduct a quantitative assessment of the investment value of the semiconductor industry across provinces. The evaluation framework requires incorporating three dimensions: first, industry scale (40% weight), measured by the inter-provincial rank percentile of total operating revenue in each province; second, profitability quality (30% weight), reflected by the inter-provincial rank percentile of operating profit margin (total operating profit divided by total operating revenue) in each province; third, technology output intensity (30% weight), measured by the inter-provincial rank percentile of the ratio of total patent applications to R&D expenditure (converted to 100 million yuan) in each province. The rank percentile for each indicator is calculated by sorting values from low to high, using the formula (rank - 1) / (total number of provinces - 1). Note that only provinces with complete data for all three indicators are included in the calculation. Under this weighted scoring system, what is the final score of the province ranked first in comprehensive investment value?
+
+Output guidelines:
+The answer should be a numerical value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.67`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_009_comprehensive_decision_hard_hard003.md b/tasks/task_009_comprehensive_decision_hard_hard003.md
new file mode 100644
index 0000000000000000000000000000000000000000..9b7e37b71fffd21e5b9688b9ab821ad36d4fb787
--- /dev/null
+++ b/tasks/task_009_comprehensive_decision_hard_hard003.md
@@ -0,0 +1,117 @@
+---
+id: task_009_comprehensive_decision_hard_hard003
+name: comprehensive_decision-hard-hard003
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard003.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, an automotive manufacturing enterprise commissioned a third-party institution to score and rate the industrial supporting capacity of each province before selecting a site for a new plant. The scoring rules are as follows: first, rank provinces in descending order by the number of government policies related to automotive manufacturing, and take the top five provinces by policy count as the candidate pool; then, within the candidate provinces, calculate the industrial supporting composite index, which is a weighted combination of three components—upstream and downstream supply chain density (weight 0.4), local labor reserve (weight 0.3), and government subsidy intensity per enterprise (weight 0.3). Supply chain density is defined as the ratio of total automotive manufacturing enterprises in the province to the national total in the industry; labor reserve is defined as the ratio of total industry employees in the province to the national total in the industry; subsidy intensity is defined as total government rewards and subsidies for automotive manufacturing in the province divided by the number of enterprises in the province (subsidy intensity must be normalized across all provinces before being used in the formula). Among the top five provinces by policy ranking, what is the composite index value of the province with the highest industrial supporting composite index?
+
+Output guidelines:
+Answer format: numerical value (4 decimal places). Output only the number, without units or text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.3187`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_010_comprehensive_decision_hard_hard004.md b/tasks/task_010_comprehensive_decision_hard_hard004.md
new file mode 100644
index 0000000000000000000000000000000000000000..b1b36c01e51cb1844de06512218cd59c8e03b232
--- /dev/null
+++ b/tasks/task_010_comprehensive_decision_hard_hard004.md
@@ -0,0 +1,117 @@
+---
+id: task_010_comprehensive_decision_hard_hard004
+name: comprehensive_decision-hard-hard004
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard004.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, a provincial development and reform commission, when reviewing the effectiveness of fiscal subsidies for the chemical raw materials and chemical products manufacturing industry, needed to identify enterprises with misallocated subsidy resources. Specifically, analysts must first define the scope: only examine enterprises located in provinces that have policy entries for "Chemical Raw Materials and Chemical Products Manufacturing" in the policy release status data; then use the industry-wide median of government subsidies and the median operating profit margin (profit margin = operating profit ÷ operating revenue × 100%) as dual thresholds to identify "capital misallocation" enterprises—those that simultaneously have "subsidy amount above the industry median" but "profit margin below the industry median". Among the valid enterprises in the policy-covered provinces, what is the proportion of capital misallocation enterprises as a percentage of total valid enterprises in those provinces (express the result as a percentage with 2 decimal places, without the % symbol)?
+
+Output guidelines:
+The answer should be a percentage value with 2 decimal places. Output only the number, without the % symbol or text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`23.18`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_011_comprehensive_decision_hard_hard005.md b/tasks/task_011_comprehensive_decision_hard_hard005.md
new file mode 100644
index 0000000000000000000000000000000000000000..3dfae02d18851bfbe75818012c105c0d37025e0f
--- /dev/null
+++ b/tasks/task_011_comprehensive_decision_hard_hard005.md
@@ -0,0 +1,117 @@
+---
+id: task_011_comprehensive_decision_hard_hard005
+name: comprehensive_decision-hard-hard005
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard005.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, for the information transmission, software and information technology services industry, an industry research institute sought to obtain a policy-adjusted comprehensive innovation efficiency indicator by superimposing the incentive effect of local policy support on top of raw innovation efficiency. The calculation logic is as follows: first, exclude from enterprise microdata any samples with missing R&D expenditure or annual domestic invention patent grants; for the remaining valid enterprises, aggregate by province and calculate the ratio of total invention patent grants to total R&D expenditure (converted to 100 million yuan) for each province as the province's raw innovation efficiency benchmark; then use the proportion of policy items in that province out of all information technology policies as the policy support coefficient, and multiply the raw efficiency benchmark by (1 plus the policy support coefficient) to obtain the final policy-adjusted innovation efficiency. Among all provinces with data, what is the specific value of the province with the highest adjusted efficiency?
+
+Output guidelines:
+The answer should be a numerical value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`63.74`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_012_comprehensive_decision_hard_hard006.md b/tasks/task_012_comprehensive_decision_hard_hard006.md
new file mode 100644
index 0000000000000000000000000000000000000000..7d68d55ae05eed8bc84c176a414db6aa9b5451ae
--- /dev/null
+++ b/tasks/task_012_comprehensive_decision_hard_hard006.md
@@ -0,0 +1,119 @@
+---
+id: task_012_comprehensive_decision_hard_hard006
+name: comprehensive_decision-hard-hard006
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard006.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, to measure the impact of different ownership backgrounds on the operating performance of specialized equipment manufacturing enterprises, an analysis team compared each enterprise's return on equity (ROE) level with the industry-wide return level in its province to calculate "excess ROE" as a relative performance indicator. Specifically: enterprise ROE is calculated as net profit divided by net assets (total assets minus total liabilities) multiplied by 100%; provincial industry benchmark ROE is extracted from provincial industry summary tables, calculated as total industry net profit divided by total industry net assets (total assets minus total liabilities) multiplied by 100%; each enterprise's excess ROE is the difference between its own ROE and its province's benchmark ROE. After grouping by ownership type, which ownership category has the highest mean excess ROE among enterprises? What is that mean value in percentage points?
+
+Output guidelines:
+The answer should be a percentage value with 2 decimal places. Output only the number, without the % symbol or text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["Collective Enterprise", 5.17]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_013_comprehensive_decision_hard_hard007.md b/tasks/task_013_comprehensive_decision_hard_hard007.md
new file mode 100644
index 0000000000000000000000000000000000000000..4d5b13ebe78432fbbbd1dda8715e82a992fa6c76
--- /dev/null
+++ b/tasks/task_013_comprehensive_decision_hard_hard007.md
@@ -0,0 +1,117 @@
+---
+id: task_013_comprehensive_decision_hard_hard007
+name: comprehensive_decision-hard-hard007
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard007.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, when a research institution was reviewing the implementation effectiveness of provincial industrial policies in the food and beverage industry, it found that although some provinces had issued many support policies, the profitability of enterprises within their jurisdictions was not ideal. To identify such "policy-heavy, low-return" provinces, the institution planned to analyze separately those provinces with a relatively large number of policies (including national-level policies, totaling 3 or more): sum the operating profit amounts of all food and beverage industry enterprises in these provinces and divide by the sum of operating revenue amounts to obtain the comprehensive operating profit margin for each province, then identify the province with the lowest profit margin. What is the profit margin value (as a percentage, rounded to two decimal places) for the province with the lowest comprehensive operating profit margin?
+
+Output guidelines:
+The answer should be a percentage value with 2 decimal places. Output only the number, without the % symbol or text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.59`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_014_comprehensive_decision_hard_hard008.md b/tasks/task_014_comprehensive_decision_hard_hard008.md
new file mode 100644
index 0000000000000000000000000000000000000000..1847c04da4a454c6ef6edb5c3b57684a6bf318d5
--- /dev/null
+++ b/tasks/task_014_comprehensive_decision_hard_hard008.md
@@ -0,0 +1,117 @@
+---
+id: task_014_comprehensive_decision_hard_hard008
+name: comprehensive_decision-hard-hard008
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard008.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, a private equity institution sought to identify high-quality provinces in the electricity, heat, gas and water production and supply industry that combine growth potential, market undervaluation, and innovation resilience. The screening logic has three layers: The first layer requires that the median year-over-year change in operating revenue of enterprises within the province be positive (>0%), to exclude regions where revenue is already shrinking; The second layer, based on the first layer results, further requires that the province's market valuation level be relatively low, i.e., the P/S ratio of all enterprises in the province must be lower than the median P/S ratio across all provinces nationwide (national median is calculated from the provincial P/S ratio series); The third layer adds an innovation requirement, i.e., the mean R&D investment ratio of enterprises in the province must be higher than the mean of all enterprises in the industry with R&D investment ratio records. How many provinces satisfy all three conditions simultaneously?
+
+Output guidelines:
+The answer should be an integer. Output only the number, without units or text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`4`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_015_comprehensive_decision_hard_hard009.md b/tasks/task_015_comprehensive_decision_hard_hard009.md
new file mode 100644
index 0000000000000000000000000000000000000000..2f9cc25d0c060f4f425db14b172d452877071cb2
--- /dev/null
+++ b/tasks/task_015_comprehensive_decision_hard_hard009.md
@@ -0,0 +1,117 @@
+---
+id: task_015_comprehensive_decision_hard_hard009
+name: comprehensive_decision-hard-hard009
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard009.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, an investment manager at a merger and acquisition fund was seeking "high R&D, low valuation" M&A targets in the textile, footwear and apparel industry, but the scope was limited to provinces covered by textile, footwear and apparel industry-related policies. The prerequisite for screening valid enterprises is: net profit amount strictly greater than zero, and both R&D investment ratio and company market cap fields have data records. On this basis, first use all valid enterprises in the industry as the benchmark population to calculate the median R&D investment ratio and the median P/E ratio respectively; then from the subset of valid enterprises located in policy-covered provinces, filter enterprises whose R&D investment ratio is higher than the industry median and whose P/E ratio is lower than the industry median. How many enterprises satisfy the above dual screening conditions? (P/E ratio = company market cap (100 million yuan) ÷ net profit amount (100 million yuan))
+
+Output guidelines:
+The answer should be an integer. Output only the number, without units or text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`9`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_016_comprehensive_decision_hard_hard010.md b/tasks/task_016_comprehensive_decision_hard_hard010.md
new file mode 100644
index 0000000000000000000000000000000000000000..e70636ad33afe4eeaf2c52fea8dd629c2fadddd4
--- /dev/null
+++ b/tasks/task_016_comprehensive_decision_hard_hard010.md
@@ -0,0 +1,117 @@
+---
+id: task_016_comprehensive_decision_hard_hard010
+name: comprehensive_decision-hard-hard010
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard010.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, to quantify the comprehensive competitive strength of the construction industry across regions, an industry association constructed a provincial competitiveness index system. The index is composed of four weighted sub-dimensions: market size share of national total (weight 30%), asset operation efficiency i.e. operating profit to total assets ratio (weight 30%), technology accumulation level i.e. cumulative invention patent grants to number of enterprises in jurisdiction ratio (weight 20%), and talent structure i.e. R&D personnel as share of total employees (weight 20%). The four raw indicators are each min-max normalized across all valid provinces, then weighted and summed to obtain the final index. Only provinces with data records for all four indicators are included in the calculation. Finally, please calculate the index difference between the first-ranked province and the last-ranked province in the competitiveness index ranking (rounded to two decimal places).
+
+Output guidelines:
+The answer should be a numeric value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.71`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_017_comprehensive_decision_hard_hard011.md b/tasks/task_017_comprehensive_decision_hard_hard011.md
new file mode 100644
index 0000000000000000000000000000000000000000..f6f575edb530e62b3052f272cd5828dacc9a1485
--- /dev/null
+++ b/tasks/task_017_comprehensive_decision_hard_hard011.md
@@ -0,0 +1,117 @@
+---
+id: task_017_comprehensive_decision_hard_hard011
+name: comprehensive_decision-hard-hard011
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard011.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, a think tank was commissioned to study the impact of policy intervention on R&D behavior in the communication transmission equipment industry. The research design divides all enterprises with R&D investment ratio data records into two groups: one group from provinces that have appeared in policy release information with "Communication Transmission Equipment" related policy entries ("National" level entries do not count as provinces and are not included in either group); the other group from provinces that have never appeared in the above policy entries. After grouping, calculate the arithmetic mean of R&D investment ratio for each group respectively, then compute the difference between them (policy-covered provinces mean minus non-policy-covered provinces mean). This difference reflects the association between policy coverage and R&D intensity of communication transmission equipment enterprises within the jurisdiction. What is this difference in percentage points?
+
+Output guidelines:
+The answer should be a numeric value with 2 decimal places. A positive number indicates policy-covered provinces are higher; a negative number indicates non-policy-covered provinces are higher. Output only the number, without units or text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`4.9`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_018_comprehensive_decision_hard_hard012.md b/tasks/task_018_comprehensive_decision_hard_hard012.md
new file mode 100644
index 0000000000000000000000000000000000000000..8a774372544a2f3d7f7adbb80c460e678879350f
--- /dev/null
+++ b/tasks/task_018_comprehensive_decision_hard_hard012.md
@@ -0,0 +1,117 @@
+---
+id: task_018_comprehensive_decision_hard_hard012
+name: comprehensive_decision-hard-hard012
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard012.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, an antitrust research team analyzed the provincial market structure of the metal smelting and rolling processing industry. To ensure statistical reliability, only provinces with operating revenue amount records and at least 5 enterprises in the industry within the jurisdiction were included. Among qualifying provinces, the Herfindahl-Hirschman Index (HHI) was used to measure market concentration in each province: calculate each enterprise's operating revenue as a share of total operating revenue of all valid enterprises in the province, sum the squares of these shares and multiply by 100% to obtain the province's HHI value. Higher HHI indicates more concentrated markets and greater monopoly risk. After identifying the province with the highest HHI, extract the province's total operating profit amount and total operating revenue amount from provincial industry summary data, and calculate the corresponding operating profit margin. What is the operating profit margin of the province with the highest HHI?
+
+Output guidelines:
+The answer should be a percentage value with 2 decimal places. Output only the number, without the % symbol or text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`4.14`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_019_comprehensive_decision_hard_hard013.md b/tasks/task_019_comprehensive_decision_hard_hard013.md
new file mode 100644
index 0000000000000000000000000000000000000000..0500eb46c554913a6c5e36e851e62e8b2394cfe3
--- /dev/null
+++ b/tasks/task_019_comprehensive_decision_hard_hard013.md
@@ -0,0 +1,119 @@
+---
+id: task_019_comprehensive_decision_hard_hard013
+name: comprehensive_decision-hard-hard013
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard013.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, a provincial industry and information department sought to evaluate the government subsidy utilization efficiency of enterprises of different sizes in the rubber and plastic products industry. After dividing enterprises into three groups by total assets—large (top 1/3 rounded up), medium (middle 1/3 rounded up), and small (bottom 1/3)—which enterprise size group has the highest subsidy utilization efficiency? What is its efficiency value?
+
+Output guidelines:
+The answer should be a numeric value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["Large", 166.33]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_020_comprehensive_decision_hard_hard014.md b/tasks/task_020_comprehensive_decision_hard_hard014.md
new file mode 100644
index 0000000000000000000000000000000000000000..744d409235abb46a5aa568cfba4c4fe5c72c0c71
--- /dev/null
+++ b/tasks/task_020_comprehensive_decision_hard_hard014.md
@@ -0,0 +1,117 @@
+---
+id: task_020_comprehensive_decision_hard_hard014
+name: comprehensive_decision-hard-hard014
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard014.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, a technology innovation fund evaluated the "R&D-patent conversion" full-chain efficiency of the consumer electronics and electrical industry across provinces, seeking to identify the province with optimal conversion efficiency (only provinces with valid enterprise count >= 3 are included). What is the R&D-patent conversion efficiency value of that province? (R&D-patent conversion efficiency = sum of annual Chinese invention patent grants / sum of annual Chinese invention patent applications × R&D output density; R&D output density = sum of annual Chinese invention patent applications / sum of R&D investment amount (100 million yuan))
+
+Output guidelines:
+The answer should be a numeric value with 2 decimal places. Output only the number, without units or text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`47.29`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_021_comprehensive_decision_hard_hard015.md b/tasks/task_021_comprehensive_decision_hard_hard015.md
new file mode 100644
index 0000000000000000000000000000000000000000..9cf9b0121a3201dd8a37f0c461606d7cc97e7958
--- /dev/null
+++ b/tasks/task_021_comprehensive_decision_hard_hard015.md
@@ -0,0 +1,117 @@
+---
+id: task_021_comprehensive_decision_hard_hard015
+name: comprehensive_decision-hard-hard015
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard015.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, a provincial government evaluated the comprehensive financial health of real estate enterprises to decide which provinces (where the province has an effective enterprise count >= 3) should face strengthened risk supervision for real estate firms. What is the health score of the province with the lowest financial health? (Financial health = Profitability score × 0.4 + Solvency score × 0.3 + Growth capability score × 0.3; Profitability is measured by the average net profit margin of enterprises in that province, where net profit margin = net profit amount / operating revenue amount; Solvency is measured as 1 − the arithmetic mean of enterprises' asset-liability ratio in that province / 100; Growth capability is measured as the median of enterprises' year-over-year change in operating revenue in that province / 100; each indicator is min-max normalized across all valid provinces before being substituted into the formula.)
+
+Output guidelines:
+The answer should be a numerical value with 2 decimal places. Output only the number without units or text explanation. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.07`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_022_comprehensive_decision_hard_hard016.md b/tasks/task_022_comprehensive_decision_hard_hard016.md
new file mode 100644
index 0000000000000000000000000000000000000000..dfd1c742fed30471ae069d88bb07080e594ce6f7
--- /dev/null
+++ b/tasks/task_022_comprehensive_decision_hard_hard016.md
@@ -0,0 +1,119 @@
+---
+id: task_022_comprehensive_decision_hard_hard016
+name: comprehensive_decision-hard-hard016
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard016.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, an institutional investor plans to build an equity portfolio among Haishan Chang Industrial Equipment Company, Zhongbai Jinmao Chain Company, and Sansan Dateng Heavy Industry Company. The total portfolio weight must equal 1, the portfolio-weighted asset-liability ratio must equal exactly 45%, and the portfolio-weighted year-on-year operating revenue change must equal exactly 0%. Based on these three companies' 2022 operating data, find their portfolio weights and compute the portfolio-weighted ROE. Note: each company's asset-liability ratio is computed as total liabilities ÷ total assets × 100%.
+
+Output guidelines:
+Answer format: weight of Haishan Chang Industrial Equipment Company, weight of Zhongbai Jinmao Chain Company, weight of Sansan Dateng Heavy Industry Company, weighted ROE. The first three weights to four decimal places; weighted ROE to three decimal places. Output numbers and commas only, with no explanatory text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[0.1318, 0.4954, 0.3728, 10.376]`
+
+Scoring rules:
+- The gold answer is a list with N=4 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_3` each as 0 or 1.
+- Return `total = (sum(part_i)) / 4` exactly.
+- If the model output is missing or cannot be parsed into 4 comparable parts, score all parts 0.
+
diff --git a/tasks/task_023_comprehensive_decision_hard_hard017.md b/tasks/task_023_comprehensive_decision_hard_hard017.md
new file mode 100644
index 0000000000000000000000000000000000000000..b7c001e4ff7c09c4ff719e069e97c711c6a3468a
--- /dev/null
+++ b/tasks/task_023_comprehensive_decision_hard_hard017.md
@@ -0,0 +1,117 @@
+---
+id: task_023_comprehensive_decision_hard_hard017
+name: comprehensive_decision-hard-hard017
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard017.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in the chemical raw materials and chemical products manufacturing industry covered by the Implementation Plan for "Three Products" in Raw Materials Industry, Hualu Runyuan Technology Co., Ltd. plans to restore profitability through product upgrade and price increases. After implementing the "Three Products" reforms, the company can obtain two types of certain profit improvements: one from process and quality improvement, equal to 1.5% of that year's operating revenue; the other from special support and subsidies. Assuming sales volume is unchanged, price increases have no effect on costs, and the goal is to bring net profit exactly to zero, find the minimum price increase rate required based on the company's 2022 operating data and policy information.
+
+Output guidelines:
+Answer format: minimum price increase rate. Four decimal places. Output the number only, no percent sign or text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`14.2792`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_024_comprehensive_decision_hard_hard018.md b/tasks/task_024_comprehensive_decision_hard_hard018.md
new file mode 100644
index 0000000000000000000000000000000000000000..09b0dd70aaf0cbc3658af494937100478ba504f2
--- /dev/null
+++ b/tasks/task_024_comprehensive_decision_hard_hard018.md
@@ -0,0 +1,117 @@
+---
+id: task_024_comprehensive_decision_hard_hard018
+name: comprehensive_decision-hard-hard018
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard018.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, does Lianji Chuangji Machine Tool Company meet the basic conditions under the Several Policies on Supporting the Construction of a Strong Province of Skilled Workers? If yes, treat it as a sample firm eligible for skills training subsidies and apply the following: all R&D personnel receive individual skill improvement subsidies at the "senior technician" rate; all non-R&D employees at the "technician" rate; assume all new subsidies are used to offset that year's R&D investment. Calculate by how many basis points the adjusted R&D investment ratio is lower than the disclosed 2022 R&D investment ratio.
+
+Output guidelines:
+Answer format: the number of basis points of decline. Two decimal places. Output the number only, no units or text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"43.33"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_025_comprehensive_decision_hard_hard019.md b/tasks/task_025_comprehensive_decision_hard_hard019.md
new file mode 100644
index 0000000000000000000000000000000000000000..81e399b93bf8090f19abd0feb45fcf7b0c5a103f
--- /dev/null
+++ b/tasks/task_025_comprehensive_decision_hard_hard019.md
@@ -0,0 +1,119 @@
+---
+id: task_025_comprehensive_decision_hard_hard019
+name: comprehensive_decision-hard-hard019
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/hard019.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，以山安泽医疗科技公司具有“微生态活菌业务”和“高温合金业务”双主业特征。某基金经理希望用以山安泽医疗科技公司与三三达腾重工公司构建一个两股票组合，来替代连机创机机床公司的增长暴露，并进一步检验该替代组合在剔除补贴后的盈利质量与研发强度溢价。若组合要求加权营业收入同比增减幅恰好等于连机创机机床公司2022年的对应指标，请基于本地数据计算：以山安泽医疗科技公司的组合权重、剔除政府奖励资金和补贴后的组合加权净利率，以及该组合研发投入占比相对连机创机机床公司高出的基点数。
+
+Output guidelines:
+答案格式为：以山安泽医疗科技公司权重,剔除补贴后的组合加权净利率,研发投入占比高出的基点数。前两项按百分比口径保留2位小数，最后一项保留2位小数。仅输出数字和逗号，不要添加单位或文字说明。如无法找到相关数据，请回答“未查询到相关数据”。
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[38.58, 14.82, 708.28]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_026_comprehensive_decision_medium_medium001.md b/tasks/task_026_comprehensive_decision_medium_medium001.md
new file mode 100644
index 0000000000000000000000000000000000000000..2d8626be1e6cee14499ff34bef50a2fed2b3e76e
--- /dev/null
+++ b/tasks/task_026_comprehensive_decision_medium_medium001.md
@@ -0,0 +1,119 @@
+---
+id: task_026_comprehensive_decision_medium_medium001
+name: comprehensive_decision-medium-medium001
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium001.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+For 2022 pharmaceutical manufacturing industry data by province, if R&D funding intensity is measured as each province's total R&D expenditure as a percentage of its total operating revenue, among all provinces with complete data records, what is the specific value of this ratio for the province with the highest level? Which company has the highest R&D funding intensity in that province?
+
+Output guidelines:
+The first answer is a numeric value (2 decimal places), unit is %; the second answer is the full company name, which must exactly match the "Company Name" field in company_profile.csv. If either question cannot be answered, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[25.48, "Kangsheng Anjian Biopharmaceutical Company"]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_027_comprehensive_decision_medium_medium002.md b/tasks/task_027_comprehensive_decision_medium_medium002.md
new file mode 100644
index 0000000000000000000000000000000000000000..16e6e1f10bcbc67a80d323f5767fab38f4eaca93
--- /dev/null
+++ b/tasks/task_027_comprehensive_decision_medium_medium002.md
@@ -0,0 +1,119 @@
+---
+id: task_027_comprehensive_decision_medium_medium002
+name: comprehensive_decision-medium-medium002
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium002.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, a semiconductor company plans to expand production and wishes to locate in the province with the highest enterprise concentration to gain industrial synergy effects. What is the proportion of that province's semiconductor industry enterprise count to the national total? What proportion does that province's total operating profit in the semiconductor industry account for of the national semiconductor industry's total operating profit?
+
+Output guidelines:
+Two answers required, both numeric values (2 decimal places), unit is %. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[31.4, 6.22]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_028_comprehensive_decision_medium_medium003.md b/tasks/task_028_comprehensive_decision_medium_medium003.md
new file mode 100644
index 0000000000000000000000000000000000000000..a77f84b032b328644a0e228dc7cee628d1a91ee2
--- /dev/null
+++ b/tasks/task_028_comprehensive_decision_medium_medium003.md
@@ -0,0 +1,119 @@
+---
+id: task_028_comprehensive_decision_medium_medium003
+name: comprehensive_decision-medium-medium003
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium003.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, measuring per capita output efficiency of automobile manufacturing by province using revenue per capita (total operating revenue ÷ total employee count), among all provinces nationwide, what is the specific value in yuan per person for the province with the highest indicator? Compared to the national average, by what percentage (1 decimal place) is that province's per capita revenue higher?
+
+Output guidelines:
+Two answers required: first is a numeric value (2 decimal places), unit yuan/person; second is a percentage (1 decimal place), unit %. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[3898878.23, 175.0]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_029_comprehensive_decision_medium_medium004.md b/tasks/task_029_comprehensive_decision_medium_medium004.md
new file mode 100644
index 0000000000000000000000000000000000000000..25639c5c3b762e4d7893a1279e6b432da3ceb6b1
--- /dev/null
+++ b/tasks/task_029_comprehensive_decision_medium_medium004.md
@@ -0,0 +1,119 @@
+---
+id: task_029_comprehensive_decision_medium_medium004
+name: comprehensive_decision-medium-medium004
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium004.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, rank provinces by profitability in chemical raw materials and chemical products manufacturing. Provincial operating profit margin is computed as total operating profit divided by total operating revenue. Using this as the ranking criterion, what is Guangdong Province's rank? Apply the same ranking to all relevant enterprises within Guangdong Province—which enterprise ranks first?
+
+Output guidelines:
+Two answers required: first is Guangdong Province's rank in the provincial ranking (integer, e.g. "6" means 6th place); second is the full name of the top-ranked enterprise in Guangdong, which must match the "Company Name" field in company_profile.csv. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[6, "Hengyi Changhua Technology Co., Ltd."]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_030_comprehensive_decision_medium_medium005.md b/tasks/task_030_comprehensive_decision_medium_medium005.md
new file mode 100644
index 0000000000000000000000000000000000000000..abcb1dff1a441bb8f567ca31338731023b4807a5
--- /dev/null
+++ b/tasks/task_030_comprehensive_decision_medium_medium005.md
@@ -0,0 +1,117 @@
+---
+id: task_030_comprehensive_decision_medium_medium005
+name: comprehensive_decision-medium-medium005
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium005.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, from all private enterprises in the food and beverage industry, aggregate government rewards and subsidy amounts by province of registration to find the province with the highest provincial subsidy total. How many hundred million yuan in government subsidies did private enterprises in that province receive in total?
+
+Output guidelines:
+The answer should be a numerical value (2 decimal places), unit is hundred million yuan. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`12.6`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_031_comprehensive_decision_medium_medium006.md b/tasks/task_031_comprehensive_decision_medium_medium006.md
new file mode 100644
index 0000000000000000000000000000000000000000..7e0b5d0511488ab99f0d55cc880602978cae95fe
--- /dev/null
+++ b/tasks/task_031_comprehensive_decision_medium_medium006.md
@@ -0,0 +1,117 @@
+---
+id: task_031_comprehensive_decision_medium_medium006
+name: comprehensive_decision-medium-medium006
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium006.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, a local government planned to introduce support policies for the specialized equipment manufacturing industry and needed to understand the province's R&D personnel investment level in this industry. Is the average proportion of R&D personnel to total employees in specialized equipment manufacturing enterprises in Zhejiang Province higher than the national average for this industry?
+
+Output guidelines:
+The answer should be "Yes" or "No". If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_032_comprehensive_decision_medium_medium007.md b/tasks/task_032_comprehensive_decision_medium_medium007.md
new file mode 100644
index 0000000000000000000000000000000000000000..15ca9956975f897dfb74f840f2f41c9f80e820da
--- /dev/null
+++ b/tasks/task_032_comprehensive_decision_medium_medium007.md
@@ -0,0 +1,119 @@
+---
+id: task_032_comprehensive_decision_medium_medium007
+name: comprehensive_decision-medium-medium007
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium007.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, analyze government subsidy leverage in the information transmission, software and information technology services industry. Define government subsidy leverage effect as the ratio of each province's total operating profit to total government subsidies. What is the specific ratio for the province with the highest government subsidy leverage effect? Which enterprise in that province has the highest leverage effect?
+
+Output guidelines:
+Two answers required: first is a numeric value (2 decimal places, unitless ratio); second is the full company name. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[67.19, "Dongche Kexin Systems Company"]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_033_comprehensive_decision_medium_medium008.md b/tasks/task_033_comprehensive_decision_medium_medium008.md
new file mode 100644
index 0000000000000000000000000000000000000000..7e53ce48a59875899268553cdb917fa4392efc8b
--- /dev/null
+++ b/tasks/task_033_comprehensive_decision_medium_medium008.md
@@ -0,0 +1,117 @@
+---
+id: task_033_comprehensive_decision_medium_medium008
+name: comprehensive_decision-medium-medium008
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium008.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, to study the capital turnover of central state-owned enterprises in the electricity, heat, gas and water production and supply industry, calculate the asset turnover ratio for each enterprise by dividing its annual operating revenue by its total assets. Find the arithmetic mean of the asset turnover ratios for these enterprises.
+
+Output guidelines:
+The answer should be a numerical value (rounded to 4 decimal places). If the relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.3266`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_034_comprehensive_decision_medium_medium009.md b/tasks/task_034_comprehensive_decision_medium_medium009.md
new file mode 100644
index 0000000000000000000000000000000000000000..fb4e1db556133ff5f269301852da84235030eb82
--- /dev/null
+++ b/tasks/task_034_comprehensive_decision_medium_medium009.md
@@ -0,0 +1,117 @@
+---
+id: task_034_comprehensive_decision_medium_medium009
+name: comprehensive_decision-medium-medium009
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium009.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, a mining industry enterprise plans to go public for financing and hopes to list on the exchange with the highest average market capitalization among enterprises in this industry. Among the four exchanges—Shenzhen Stock Exchange, Hong Kong Stock Exchange, Shanghai Stock Exchange, and Beijing Stock Exchange—which exchange has the highest average market capitalization of listed mining enterprises?
+
+Output guidelines:
+The answer should be the exchange name (e.g., "Shenzhen Stock Exchange", "Hong Kong Stock Exchange", "Shanghai Stock Exchange", "Beijing Stock Exchange"). If the relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Shanghai Stock Exchange"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_035_comprehensive_decision_medium_medium010.md b/tasks/task_035_comprehensive_decision_medium_medium010.md
new file mode 100644
index 0000000000000000000000000000000000000000..d71a559307d4dc42b8aa4876c0cb0e7746e33457
--- /dev/null
+++ b/tasks/task_035_comprehensive_decision_medium_medium010.md
@@ -0,0 +1,119 @@
+---
+id: task_035_comprehensive_decision_medium_medium010
+name: comprehensive_decision-medium-medium010
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium010.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in the provincial data for the construction industry, each province has an indicator reflecting the average asset-liability ratio (financial leverage level) of enterprises in that province's industry (considering only enterprises with valid total assets and total liabilities). Among the provinces covered by valid data, which province has the lowest value for this mean indicator, and what is that value?
+
+Output guidelines:
+The answer should be a numerical value (rounded to 2 decimal places), in units of %. If the relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["Shanxi Province", 27.17]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_036_comprehensive_decision_medium_medium011.md b/tasks/task_036_comprehensive_decision_medium_medium011.md
new file mode 100644
index 0000000000000000000000000000000000000000..f1a3311f6eeab5eb500387dfb9a976e040e3cae6
--- /dev/null
+++ b/tasks/task_036_comprehensive_decision_medium_medium011.md
@@ -0,0 +1,117 @@
+---
+id: task_036_comprehensive_decision_medium_medium011
+name: comprehensive_decision-medium-medium011
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium011.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, among all enterprises in the rubber and plastic products industry with R&D investment records, what is the R&D concentration CR5?
+
+Output guidelines:
+The answer should be a numerical value (rounded to 2 decimal places), in units of %. If the relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`34.66`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_037_comprehensive_decision_medium_medium012.md b/tasks/task_037_comprehensive_decision_medium_medium012.md
new file mode 100644
index 0000000000000000000000000000000000000000..68993bd2325d5cb79850ee698ee5c1525cb3c612
--- /dev/null
+++ b/tasks/task_037_comprehensive_decision_medium_medium012.md
@@ -0,0 +1,119 @@
+---
+id: task_037_comprehensive_decision_medium_medium012
+name: comprehensive_decision-medium-medium012
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium012.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, among all enterprises in Guangdong Province belonging to the wholesale and retail trade industry, using each enterprise's net profit margin as the comparison standard, what is the indicator value for the enterprise with the highest net profit margin? What is that enterprise's rank among all enterprises in this industry nationwide?
+
+Output guidelines:
+Two answers required: first is a numeric value (2 decimal places, unit %); second is the rank number (integer, e.g. "7" means 7th place). If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[31.25, 7]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_038_comprehensive_decision_medium_medium013.md b/tasks/task_038_comprehensive_decision_medium_medium013.md
new file mode 100644
index 0000000000000000000000000000000000000000..f0152a0894f0778fac5d66cf3a11b07fe6433871
--- /dev/null
+++ b/tasks/task_038_comprehensive_decision_medium_medium013.md
@@ -0,0 +1,119 @@
+---
+id: task_038_comprehensive_decision_medium_medium013
+name: comprehensive_decision-medium-medium013
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium013.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, a scientific research and technical services enterprise wishes to identify the province with the fastest net profit growth in the industry to guide market expansion. What is the indicator value for the province with the highest median year-on-year net profit growth rate in the national scientific research and technical services industry? What is the nationwide rank of the enterprise with the highest such indicator in that province among all enterprises in this industry?
+
+Output guidelines:
+Two answers required: first is a numeric value (2 decimal places, unit %), i.e. the indicator value for the province with the highest "median year-on-year net profit growth rate" in this industry nationwide; second is a rank number (integer, e.g. "23" means 23rd place), i.e. the nationwide rank of the enterprise with the highest "year-on-year net profit growth rate" in that province among all enterprises in this industry. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[13.81, 23]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_039_comprehensive_decision_medium_medium014.md b/tasks/task_039_comprehensive_decision_medium_medium014.md
new file mode 100644
index 0000000000000000000000000000000000000000..16066f6b6bdc7374b2b1d84f97ad5160e0fb5cd3
--- /dev/null
+++ b/tasks/task_039_comprehensive_decision_medium_medium014.md
@@ -0,0 +1,119 @@
+---
+id: task_039_comprehensive_decision_medium_medium014
+name: comprehensive_decision-medium-medium014
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium014.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, for the metal smelting and rolling processing industry, among provinces with valid records for both total government subsidies and total industry employee count, per capita subsidy is computed as each province's total government rewards and subsidies divided by that province's industry employee count. What is the per capita subsidy in yuan for the province with the highest per capita subsidy? What is the nationwide rank of the enterprise with the highest such indicator in that province among all enterprises in this industry?
+
+Output guidelines:
+Two answers required: first is a numeric value (2 decimal places, unit yuan/person), i.e. the highest provincial per capita subsidy; second is a rank number (integer), indicating the nationwide rank of the enterprise with the highest indicator in that province among all enterprises in this industry. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[17569.95, 12]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_040_comprehensive_decision_medium_medium015.md b/tasks/task_040_comprehensive_decision_medium_medium015.md
new file mode 100644
index 0000000000000000000000000000000000000000..e0ee5884201e5ca5e3888c1ea5e6aeb91ed250e2
--- /dev/null
+++ b/tasks/task_040_comprehensive_decision_medium_medium015.md
@@ -0,0 +1,119 @@
+---
+id: task_040_comprehensive_decision_medium_medium015
+name: comprehensive_decision-medium-medium015
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium015.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+List the 2022 indicators for which Shandong Province's financial industry enterprise averages are below the national financial industry medians.
+
+Output guidelines:
+The answer must list all qualifying indicator names, separated by semicolons. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["Year-on-year R&D personnel growth rate", "Year-on-year operating profit growth rate", "Year-on-year net profit growth rate", "Year-on-year employee growth rate", "Capitalized R&D expenditure", "Year-on-year capitalized R&D expenditure growth rate", "Annual PCT patent applications", "Annual PCT invention patent applications", "Provincial/ministerial science and technology progress award", "Participation in drafting national standards", "Participation in drafting industry standards", "Annual Chinese patent applications", "Annual Chinese invention patent applications", "Annual Chinese patent grants", "Cumulative Chinese invention patent applications", "Annual Chinese invention patent grants", "Cumulative PCT patent applications", "Cumulative PCT invention patent applications", "Cumulative Chinese patent applications", "Cumulative Chinese invention patent grants", "Cumulative patent citations", "R&D personnel ratio", "R&D personnel count", "Year-on-year R&D expenditure growth rate", "Cumulative Chinese invention patent lapses", "Company market value", "Asset-liability ratio", "Total employee count"]`
+
+Scoring rules:
+- The gold answer is a list with N=28 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_27` each as 0 or 1.
+- Return `total = (sum(part_i)) / 28` exactly.
+- If the model output is missing or cannot be parsed into 28 comparable parts, score all parts 0.
+
diff --git a/tasks/task_041_comprehensive_decision_medium_medium016.md b/tasks/task_041_comprehensive_decision_medium_medium016.md
new file mode 100644
index 0000000000000000000000000000000000000000..1512fbb5a608a6cbe292771b538ea4f14f49d735
--- /dev/null
+++ b/tasks/task_041_comprehensive_decision_medium_medium016.md
@@ -0,0 +1,117 @@
+---
+id: task_041_comprehensive_decision_medium_medium016
+name: comprehensive_decision-medium-medium016
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium016.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, between Sichuan Province's top enterprise by operating revenue and Shandong Province's top enterprise by net profit in the pharmaceutical manufacturing industry, which one has the highest tax payment?
+
+Output guidelines:
+The answer must be the company name. Output only the company name, with no additional text. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No relevant data found"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_042_comprehensive_decision_medium_medium017.md b/tasks/task_042_comprehensive_decision_medium_medium017.md
new file mode 100644
index 0000000000000000000000000000000000000000..43fbf4ab857090f802828fd92eae9da4badc3afa
--- /dev/null
+++ b/tasks/task_042_comprehensive_decision_medium_medium017.md
@@ -0,0 +1,117 @@
+---
+id: task_042_comprehensive_decision_medium_medium017
+name: comprehensive_decision-medium-medium017
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium017.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, the industry where Zhongbai Jinmao Chain Company operates, is the enterprise with the highest year-on-year R&D expenditure growth rate also the one with the highest R&D expenditure?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only the answer, with no additional text. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_043_comprehensive_decision_medium_medium018.md b/tasks/task_043_comprehensive_decision_medium_medium018.md
new file mode 100644
index 0000000000000000000000000000000000000000..0b966200d7fd8a68565f316a4421ff7a1cab7a65
--- /dev/null
+++ b/tasks/task_043_comprehensive_decision_medium_medium018.md
@@ -0,0 +1,117 @@
+---
+id: task_043_comprehensive_decision_medium_medium018
+name: comprehensive_decision-medium-medium018
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium018.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022 Nationwide, is the province with the highest R&D expenditure growth rate also the province with the lowest R&D expenditure?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only the answer, with no additional text. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_044_comprehensive_decision_medium_medium019.md b/tasks/task_044_comprehensive_decision_medium_medium019.md
new file mode 100644
index 0000000000000000000000000000000000000000..af4a825009efeff07c2770476ae002e7e6a3552e
--- /dev/null
+++ b/tasks/task_044_comprehensive_decision_medium_medium019.md
@@ -0,0 +1,117 @@
+---
+id: task_044_comprehensive_decision_medium_medium019
+name: comprehensive_decision-medium-medium019
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium019.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, Sichuan Province, is Zhongbai Jinmao Chain Company's R&D expenditure higher than the R&D expenditure of the enterprise ranked 15th nationwide in its industry?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only the answer, with no additional text. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_045_comprehensive_decision_medium_medium020.md b/tasks/task_045_comprehensive_decision_medium_medium020.md
new file mode 100644
index 0000000000000000000000000000000000000000..2c7a66f0447f8edba2304e930ee1d3da3eb8b434
--- /dev/null
+++ b/tasks/task_045_comprehensive_decision_medium_medium020.md
@@ -0,0 +1,117 @@
+---
+id: task_045_comprehensive_decision_medium_medium020
+name: comprehensive_decision-medium-medium020
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium020.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, the chemical raw materials and chemical products manufacturing industry in Shandong Province, is the market capitalization of the leading enterprises by operating revenue among the top 3 in this industry nationwide?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only the answer, with no additional text. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_046_comprehensive_decision_medium_medium021.md b/tasks/task_046_comprehensive_decision_medium_medium021.md
new file mode 100644
index 0000000000000000000000000000000000000000..418dff11e2434e2aa649349b44db9fe040f8a9fb
--- /dev/null
+++ b/tasks/task_046_comprehensive_decision_medium_medium021.md
@@ -0,0 +1,117 @@
+---
+id: task_046_comprehensive_decision_medium_medium021
+name: comprehensive_decision-medium-medium021
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium021.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, the chemical raw materials and chemical products manufacturing industry in Shandong Province, is the market capitalization of the enterprise with the highest operating revenue among the top 3 in this industry nationwide?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only the answer, with no additional text. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_047_comprehensive_decision_medium_medium022.md b/tasks/task_047_comprehensive_decision_medium_medium022.md
new file mode 100644
index 0000000000000000000000000000000000000000..7cbbc3307beda76374a0bf5b36d06aed2bf5bc4d
--- /dev/null
+++ b/tasks/task_047_comprehensive_decision_medium_medium022.md
@@ -0,0 +1,117 @@
+---
+id: task_047_comprehensive_decision_medium_medium022
+name: comprehensive_decision-medium-medium022
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium022.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022 is the region with the highest average R&D investment growth rate also the one with the most policy quantity?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only the answer, with no additional text. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_048_comprehensive_decision_medium_medium023.md b/tasks/task_048_comprehensive_decision_medium_medium023.md
new file mode 100644
index 0000000000000000000000000000000000000000..2fc20e5ffbee96bd42e8f13fc9d5940c260d466c
--- /dev/null
+++ b/tasks/task_048_comprehensive_decision_medium_medium023.md
@@ -0,0 +1,117 @@
+---
+id: task_048_comprehensive_decision_medium_medium023
+name: comprehensive_decision-medium-medium023
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium023.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022 is the average year-on-year net profit growth rate for the industry where Haishan Changgong Equipment Company operates higher than the average R&D investment growth rate?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only the answer, with no additional text. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_049_comprehensive_decision_medium_medium024.md b/tasks/task_049_comprehensive_decision_medium_medium024.md
new file mode 100644
index 0000000000000000000000000000000000000000..32b56a03d139ec7f87775cb9d40275244ea12d22
--- /dev/null
+++ b/tasks/task_049_comprehensive_decision_medium_medium024.md
@@ -0,0 +1,117 @@
+---
+id: task_049_comprehensive_decision_medium_medium024
+name: comprehensive_decision-medium-medium024
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium024.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the region with the highest total operating revenue nationwide also the one with the largest total enterprise market capitalization?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only the answer, with no additional text. If relevant data cannot be found, respond with "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_050_comprehensive_decision_medium_medium025.md b/tasks/task_050_comprehensive_decision_medium_medium025.md
new file mode 100644
index 0000000000000000000000000000000000000000..99ccf8abbcfa7232365e4e8cf7c291eae6946fa8
--- /dev/null
+++ b/tasks/task_050_comprehensive_decision_medium_medium025.md
@@ -0,0 +1,117 @@
+---
+id: task_050_comprehensive_decision_medium_medium025
+name: comprehensive_decision-medium-medium025
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium025.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the industry with the highest R&D investment growth rate also the industry with the largest R&D investment?
+
+Output guidelines:
+The answer must be "Yes" or "No". Only output the answer, do not add any explanation. If the relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_051_comprehensive_decision_medium_medium026.md b/tasks/task_051_comprehensive_decision_medium_medium026.md
new file mode 100644
index 0000000000000000000000000000000000000000..2a4be4496ba1c58603fdf11316ac4092760413b5
--- /dev/null
+++ b/tasks/task_051_comprehensive_decision_medium_medium026.md
@@ -0,0 +1,117 @@
+---
+id: task_051_comprehensive_decision_medium_medium026
+name: comprehensive_decision-medium-medium026
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium026.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, does the industry with the highest average asset-liability ratio in Guangdong Province belong to the industry with the highest average asset-liability ratio nationwide?
+
+Output guidelines:
+The answer must be "Yes" or "No". Only output the answer, do not add any explanation. If the relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_052_comprehensive_decision_medium_medium027.md b/tasks/task_052_comprehensive_decision_medium_medium027.md
new file mode 100644
index 0000000000000000000000000000000000000000..4abb58bbe3c24d16df103a9f006c834a387b2e89
--- /dev/null
+++ b/tasks/task_052_comprehensive_decision_medium_medium027.md
@@ -0,0 +1,117 @@
+---
+id: task_052_comprehensive_decision_medium_medium027
+name: comprehensive_decision-medium-medium027
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium027.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, Compare the enterprise with the highest R&D input-output ratio in Guangdong Province and the enterprise with the highest R&D input-output ratio in Sichuan Province (R&D input-output ratio = operating revenue amount / R&D investment amount). Which enterprise has higher operating revenue? The unit of operating revenue is yuan.
+
+Output guidelines:
+The answer must be "yes" or "no". Output only the answer, without any additional explanation. If the relevant data cannot be found, please answer "No relevant data retrieved".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"The enterprise with the highest R&D input-output ratio in Guangdong Province (Wulichangyuan Wholesale Company) has higher operating revenue"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_053_comprehensive_decision_medium_medium028.md b/tasks/task_053_comprehensive_decision_medium_medium028.md
new file mode 100644
index 0000000000000000000000000000000000000000..807e3cc3bfd1aeb9fd31c4ad3ad3cc674066ed24
--- /dev/null
+++ b/tasks/task_053_comprehensive_decision_medium_medium028.md
@@ -0,0 +1,119 @@
+---
+id: task_053_comprehensive_decision_medium_medium028
+name: comprehensive_decision-medium-medium028
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium028.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, for the enterprise ranked first in year-over-year growth rate of R&D investment in the chemical raw materials and chemical products manufacturing industry, which of its indicators perform below the industry average?
+
+Output guidelines:
+Output only the indicator name(s) as the answer, without any additional explanation. If the relevant data cannot be found, please answer "No relevant data retrieved"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["Total assets", "R&D investment ratio", "Capitalized R&D investment", "Year-over-year change in capitalized R&D investment", "Company market capitalization", "Cumulative China patent applications", "Cumulative China invention patent grants", "Cumulative China invention patent invalidations", "Annual China patent applications", "Annual China invention patent applications", "Annual China patent grants", "Annual China invention patent grants", "R&D personnel ratio", "Total liabilities", "Cumulative citations of all patents", "Cumulative China invention patent applications", "Participation in drafting national standards"]`
+
+Scoring rules:
+- The gold answer is a list with N=17 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_16` each as 0 or 1.
+- Return `total = (sum(part_i)) / 17` exactly.
+- If the model output is missing or cannot be parsed into 17 comparable parts, score all parts 0.
+
diff --git a/tasks/task_054_comprehensive_decision_medium_medium029.md b/tasks/task_054_comprehensive_decision_medium_medium029.md
new file mode 100644
index 0000000000000000000000000000000000000000..d709333d04e843d8fa95d61a6d009ed7e1987d26
--- /dev/null
+++ b/tasks/task_054_comprehensive_decision_medium_medium029.md
@@ -0,0 +1,117 @@
+---
+id: task_054_comprehensive_decision_medium_medium029
+name: comprehensive_decision-medium-medium029
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium029.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Total number of all enterprises affected by the policy 'Notice of the General Office of Guangdong Provincial People's Government on Printing and Distributing Several Measures of Guangdong Province for Further Promoting Steady Growth of Industrial Economy' in 2022
+
+Output guidelines:
+The answer must be "yes" or "no". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`416`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_055_comprehensive_decision_medium_medium030.md b/tasks/task_055_comprehensive_decision_medium_medium030.md
new file mode 100644
index 0000000000000000000000000000000000000000..4d17d372422fa314388263202aaaf06727da41d3
--- /dev/null
+++ b/tasks/task_055_comprehensive_decision_medium_medium030.md
@@ -0,0 +1,117 @@
+---
+id: task_055_comprehensive_decision_medium_medium030
+name: comprehensive_decision-medium-medium030
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium030.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Did both the operating revenue and R&D investment of Haishan Changgong Equipment Company and Sansan Daten Heavy Industry Company show an upward trend in 2022?
+
+Output guidelines:
+The answer must be "yes" or "no". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"no"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_056_comprehensive_decision_medium_medium031.md b/tasks/task_056_comprehensive_decision_medium_medium031.md
new file mode 100644
index 0000000000000000000000000000000000000000..b35884d642a41f6707c2563c3ac3fef52d855bc1
--- /dev/null
+++ b/tasks/task_056_comprehensive_decision_medium_medium031.md
@@ -0,0 +1,117 @@
+---
+id: task_056_comprehensive_decision_medium_medium031
+name: comprehensive_decision-medium-medium031
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium031.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, was Haishan Changgong Equipment Company the only enterprise in its region with both increased operating revenue and R&D investment?
+
+Output guidelines:
+The answer must be "yes" or "no". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"no"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_057_comprehensive_decision_medium_medium032.md b/tasks/task_057_comprehensive_decision_medium_medium032.md
new file mode 100644
index 0000000000000000000000000000000000000000..d3aa98e2ad694bee04939c2d475dd4af217cef5e
--- /dev/null
+++ b/tasks/task_057_comprehensive_decision_medium_medium032.md
@@ -0,0 +1,117 @@
+---
+id: task_057_comprehensive_decision_medium_medium032
+name: comprehensive_decision-medium-medium032
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium032.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, was Haishan Changgong Equipment Company the only enterprise in its region with both declined operating revenue and R&D investment?
+
+Output guidelines:
+The answer must be "yes" or "no". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"no"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_058_comprehensive_decision_medium_medium033.md b/tasks/task_058_comprehensive_decision_medium_medium033.md
new file mode 100644
index 0000000000000000000000000000000000000000..337920246b53f8213f41cd77b03d1139850921bf
--- /dev/null
+++ b/tasks/task_058_comprehensive_decision_medium_medium033.md
@@ -0,0 +1,119 @@
+---
+id: task_058_comprehensive_decision_medium_medium033
+name: comprehensive_decision-medium-medium033
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium033.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In the industry with the most invention patent grants in 2022, what is the average asset-liability ratio of enterprises? In which province is this industry concentrated?
+
+Output guidelines:
+Answer data should retain two decimal places. Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[85.7, "Guangdong Province"]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_059_comprehensive_decision_medium_medium034.md b/tasks/task_059_comprehensive_decision_medium_medium034.md
new file mode 100644
index 0000000000000000000000000000000000000000..17b3acef6a40d3246dbe1ceb9f1763b2ed2d8686
--- /dev/null
+++ b/tasks/task_059_comprehensive_decision_medium_medium034.md
@@ -0,0 +1,117 @@
+---
+id: task_059_comprehensive_decision_medium_medium034
+name: comprehensive_decision-medium-medium034
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium034.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Among the provinces that performed best in cultivating high-tech enterprises in 2022 (measured by the number of high-tech enterprises), which high-tech enterprise has the highest year-over-year growth in average operating revenue?
+
+Output guidelines:
+The answer must be an enterprise name. Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Langji Ruanchuang Information Technology Company"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_060_comprehensive_decision_medium_medium035.md b/tasks/task_060_comprehensive_decision_medium_medium035.md
new file mode 100644
index 0000000000000000000000000000000000000000..fae64c02c631a548a751a4e316720bbad5e0559b
--- /dev/null
+++ b/tasks/task_060_comprehensive_decision_medium_medium035.md
@@ -0,0 +1,119 @@
+---
+id: task_060_comprehensive_decision_medium_medium035
+name: comprehensive_decision-medium-medium035
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium035.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which provinces ranked in the top three in terms of market share among enterprises in the cloud computing services sector?
+
+Output guidelines:
+The answer must be province names. Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["Beijing", "Shanghai", "Guangdong Province"]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_061_comprehensive_decision_medium_medium036.md b/tasks/task_061_comprehensive_decision_medium_medium036.md
new file mode 100644
index 0000000000000000000000000000000000000000..4b725409a94233cbce30b0f79d17888a316b7155
--- /dev/null
+++ b/tasks/task_061_comprehensive_decision_medium_medium036.md
@@ -0,0 +1,117 @@
+---
+id: task_061_comprehensive_decision_medium_medium036
+name: comprehensive_decision-medium-medium036
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium036.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, what percentage of total operating revenue in the market do the top 20% enterprises by R&D investment in Sichuan Province's pharmaceutical manufacturing industry account for?
+
+Output guidelines:
+The answer must be an exact number with 2 decimal places. Output only the number, do not add units, commas, or any explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`61.92`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_062_comprehensive_decision_medium_medium037.md b/tasks/task_062_comprehensive_decision_medium_medium037.md
new file mode 100644
index 0000000000000000000000000000000000000000..c17dc7ff2baf34272eec9ab9d2621f9948a079f4
--- /dev/null
+++ b/tasks/task_062_comprehensive_decision_medium_medium037.md
@@ -0,0 +1,117 @@
+---
+id: task_062_comprehensive_decision_medium_medium037
+name: comprehensive_decision-medium-medium037
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium037.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, what percentage of total operating revenue in the market do the top enterprises by R&D investment (top 3) in Sichuan Province's pharmaceutical manufacturing industry account for?
+
+Output guidelines:
+The answer must be an exact number with 2 decimal places. Output only the number, do not add units, commas, or any explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`61.92`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_063_comprehensive_decision_medium_medium038.md b/tasks/task_063_comprehensive_decision_medium_medium038.md
new file mode 100644
index 0000000000000000000000000000000000000000..f320a4cf2e26eda56310d2817d670bd29364b41e
--- /dev/null
+++ b/tasks/task_063_comprehensive_decision_medium_medium038.md
@@ -0,0 +1,117 @@
+---
+id: task_063_comprehensive_decision_medium_medium038
+name: comprehensive_decision-medium-medium038
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium038.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which enterprise has the highest operating revenue in the same region and industry as Sansan Daten Heavy Industry Company?
+
+Output guidelines:
+The answer must be an enterprise name. Output only the enterprise name, do not add any other explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Lingong Hangteng Heavy Industry Company"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_064_comprehensive_decision_medium_medium039.md b/tasks/task_064_comprehensive_decision_medium_medium039.md
new file mode 100644
index 0000000000000000000000000000000000000000..4b497c2e95581355f11a9326366ddc1736e1a052
--- /dev/null
+++ b/tasks/task_064_comprehensive_decision_medium_medium039.md
@@ -0,0 +1,119 @@
+---
+id: task_064_comprehensive_decision_medium_medium039
+name: comprehensive_decision-medium-medium039
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium039.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, list all indicators for which Guangdong Province's Information Transmission, Software and Information Technology Services industry has mean values superior to the national average, and sort them by advantage magnitude from high to low.
+
+Output guidelines:
+Answer format: [Indicator 1, Indicator 2, Indicator 3, ...]. Output only indicator names, separated by commas and spaces, do not add any other explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["Mean of year-over-year change in R&D investment ratio", "Mean of year-over-year change in net profit", "Mean of annual PCT patent applications", "Mean of annual PCT invention patent applications", "Mean of government incentive funds and subsidies", "Mean of net profit amount", "Mean of year-over-year change in capitalized R&D investment", "Mean of cumulative PCT patent applications", "Mean of cumulative PCT invention patent applications", "Mean of cumulative citations of all patents", "Mean of cumulative China invention patent grants", "Mean of participation in drafting national standards", "Mean of R&D personnel ratio", "Mean of R&D investment ratio", "Mean of asset-liability ratio"]`
+
+Scoring rules:
+- The gold answer is a list with N=15 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_14` each as 0 or 1.
+- Return `total = (sum(part_i)) / 15` exactly.
+- If the model output is missing or cannot be parsed into 15 comparable parts, score all parts 0.
+
diff --git a/tasks/task_065_comprehensive_decision_medium_medium040.md b/tasks/task_065_comprehensive_decision_medium_medium040.md
new file mode 100644
index 0000000000000000000000000000000000000000..203dbac4f2b936d8dd5a4b22e74860518bd377c0
--- /dev/null
+++ b/tasks/task_065_comprehensive_decision_medium_medium040.md
@@ -0,0 +1,117 @@
+---
+id: task_065_comprehensive_decision_medium_medium040
+name: comprehensive_decision-medium-medium040
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium040.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in Guangdong Province's Raw Chemical Materials and Chemical Products Manufacturing industry, what share of the market's operating revenue does the company with the highest market capitalization account for?
+
+Output guidelines:
+The answer must be an exact number, rounded to 2 decimal places. Output only the number, do not add units, commas, or any textual explanation. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`21.92`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_066_comprehensive_decision_medium_medium041.md b/tasks/task_066_comprehensive_decision_medium_medium041.md
new file mode 100644
index 0000000000000000000000000000000000000000..58baef54e98437157b66752e56285a99bb409c39
--- /dev/null
+++ b/tasks/task_066_comprehensive_decision_medium_medium041.md
@@ -0,0 +1,117 @@
+---
+id: task_066_comprehensive_decision_medium_medium041
+name: comprehensive_decision-medium-medium041
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium041.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, nationwide, is the province with the highest mean market capitalization also the province with the highest total net profit?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only the answer, do not add any other explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_067_comprehensive_decision_medium_medium042.md b/tasks/task_067_comprehensive_decision_medium_medium042.md
new file mode 100644
index 0000000000000000000000000000000000000000..adc0957fd92fb4b326abe870cc881818c8ceeb79
--- /dev/null
+++ b/tasks/task_067_comprehensive_decision_medium_medium042.md
@@ -0,0 +1,117 @@
+---
+id: task_067_comprehensive_decision_medium_medium042
+name: comprehensive_decision-medium-medium042
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium042.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, among enterprises in regions applicable to the "Notice on Organizing Applications for First Home Purchase Subsidies for Outstanding Young Talents in the Biomedicine Industry" policy, what is the net profit of the enterprise with the best operating revenue performance?
+
+Output guidelines:
+The answer must be an exact number, in CNY, rounded to 2 decimal places. Output only the number, do not add units, commas, or any textual explanation. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`4253373290.96`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_068_comprehensive_decision_medium_medium043.md b/tasks/task_068_comprehensive_decision_medium_medium043.md
new file mode 100644
index 0000000000000000000000000000000000000000..663753822abae88ac96f181297466db1636d344b
--- /dev/null
+++ b/tasks/task_068_comprehensive_decision_medium_medium043.md
@@ -0,0 +1,117 @@
+---
+id: task_068_comprehensive_decision_medium_medium043
+name: comprehensive_decision-medium-medium043
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium043.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, for Pharmaceutical Manufacturing, is the province with the highest total assets also the province with the highest R&D investment?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only "Yes" or "No", do not add any explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_069_comprehensive_decision_medium_medium044.md b/tasks/task_069_comprehensive_decision_medium_medium044.md
new file mode 100644
index 0000000000000000000000000000000000000000..71b9fde4f66f5ea2efccc5ab41795ff0639e89e4
--- /dev/null
+++ b/tasks/task_069_comprehensive_decision_medium_medium044.md
@@ -0,0 +1,117 @@
+---
+id: task_069_comprehensive_decision_medium_medium044
+name: comprehensive_decision-medium-medium044
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium044.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in Shandong Province's Pharmaceutical Manufacturing industry, is Haishan Changgong Equipment Company's R&D investment ratio higher than the R&D investment ratio of the 10th-ranked Pharmaceutical Manufacturing enterprise in Hunan Province?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_070_comprehensive_decision_medium_medium045.md b/tasks/task_070_comprehensive_decision_medium_medium045.md
new file mode 100644
index 0000000000000000000000000000000000000000..27a2f5163b2b578d73d8f9df33f39a9a4feb8983
--- /dev/null
+++ b/tasks/task_070_comprehensive_decision_medium_medium045.md
@@ -0,0 +1,117 @@
+---
+id: task_070_comprehensive_decision_medium_medium045
+name: comprehensive_decision-medium-medium045
+category: comprehensive_decision
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/comprehensive_decision/medium045.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in Sichuan Province, is Zhongbai Jinmao Chain Company's R&D investment higher than the R&D investment of the 15th-ranked enterprise nationwide in its industry?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only the answer, do not add any explanatory text. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_071_enterprise_industry_analysis_easy_easy001.md b/tasks/task_071_enterprise_industry_analysis_easy_easy001.md
new file mode 100644
index 0000000000000000000000000000000000000000..2133429cdf91127f32b3605f66979d2dfe9b0b83
--- /dev/null
+++ b/tasks/task_071_enterprise_industry_analysis_easy_easy001.md
@@ -0,0 +1,117 @@
+---
+id: task_071_enterprise_industry_analysis_easy_easy001
+name: enterprise_industry_analysis-easy-easy001
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy001.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is higher: the total number of citations of all patents of Zhong Ji Da Chang Tong Ye Co., Ltd. or the industry median benchmark?
+
+Output guidelines:
+The answer must be either the company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Zhong Ji Da Chang Tong Ye Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_072_enterprise_industry_analysis_easy_easy002.md b/tasks/task_072_enterprise_industry_analysis_easy_easy002.md
new file mode 100644
index 0000000000000000000000000000000000000000..19b25ec32d121f0043c1361cdfbeb8c8bf91dd6e
--- /dev/null
+++ b/tasks/task_072_enterprise_industry_analysis_easy_easy002.md
@@ -0,0 +1,117 @@
+---
+id: task_072_enterprise_industry_analysis_easy_easy002
+name: enterprise_industry_analysis-easy-easy002
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy002.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, what is the difference between the total liabilities of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. and the median total liabilities of its industry?
+
+Output guidelines:
+The answer must be a number with three decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`493355813.605`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_073_enterprise_industry_analysis_easy_easy003.md b/tasks/task_073_enterprise_industry_analysis_easy_easy003.md
new file mode 100644
index 0000000000000000000000000000000000000000..d271d0041d17aeca01aaad7eb4d39667d27a9c5a
--- /dev/null
+++ b/tasks/task_073_enterprise_industry_analysis_easy_easy003.md
@@ -0,0 +1,117 @@
+---
+id: task_073_enterprise_industry_analysis_easy_easy003
+name: enterprise_industry_analysis-easy-easy003
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy003.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is lower: the year-over-year change rate of operating profit of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. or the average of this indicator in its industry?
+
+Output guidelines:
+The answer must be either the company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_074_enterprise_industry_analysis_easy_easy004.md b/tasks/task_074_enterprise_industry_analysis_easy_easy004.md
new file mode 100644
index 0000000000000000000000000000000000000000..c9bd92b7b76b682f23772fbf1abaf35033eed462
--- /dev/null
+++ b/tasks/task_074_enterprise_industry_analysis_easy_easy004.md
@@ -0,0 +1,117 @@
+---
+id: task_074_enterprise_industry_analysis_easy_easy004
+name: enterprise_industry_analysis-easy-easy004
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy004.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the year-over-year net profit growth rate of Zhong Tong Jie Tong Yun Shu Co., Ltd. lower than that of Yun Da Hang Chang Kuai Di Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_075_enterprise_industry_analysis_easy_easy005.md b/tasks/task_075_enterprise_industry_analysis_easy_easy005.md
new file mode 100644
index 0000000000000000000000000000000000000000..a82a253af4660e1f3c0a4d84898dd78067887622
--- /dev/null
+++ b/tasks/task_075_enterprise_industry_analysis_easy_easy005.md
@@ -0,0 +1,117 @@
+---
+id: task_075_enterprise_industry_analysis_easy_easy005
+name: enterprise_industry_analysis-easy-easy005
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy005.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the market capitalization of Zhong Tong Jie Tong Yun Shu Co., Ltd. lower than that of Yun Da Hang Chang Kuai Di Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_076_enterprise_industry_analysis_easy_easy006.md b/tasks/task_076_enterprise_industry_analysis_easy_easy006.md
new file mode 100644
index 0000000000000000000000000000000000000000..bd944335d138c8de7d1e1af760c4177d8509d9e4
--- /dev/null
+++ b/tasks/task_076_enterprise_industry_analysis_easy_easy006.md
@@ -0,0 +1,117 @@
+---
+id: task_076_enterprise_industry_analysis_easy_easy006
+name: enterprise_industry_analysis-easy-easy006
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy006.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the annual number of authorized Chinese invention patents of Huan Xing Jin Ya Shi Shang Co., Ltd. lower than that of Li Ding Sheng Shang Fang Zhi Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_077_enterprise_industry_analysis_easy_easy007.md b/tasks/task_077_enterprise_industry_analysis_easy_easy007.md
new file mode 100644
index 0000000000000000000000000000000000000000..2bc5338a0996649e871ab5b04587f4ffd08b327c
--- /dev/null
+++ b/tasks/task_077_enterprise_industry_analysis_easy_easy007.md
@@ -0,0 +1,117 @@
+---
+id: task_077_enterprise_industry_analysis_easy_easy007
+name: enterprise_industry_analysis-easy-easy007
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy007.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the number of national standards participated in drafting by Bao Xin Ke Hui Ruan Jian Co., Ltd. the same as that of Zhong Ke Chuang Xin Ruan Jian Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_078_enterprise_industry_analysis_easy_easy008.md b/tasks/task_078_enterprise_industry_analysis_easy_easy008.md
new file mode 100644
index 0000000000000000000000000000000000000000..451ccd7e258655d2f5dae3e9d125ec27481042ff
--- /dev/null
+++ b/tasks/task_078_enterprise_industry_analysis_easy_easy008.md
@@ -0,0 +1,117 @@
+---
+id: task_078_enterprise_industry_analysis_easy_easy008
+name: enterprise_industry_analysis-easy-easy008
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy008.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, compared with Zhong Ke Chuang Xin Ruan Jian Co., Ltd., what is the difference in cumulative PCT patent applications of Bao Xin Ke Hui Ruan Jian Co., Ltd.?
+
+Output guidelines:
+The answer must be a number with one decimal place. Output only the number without units, commas, or any text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-483.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_079_enterprise_industry_analysis_easy_easy009.md b/tasks/task_079_enterprise_industry_analysis_easy_easy009.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b63f8b6913233b5ed07315a667421395cf29c52
--- /dev/null
+++ b/tasks/task_079_enterprise_industry_analysis_easy_easy009.md
@@ -0,0 +1,117 @@
+---
+id: task_079_enterprise_industry_analysis_easy_easy009
+name: enterprise_industry_analysis-easy-easy009
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy009.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the total number of employees of Can Xin Hui Xin Semiconductor Co., Ltd. higher than that of Rui Xin Xin Yao Materials Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_080_enterprise_industry_analysis_easy_easy010.md b/tasks/task_080_enterprise_industry_analysis_easy_easy010.md
new file mode 100644
index 0000000000000000000000000000000000000000..7781628af26676f84771d1f2ed8a0c041eacfd7d
--- /dev/null
+++ b/tasks/task_080_enterprise_industry_analysis_easy_easy010.md
@@ -0,0 +1,117 @@
+---
+id: task_080_enterprise_industry_analysis_easy_easy010
+name: enterprise_industry_analysis-easy-easy010
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy010.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the cumulative number of granted Chinese invention patents of Chuang Wei Yao Yao Dian Qi Co., Ltd. lower than that of Mei Neng Dian Jin Technology Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_081_enterprise_industry_analysis_easy_easy011.md b/tasks/task_081_enterprise_industry_analysis_easy_easy011.md
new file mode 100644
index 0000000000000000000000000000000000000000..ae87b631cf4e13cdb906f60859bd9eca94076b03
--- /dev/null
+++ b/tasks/task_081_enterprise_industry_analysis_easy_easy011.md
@@ -0,0 +1,117 @@
+---
+id: task_081_enterprise_industry_analysis_easy_easy011
+name: enterprise_industry_analysis-easy-easy011
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy011.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the R&D investment ratio of Yong Feng Xin Ruan Network Co., Ltd. higher than that of Jin Fei Shu Ruan Data Services Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_082_enterprise_industry_analysis_easy_easy012.md b/tasks/task_082_enterprise_industry_analysis_easy_easy012.md
new file mode 100644
index 0000000000000000000000000000000000000000..9b8e9bcaa3958325d4ef41a5adc23ad91bc8a8f2
--- /dev/null
+++ b/tasks/task_082_enterprise_industry_analysis_easy_easy012.md
@@ -0,0 +1,117 @@
+---
+id: task_082_enterprise_industry_analysis_easy_easy012
+name: enterprise_industry_analysis-easy-easy012
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy012.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the total assets of Yong Feng Xin Ruan Network Co., Ltd. lower than that of Jin Fei Shu Ruan Data Services Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_083_enterprise_industry_analysis_easy_easy013.md b/tasks/task_083_enterprise_industry_analysis_easy_easy013.md
new file mode 100644
index 0000000000000000000000000000000000000000..4ccbc83cfadbb0331c339dc16d2985ccf1a97182
--- /dev/null
+++ b/tasks/task_083_enterprise_industry_analysis_easy_easy013.md
@@ -0,0 +1,117 @@
+---
+id: task_083_enterprise_industry_analysis_easy_easy013
+name: enterprise_industry_analysis-easy-easy013
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy013.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the company market value of Wu Li Chang Yuan Wholesale Co., Ltd. higher than that of Wu Li Hui Jin Retail Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_084_enterprise_industry_analysis_easy_easy014.md b/tasks/task_084_enterprise_industry_analysis_easy_easy014.md
new file mode 100644
index 0000000000000000000000000000000000000000..d776e1c04c5b3b7234dcfdaf92632949c06db383
--- /dev/null
+++ b/tasks/task_084_enterprise_industry_analysis_easy_easy014.md
@@ -0,0 +1,117 @@
+---
+id: task_084_enterprise_industry_analysis_easy_easy014
+name: enterprise_industry_analysis-easy-easy014
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy014.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the operating revenue of Wu Li Chang Yuan Pi Fa Co., Ltd. lower than that of Wu Li Hui Jin Ling Shou Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_085_enterprise_industry_analysis_easy_easy015.md b/tasks/task_085_enterprise_industry_analysis_easy_easy015.md
new file mode 100644
index 0000000000000000000000000000000000000000..b758ae4b67f92bd999f08812ae4e6535a32d1940
--- /dev/null
+++ b/tasks/task_085_enterprise_industry_analysis_easy_easy015.md
@@ -0,0 +1,117 @@
+---
+id: task_085_enterprise_industry_analysis_easy_easy015
+name: enterprise_industry_analysis-easy-easy015
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy015.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the annual number of Chinese patent applications of Mei Neng Xuan Jin Dian Qi Co., Ltd. higher than that of Li Xin Yao Yue Dian Qi Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_086_enterprise_industry_analysis_easy_easy016.md b/tasks/task_086_enterprise_industry_analysis_easy_easy016.md
new file mode 100644
index 0000000000000000000000000000000000000000..d116c0d7ea9f97e1ad5784c83b130c4a250dfa9a
--- /dev/null
+++ b/tasks/task_086_enterprise_industry_analysis_easy_easy016.md
@@ -0,0 +1,117 @@
+---
+id: task_086_enterprise_industry_analysis_easy_easy016
+name: enterprise_industry_analysis-easy-easy016
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy016.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the annual number of authorized Chinese invention patents of Mei Neng Xuan Jin Dian Qi Co., Ltd. higher than the annual number of Chinese invention patent applications of Hai Li Chuang Yao Jia Dian Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_087_enterprise_industry_analysis_easy_easy017.md b/tasks/task_087_enterprise_industry_analysis_easy_easy017.md
new file mode 100644
index 0000000000000000000000000000000000000000..ba1d53deddc7768683d35979820ab024d055b411
--- /dev/null
+++ b/tasks/task_087_enterprise_industry_analysis_easy_easy017.md
@@ -0,0 +1,117 @@
+---
+id: task_087_enterprise_industry_analysis_easy_easy017
+name: enterprise_industry_analysis-easy-easy017
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy017.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the number of provincial or ministerial science and technology progress awards of Mei Neng Xuan Jin Dian Qi Co., Ltd. lower than that of Hai Li Chuang Yao Jia Dian Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_088_enterprise_industry_analysis_easy_easy018.md b/tasks/task_088_enterprise_industry_analysis_easy_easy018.md
new file mode 100644
index 0000000000000000000000000000000000000000..6d6bab9d571ac2c9a8cf8bad67087854b22ade81
--- /dev/null
+++ b/tasks/task_088_enterprise_industry_analysis_easy_easy018.md
@@ -0,0 +1,117 @@
+---
+id: task_088_enterprise_industry_analysis_easy_easy018
+name: enterprise_industry_analysis-easy-easy018
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy018.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, compared with Jing Neng Dian Re Ran Qi Co., Ltd., what is the difference in total liabilities of San Xia Ze Neng Dian Li Co., Ltd.?
+
+Output guidelines:
+The answer must be a number with two decimal places. Output only the number without units, commas, or any text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`2070319837.92`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_089_enterprise_industry_analysis_easy_easy019.md b/tasks/task_089_enterprise_industry_analysis_easy_easy019.md
new file mode 100644
index 0000000000000000000000000000000000000000..67c8540dbfb66bf3cfa41b14e9457025b5006adf
--- /dev/null
+++ b/tasks/task_089_enterprise_industry_analysis_easy_easy019.md
@@ -0,0 +1,117 @@
+---
+id: task_089_enterprise_industry_analysis_easy_easy019
+name: enterprise_industry_analysis-easy-easy019
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy019.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the R&D investment ratio of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. lower than that of Yi Shan Tai Tai Medical Devices Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_090_enterprise_industry_analysis_easy_easy020.md b/tasks/task_090_enterprise_industry_analysis_easy_easy020.md
new file mode 100644
index 0000000000000000000000000000000000000000..782716d9e478910cee7153c4e6ec3dc6b09e6322
--- /dev/null
+++ b/tasks/task_090_enterprise_industry_analysis_easy_easy020.md
@@ -0,0 +1,117 @@
+---
+id: task_090_enterprise_industry_analysis_easy_easy020
+name: enterprise_industry_analysis-easy-easy020
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy020.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the debt-to-asset ratio of Zhong Hai Gong Zhu Jin Jian Zhu She Ji Co., Ltd. higher than that of Yi Shan Tai Tai Medical Devices Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_091_enterprise_industry_analysis_easy_easy021.md b/tasks/task_091_enterprise_industry_analysis_easy_easy021.md
new file mode 100644
index 0000000000000000000000000000000000000000..e012a3626c6df4f2b293b9e56a9d08e136018bcd
--- /dev/null
+++ b/tasks/task_091_enterprise_industry_analysis_easy_easy021.md
@@ -0,0 +1,117 @@
+---
+id: task_091_enterprise_industry_analysis_easy_easy021
+name: enterprise_industry_analysis-easy-easy021
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy021.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the total citations of all patents of Guang Sheng Chang Ze Group Co., Ltd. higher than the corresponding indicator of Lang Ji Yun Hui Technology Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_092_enterprise_industry_analysis_easy_easy022.md b/tasks/task_092_enterprise_industry_analysis_easy_easy022.md
new file mode 100644
index 0000000000000000000000000000000000000000..8dccfbcd49cc075323d806fdcf5c75d7d2994b42
--- /dev/null
+++ b/tasks/task_092_enterprise_industry_analysis_easy_easy022.md
@@ -0,0 +1,117 @@
+---
+id: task_092_enterprise_industry_analysis_easy_easy022
+name: enterprise_industry_analysis-easy-easy022
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy022.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the annual number of Chinese patent applications of Pu Ge Jian Chen Pharmaceutical Co., Ltd. higher than that of Jin Hu Real Estate Construction Development Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_093_enterprise_industry_analysis_easy_easy023.md b/tasks/task_093_enterprise_industry_analysis_easy_easy023.md
new file mode 100644
index 0000000000000000000000000000000000000000..ba45f6a4bc445d2c90076b64a7560dad1853e395
--- /dev/null
+++ b/tasks/task_093_enterprise_industry_analysis_easy_easy023.md
@@ -0,0 +1,117 @@
+---
+id: task_093_enterprise_industry_analysis_easy_easy023
+name: enterprise_industry_analysis-easy-easy023
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy023.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is larger: the R&D investment ratio of Hua Cheng Sheng Yuan Integrated Development Co., Ltd. or the year-over-year net profit change rate of Long He Zhi Jin Real Estate Co., Ltd.?
+
+Output guidelines:
+The answer must be either the company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Hua Cheng Sheng Yuan Integrated Development Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_094_enterprise_industry_analysis_easy_easy024.md b/tasks/task_094_enterprise_industry_analysis_easy_easy024.md
new file mode 100644
index 0000000000000000000000000000000000000000..24f3db55ff535c218135f7905c1d8eb6b9a6d1f3
--- /dev/null
+++ b/tasks/task_094_enterprise_industry_analysis_easy_easy024.md
@@ -0,0 +1,117 @@
+---
+id: task_094_enterprise_industry_analysis_easy_easy024
+name: enterprise_industry_analysis-easy-easy024
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy024.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the cumulative number of PCT patent applications of Shi Yang Zhi Guang Dian Qi Co., Ltd. lower than the corresponding value of Xu Ye Zhi Gong Technology Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_095_enterprise_industry_analysis_easy_easy025.md b/tasks/task_095_enterprise_industry_analysis_easy_easy025.md
new file mode 100644
index 0000000000000000000000000000000000000000..1f8512a481b35a17ddb3e56c24999a25dc74f8ce
--- /dev/null
+++ b/tasks/task_095_enterprise_industry_analysis_easy_easy025.md
@@ -0,0 +1,117 @@
+---
+id: task_095_enterprise_industry_analysis_easy_easy025
+name: enterprise_industry_analysis-easy-easy025
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy025.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the government award funding or subsidy of Shi Yang Zhi Guang Dian Qi Co., Ltd. lower than that of Xu Ye Zhi Gong Technology Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_096_enterprise_industry_analysis_easy_easy026.md b/tasks/task_096_enterprise_industry_analysis_easy_easy026.md
new file mode 100644
index 0000000000000000000000000000000000000000..497f1f0cbc81a1fac901984a98f6dfd3e2567916
--- /dev/null
+++ b/tasks/task_096_enterprise_industry_analysis_easy_easy026.md
@@ -0,0 +1,117 @@
+---
+id: task_096_enterprise_industry_analysis_easy_easy026
+name: enterprise_industry_analysis-easy-easy026
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy026.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the total number of employees of Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. higher than that of Shen Zhou Wu Jin Zi Xun Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_097_enterprise_industry_analysis_easy_easy027.md b/tasks/task_097_enterprise_industry_analysis_easy_easy027.md
new file mode 100644
index 0000000000000000000000000000000000000000..83b8ae0ed66f1e2eb29d14a6c50bc73763459cf9
--- /dev/null
+++ b/tasks/task_097_enterprise_industry_analysis_easy_easy027.md
@@ -0,0 +1,117 @@
+---
+id: task_097_enterprise_industry_analysis_easy_easy027
+name: enterprise_industry_analysis-easy-easy027
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy027.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the cumulative number of invalid Chinese invention patents of Yuan Tong Sheng Sheng Gong Ying Lian Co., Ltd. lower than the same indicator of Shen Zhou Wu Jin Zi Xun Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_098_enterprise_industry_analysis_easy_easy028.md b/tasks/task_098_enterprise_industry_analysis_easy_easy028.md
new file mode 100644
index 0000000000000000000000000000000000000000..812f191fe52d7a9d9f3c68ea955c7c6a89b11862
--- /dev/null
+++ b/tasks/task_098_enterprise_industry_analysis_easy_easy028.md
@@ -0,0 +1,117 @@
+---
+id: task_098_enterprise_industry_analysis_easy_easy028
+name: enterprise_industry_analysis-easy-easy028
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy028.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the government reward fund or subsidy of Guang Sheng Chang Ze Group Co., Ltd. higher than that of Zhong You Zheng Yun Yun Port Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_099_enterprise_industry_analysis_easy_easy029.md b/tasks/task_099_enterprise_industry_analysis_easy_easy029.md
new file mode 100644
index 0000000000000000000000000000000000000000..309cbd1557976752da07d1c4883adfe3a2167f76
--- /dev/null
+++ b/tasks/task_099_enterprise_industry_analysis_easy_easy029.md
@@ -0,0 +1,117 @@
+---
+id: task_099_enterprise_industry_analysis_easy_easy029
+name: enterprise_industry_analysis-easy-easy029
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy029.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the year-over-year change rate of R&D investment of Yong Feng Ke Lian Software Co., Ltd. higher than that of Hai Li Xuan Yue Electric Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_100_enterprise_industry_analysis_easy_easy030.md b/tasks/task_100_enterprise_industry_analysis_easy_easy030.md
new file mode 100644
index 0000000000000000000000000000000000000000..c02f06f44415f5d07946221633af35e851c10e29
--- /dev/null
+++ b/tasks/task_100_enterprise_industry_analysis_easy_easy030.md
@@ -0,0 +1,117 @@
+---
+id: task_100_enterprise_industry_analysis_easy_easy030
+name: enterprise_industry_analysis-easy-easy030
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy030.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the total assets of Yong Feng Ke Lian Software Co., Ltd. higher than that of Hai Li Xuan Yue Electric Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_101_enterprise_industry_analysis_easy_easy031.md b/tasks/task_101_enterprise_industry_analysis_easy_easy031.md
new file mode 100644
index 0000000000000000000000000000000000000000..ffae6d53fd586f3360d9663838777dd44dbd39fa
--- /dev/null
+++ b/tasks/task_101_enterprise_industry_analysis_easy_easy031.md
@@ -0,0 +1,117 @@
+---
+id: task_101_enterprise_industry_analysis_easy_easy031
+name: enterprise_industry_analysis-easy-easy031
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy031.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is higher: the operating revenue amount of Heng Yi Run Heng Technology Co., Ltd. or the total assets of Lian Ji Ji Jin Ji Chuang Co., Ltd.?
+
+Output guidelines:
+The answer must be either the company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Lian Ji Ji Jin Ji Chuang Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_102_enterprise_industry_analysis_easy_easy032.md b/tasks/task_102_enterprise_industry_analysis_easy_easy032.md
new file mode 100644
index 0000000000000000000000000000000000000000..47013ff9b73026d53eaca003a3ec94955053a039
--- /dev/null
+++ b/tasks/task_102_enterprise_industry_analysis_easy_easy032.md
@@ -0,0 +1,117 @@
+---
+id: task_102_enterprise_industry_analysis_easy_easy032
+name: enterprise_industry_analysis-easy-easy032
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy032.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the number of R&D personnel of Heng Yi Run Heng Technology Co., Ltd. lower than the annual number of Chinese patent applications of Lian Ji Ji Jin Ji Chuang Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_103_enterprise_industry_analysis_easy_easy033.md b/tasks/task_103_enterprise_industry_analysis_easy_easy033.md
new file mode 100644
index 0000000000000000000000000000000000000000..9a0524f710dcd24e6760ef87ef58f3d661255e0c
--- /dev/null
+++ b/tasks/task_103_enterprise_industry_analysis_easy_easy033.md
@@ -0,0 +1,117 @@
+---
+id: task_103_enterprise_industry_analysis_easy_easy033
+name: enterprise_industry_analysis-easy-easy033
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy033.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, compared with the total liabilities of Wei Xing Run Jin Ke Ji Co., Ltd., what is the difference in total assets of Ping Ru Gang Tong Yun Wu Liu Co., Ltd.?
+
+Output guidelines:
+The answer must be a number with two decimal places. Output only the number without units, commas, or any text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`37950292402.87`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_104_enterprise_industry_analysis_easy_easy034.md b/tasks/task_104_enterprise_industry_analysis_easy_easy034.md
new file mode 100644
index 0000000000000000000000000000000000000000..37ec6314926b2d5080a19f5d9bbb7eeb03a02a41
--- /dev/null
+++ b/tasks/task_104_enterprise_industry_analysis_easy_easy034.md
@@ -0,0 +1,117 @@
+---
+id: task_104_enterprise_industry_analysis_easy_easy034
+name: enterprise_industry_analysis-easy-easy034
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy034.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, between the year-over-year net profit growth rate of Ping Ru Gang Tong Yun Wu Liu Co., Ltd. and the net profit amount of Wei Xing Run Jin Ke Ji Co., Ltd., which value is larger?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Wei Xing Run Jin Ke Ji Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_105_enterprise_industry_analysis_easy_easy035.md b/tasks/task_105_enterprise_industry_analysis_easy_easy035.md
new file mode 100644
index 0000000000000000000000000000000000000000..bca54f18338aa516f0ea234a83d7bc344fd2dd86
--- /dev/null
+++ b/tasks/task_105_enterprise_industry_analysis_easy_easy035.md
@@ -0,0 +1,117 @@
+---
+id: task_105_enterprise_industry_analysis_easy_easy035
+name: enterprise_industry_analysis-easy-easy035
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy035.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, are the industries of Gao Yin Ze Tong Pi Fa Co., Ltd. and Yong Hui Ze Hui Pi Fa Co., Ltd. the same?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_106_enterprise_industry_analysis_easy_easy036.md b/tasks/task_106_enterprise_industry_analysis_easy_easy036.md
new file mode 100644
index 0000000000000000000000000000000000000000..ef430e4a87b7f5a89f4ad9cdb81d2af782855fc2
--- /dev/null
+++ b/tasks/task_106_enterprise_industry_analysis_easy_easy036.md
@@ -0,0 +1,117 @@
+---
+id: task_106_enterprise_industry_analysis_easy_easy036
+name: enterprise_industry_analysis-easy-easy036
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy036.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Compared with the listing date of Yong Hui Ze Hui Pi Fa Co., Ltd., which listing date is earlier for Gao Yin Ze Tong Pi Fa Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Gao Yin Ze Tong Pi Fa Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_107_enterprise_industry_analysis_easy_easy037.md b/tasks/task_107_enterprise_industry_analysis_easy_easy037.md
new file mode 100644
index 0000000000000000000000000000000000000000..a6fdf0a9496bfb66e7bdb8104bc9d442b81ec139
--- /dev/null
+++ b/tasks/task_107_enterprise_industry_analysis_easy_easy037.md
@@ -0,0 +1,117 @@
+---
+id: task_107_enterprise_industry_analysis_easy_easy037
+name: enterprise_industry_analysis-easy-easy037
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy037.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, are the industries of Ma Gang Tai Jin Cai Liao Co., Ltd. and Ma Gang Gang Sheng Bu Xiu Gang Co., Ltd. the same?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_108_enterprise_industry_analysis_easy_easy038.md b/tasks/task_108_enterprise_industry_analysis_easy_easy038.md
new file mode 100644
index 0000000000000000000000000000000000000000..e172a0205c7f665cc5d50899662d55eaa33e63a6
--- /dev/null
+++ b/tasks/task_108_enterprise_industry_analysis_easy_easy038.md
@@ -0,0 +1,117 @@
+---
+id: task_108_enterprise_industry_analysis_easy_easy038
+name: enterprise_industry_analysis-easy-easy038
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy038.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the enterprise type of Magang Taijin Materials Co., Ltd. the same as that of Magang Gangsheng Stainless Steel Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_109_enterprise_industry_analysis_easy_easy039.md b/tasks/task_109_enterprise_industry_analysis_easy_easy039.md
new file mode 100644
index 0000000000000000000000000000000000000000..92f90097daf1d3c6438105ef2399fb037dff6b8b
--- /dev/null
+++ b/tasks/task_109_enterprise_industry_analysis_easy_easy039.md
@@ -0,0 +1,117 @@
+---
+id: task_109_enterprise_industry_analysis_easy_easy039
+name: enterprise_industry_analysis-easy-easy039
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy039.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the industry of Chuangwei Yaosheng Electric Co., Ltd. the same as that of Lixin Zhichuang Home Appliances Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_110_enterprise_industry_analysis_easy_easy040.md b/tasks/task_110_enterprise_industry_analysis_easy_easy040.md
new file mode 100644
index 0000000000000000000000000000000000000000..5ebc89fdf721ecc83b015a4153664319d6e344a9
--- /dev/null
+++ b/tasks/task_110_enterprise_industry_analysis_easy_easy040.md
@@ -0,0 +1,117 @@
+---
+id: task_110_enterprise_industry_analysis_easy_easy040
+name: enterprise_industry_analysis-easy-easy040
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy040.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the stock exchange of Chuangwei Yaosheng Electric Co., Ltd. the same as that of Lixin Zhichuang Home Appliances Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_111_enterprise_industry_analysis_easy_easy041.md b/tasks/task_111_enterprise_industry_analysis_easy_easy041.md
new file mode 100644
index 0000000000000000000000000000000000000000..730b3f2a228a0946510fa2d626e262cfa4050607
--- /dev/null
+++ b/tasks/task_111_enterprise_industry_analysis_easy_easy041.md
@@ -0,0 +1,117 @@
+---
+id: task_111_enterprise_industry_analysis_easy_easy041
+name: enterprise_industry_analysis-easy-easy041
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy041.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Comparing the incorporation dates of Biyuan Chanhua Real Estate Holdings Co., Ltd. and Huarun Zhijin Construction Development Co., Ltd., which one was established earlier?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Biyuan Chanhua Real Estate Holdings Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_112_enterprise_industry_analysis_easy_easy042.md b/tasks/task_112_enterprise_industry_analysis_easy_easy042.md
new file mode 100644
index 0000000000000000000000000000000000000000..9bb13e78e993262ae73023fe9a3de773bdb6270b
--- /dev/null
+++ b/tasks/task_112_enterprise_industry_analysis_easy_easy042.md
@@ -0,0 +1,117 @@
+---
+id: task_112_enterprise_industry_analysis_easy_easy042
+name: enterprise_industry_analysis-easy-easy042
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy042.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, are the industries of Biyuan Chanhua Real Estate Holdings Co., Ltd. and Huarun Zhijin Construction Development Co., Ltd. the same?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_113_enterprise_industry_analysis_easy_easy043.md b/tasks/task_113_enterprise_industry_analysis_easy_easy043.md
new file mode 100644
index 0000000000000000000000000000000000000000..17cae63a0d54528bdef90d528369628f1c2474b1
--- /dev/null
+++ b/tasks/task_113_enterprise_industry_analysis_easy_easy043.md
@@ -0,0 +1,117 @@
+---
+id: task_113_enterprise_industry_analysis_easy_easy043
+name: enterprise_industry_analysis-easy-easy043
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy043.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Comparing the incorporation dates of Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd. and Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd., which one was established earlier?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_114_enterprise_industry_analysis_easy_easy044.md b/tasks/task_114_enterprise_industry_analysis_easy_easy044.md
new file mode 100644
index 0000000000000000000000000000000000000000..f7c7f25e548eabc0b99893dd12e2bdf50a926706
--- /dev/null
+++ b/tasks/task_114_enterprise_industry_analysis_easy_easy044.md
@@ -0,0 +1,117 @@
+---
+id: task_114_enterprise_industry_analysis_easy_easy044
+name: enterprise_industry_analysis-easy-easy044
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy044.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, are the stock exchanges of Jian Ming Sheng Kang Yi Liao Qi Xie Co., Ltd. and Kang Sheng Kang Rui Yao Ye Joint Stock Co., Ltd. the same?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_115_enterprise_industry_analysis_easy_easy045.md b/tasks/task_115_enterprise_industry_analysis_easy_easy045.md
new file mode 100644
index 0000000000000000000000000000000000000000..7634d246d4734385c89d9237c025d6cb161ceaed
--- /dev/null
+++ b/tasks/task_115_enterprise_industry_analysis_easy_easy045.md
@@ -0,0 +1,117 @@
+---
+id: task_115_enterprise_industry_analysis_easy_easy045
+name: enterprise_industry_analysis-easy-easy045
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy045.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the industry of Chuang Wei Jin Zhi Electrical Appliances Co., Ltd. the same as that of Long He Chan Zhi Di Chan Holdings Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_116_enterprise_industry_analysis_easy_easy046.md b/tasks/task_116_enterprise_industry_analysis_easy_easy046.md
new file mode 100644
index 0000000000000000000000000000000000000000..37f017979af9b9ddc514cd05b0201c1de8f3e3e6
--- /dev/null
+++ b/tasks/task_116_enterprise_industry_analysis_easy_easy046.md
@@ -0,0 +1,117 @@
+---
+id: task_116_enterprise_industry_analysis_easy_easy046
+name: enterprise_industry_analysis-easy-easy046
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy046.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, are the listing boards of Chuang Wei Jin Zhi Electrical Appliances Co., Ltd. and Long He Chan Zhi Di Chan Holdings Co., Ltd. the same?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_117_enterprise_industry_analysis_easy_easy047.md b/tasks/task_117_enterprise_industry_analysis_easy_easy047.md
new file mode 100644
index 0000000000000000000000000000000000000000..226007fd0c8ae4ea3d264c3d9daf13eba7662f40
--- /dev/null
+++ b/tasks/task_117_enterprise_industry_analysis_easy_easy047.md
@@ -0,0 +1,117 @@
+---
+id: task_117_enterprise_industry_analysis_easy_easy047
+name: enterprise_industry_analysis-easy-easy047
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy047.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, are the industries of Hua Ying Tai Sheng Wealth Management Co., Ltd. and Yong Feng Lian Chuang Xi Tong Co., Ltd. the same?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_118_enterprise_industry_analysis_easy_easy048.md b/tasks/task_118_enterprise_industry_analysis_easy_easy048.md
new file mode 100644
index 0000000000000000000000000000000000000000..decda4128b2482380ec6291a5952be8ca912176f
--- /dev/null
+++ b/tasks/task_118_enterprise_industry_analysis_easy_easy048.md
@@ -0,0 +1,117 @@
+---
+id: task_118_enterprise_industry_analysis_easy_easy048
+name: enterprise_industry_analysis-easy-easy048
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy048.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the stock exchange of Huaying Taisheng Wealth Management Co., Ltd. the same as that of Yongfeng Lianchuang Systems Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_119_enterprise_industry_analysis_easy_easy049.md b/tasks/task_119_enterprise_industry_analysis_easy_easy049.md
new file mode 100644
index 0000000000000000000000000000000000000000..ac1ec440158f3104b57521dfd61e07aa488d1380
--- /dev/null
+++ b/tasks/task_119_enterprise_industry_analysis_easy_easy049.md
@@ -0,0 +1,117 @@
+---
+id: task_119_enterprise_industry_analysis_easy_easy049
+name: enterprise_industry_analysis-easy-easy049
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy049.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the board segment of Zhongbai Damao Wholesale Co., Ltd. the same as that of Luxi Runheng Chemical Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_120_enterprise_industry_analysis_easy_easy050.md b/tasks/task_120_enterprise_industry_analysis_easy_easy050.md
new file mode 100644
index 0000000000000000000000000000000000000000..7d28ab51e725836e93b55a099920129ae9106413
--- /dev/null
+++ b/tasks/task_120_enterprise_industry_analysis_easy_easy050.md
@@ -0,0 +1,117 @@
+---
+id: task_120_enterprise_industry_analysis_easy_easy050
+name: enterprise_industry_analysis-easy-easy050
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy050.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the ownership type of Zhongbai Damao Wholesale Co., Ltd. the same as that of Luxi Runheng Chemical Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_121_enterprise_industry_analysis_easy_easy051.md b/tasks/task_121_enterprise_industry_analysis_easy_easy051.md
new file mode 100644
index 0000000000000000000000000000000000000000..916d072dad0de5e2c9ade1877fb5790dece9c210
--- /dev/null
+++ b/tasks/task_121_enterprise_industry_analysis_easy_easy051.md
@@ -0,0 +1,117 @@
+---
+id: task_121_enterprise_industry_analysis_easy_easy051
+name: enterprise_industry_analysis-easy-easy051
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy051.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the enterprise type of Meineng Xuanyue Electric Co., Ltd. the same as that of Baotie Yuanchang Metal Products Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_122_enterprise_industry_analysis_easy_easy052.md b/tasks/task_122_enterprise_industry_analysis_easy_easy052.md
new file mode 100644
index 0000000000000000000000000000000000000000..8284f7c9dccdb3f769df6bd16e0ebe8017a5a78f
--- /dev/null
+++ b/tasks/task_122_enterprise_industry_analysis_easy_easy052.md
@@ -0,0 +1,117 @@
+---
+id: task_122_enterprise_industry_analysis_easy_easy052
+name: enterprise_industry_analysis-easy-easy052
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy052.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, are the stock exchanges of Mei Neng Xuan Yue Dian Qi Co., Ltd. and Bao Tie Yuan Chang Jin Shu Zhi Pin Co., Ltd. the same?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_123_enterprise_industry_analysis_easy_easy053.md b/tasks/task_123_enterprise_industry_analysis_easy_easy053.md
new file mode 100644
index 0000000000000000000000000000000000000000..3e45303eecd05b14503411d1fae943cb23569435
--- /dev/null
+++ b/tasks/task_123_enterprise_industry_analysis_easy_easy053.md
@@ -0,0 +1,117 @@
+---
+id: task_123_enterprise_industry_analysis_easy_easy053
+name: enterprise_industry_analysis-easy-easy053
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy053.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, does the secondary industry of Bao He Hua Chang Jian She Kai Fa Co., Ltd. belong to the same industry as that of Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_124_enterprise_industry_analysis_easy_easy054.md b/tasks/task_124_enterprise_industry_analysis_easy_easy054.md
new file mode 100644
index 0000000000000000000000000000000000000000..a59a812ff88ee8b4b81352ead2b62001bd250617
--- /dev/null
+++ b/tasks/task_124_enterprise_industry_analysis_easy_easy054.md
@@ -0,0 +1,117 @@
+---
+id: task_124_enterprise_industry_analysis_easy_easy054
+name: enterprise_industry_analysis-easy-easy054
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy054.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, are the industries of Bao He Hua Chang Jian She Kai Fa Co., Ltd. and Li Qun Tong Tong Dian Zi Shang Wu Co., Ltd. the same?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_125_enterprise_industry_analysis_easy_easy055.md b/tasks/task_125_enterprise_industry_analysis_easy_easy055.md
new file mode 100644
index 0000000000000000000000000000000000000000..34e13a6bb564fd77ae4d7519724fa69296ffb297
--- /dev/null
+++ b/tasks/task_125_enterprise_industry_analysis_easy_easy055.md
@@ -0,0 +1,117 @@
+---
+id: task_125_enterprise_industry_analysis_easy_easy055
+name: enterprise_industry_analysis-easy-easy055
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy055.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, compared with Chuang Xin Yao Rui Integrated Circuit Co., Ltd., did the core competitiveness of Ya Wei Ze Zhi Technology Co., Ltd. also emphasize technological innovation and high-quality customer resources?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_126_enterprise_industry_analysis_easy_easy056.md b/tasks/task_126_enterprise_industry_analysis_easy_easy056.md
new file mode 100644
index 0000000000000000000000000000000000000000..b1b735b5be049b38b4ef059ed4967a0380afb01d
--- /dev/null
+++ b/tasks/task_126_enterprise_industry_analysis_easy_easy056.md
@@ -0,0 +1,117 @@
+---
+id: task_126_enterprise_industry_analysis_easy_easy056
+name: enterprise_industry_analysis-easy-easy056
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy056.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, did the core competitiveness of Zhong Ke Sheng Ke Ji Shu Yan Jiu Yuan Co., Ltd. and that of Lian Ji Zhi Sheng Ji Xie Co., Ltd. in the same province show a competitive relationship in terms of technical capabilities?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_127_enterprise_industry_analysis_easy_easy057.md b/tasks/task_127_enterprise_industry_analysis_easy_easy057.md
new file mode 100644
index 0000000000000000000000000000000000000000..f4af2136c56d1e8c54947c86c52b759f3887917f
--- /dev/null
+++ b/tasks/task_127_enterprise_industry_analysis_easy_easy057.md
@@ -0,0 +1,117 @@
+---
+id: task_127_enterprise_industry_analysis_easy_easy057
+name: enterprise_industry_analysis-easy-easy057
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy057.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, did the products of He Lian Chuang Hang She Bei Co., Ltd. and Shan La Da Chuang Zhi Neng Zhuang Bei Co., Ltd. have a competitive relationship?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_128_enterprise_industry_analysis_easy_easy058.md b/tasks/task_128_enterprise_industry_analysis_easy_easy058.md
new file mode 100644
index 0000000000000000000000000000000000000000..291eabff04b463fb481d2aaa81cee03614a3c210
--- /dev/null
+++ b/tasks/task_128_enterprise_industry_analysis_easy_easy058.md
@@ -0,0 +1,117 @@
+---
+id: task_128_enterprise_industry_analysis_easy_easy058
+name: enterprise_industry_analysis-easy-easy058
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy058.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, compared with the products of Jiejie Dahang Equipment Co., Ltd., are the products of Sansan Gongzhi Technology Co., Ltd. in a competitive relationship?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_129_enterprise_industry_analysis_easy_easy059.md b/tasks/task_129_enterprise_industry_analysis_easy_easy059.md
new file mode 100644
index 0000000000000000000000000000000000000000..294f48128acf821ed3f976e5bf28bb6193dafe67
--- /dev/null
+++ b/tasks/task_129_enterprise_industry_analysis_easy_easy059.md
@@ -0,0 +1,117 @@
+---
+id: task_129_enterprise_industry_analysis_easy_easy059
+name: enterprise_industry_analysis-easy-easy059
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy059.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is Haomei Company's core competitiveness in R&D and technology more focused on innovation and diversification than that of Zhongjin Yeye Resources Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_130_enterprise_industry_analysis_easy_easy060.md b/tasks/task_130_enterprise_industry_analysis_easy_easy060.md
new file mode 100644
index 0000000000000000000000000000000000000000..eeb3eb6f4da4caa0aa4c3e637554879a32bbfbb5
--- /dev/null
+++ b/tasks/task_130_enterprise_industry_analysis_easy_easy060.md
@@ -0,0 +1,117 @@
+---
+id: task_130_enterprise_industry_analysis_easy_easy060
+name: enterprise_industry_analysis-easy-easy060
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy060.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, compared with Sansong Shijin Condiment Co., Ltd.'s strengths in technology development and independent innovation, whose strengths are more focused on market influence: Haishan Weixiang Catering Management Co., Ltd. or Sansong Shijin Condiment Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Haishan Weixiang Catering Management Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_131_enterprise_industry_analysis_easy_easy061.md b/tasks/task_131_enterprise_industry_analysis_easy_easy061.md
new file mode 100644
index 0000000000000000000000000000000000000000..815877206321eb2b516d0d5a32511939dbfae4a4
--- /dev/null
+++ b/tasks/task_131_enterprise_industry_analysis_easy_easy061.md
@@ -0,0 +1,117 @@
+---
+id: task_131_enterprise_industry_analysis_easy_easy061
+name: enterprise_industry_analysis-easy-easy061
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy061.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, comparing the R&D and technology of Meineng Dianguang Home Appliances Co., Ltd. with the technological innovation of Lixin Shengyue Intelligent Technology Co., Ltd., which company is more competitive in the connector field?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Meineng Dianguang Home Appliances Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_132_enterprise_industry_analysis_easy_easy062.md b/tasks/task_132_enterprise_industry_analysis_easy_easy062.md
new file mode 100644
index 0000000000000000000000000000000000000000..bf10172747529e29c5b8b32b8a14fc6aa6a47272
--- /dev/null
+++ b/tasks/task_132_enterprise_industry_analysis_easy_easy062.md
@@ -0,0 +1,117 @@
+---
+id: task_132_enterprise_industry_analysis_easy_easy062
+name: enterprise_industry_analysis-easy-easy062
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy062.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, what is the difference between the number of SSE-listed enterprises in the education industry in Beijing and the number of SSE-listed central state-owned enterprises in the national consumer electronics and electrical industry?
+
+Output guidelines:
+The answer must be a number with one decimal place. Output only the number without units, commas, or any explanatory text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-2.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_133_enterprise_industry_analysis_easy_easy063.md b/tasks/task_133_enterprise_industry_analysis_easy_easy063.md
new file mode 100644
index 0000000000000000000000000000000000000000..db06c283004801600a941ddd7207ca90f3895265
--- /dev/null
+++ b/tasks/task_133_enterprise_industry_analysis_easy_easy063.md
@@ -0,0 +1,117 @@
+---
+id: task_133_enterprise_industry_analysis_easy_easy063
+name: enterprise_industry_analysis-easy-easy063
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy063.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is lower: the average year-on-year net profit growth rate of Beijing's furniture manufacturing industry or that of the nationwide chemical fiber manufacturing industry?
+
+Output guidelines:
+The answer must be either "Beijing Furniture Manufacturing" or "Nationwide Chemical Fiber Manufacturing". If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Beijing Furniture Manufacturing"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_134_enterprise_industry_analysis_easy_easy064.md b/tasks/task_134_enterprise_industry_analysis_easy_easy064.md
new file mode 100644
index 0000000000000000000000000000000000000000..9cf35248ba32feef69bf8b75128b21997f55ecc5
--- /dev/null
+++ b/tasks/task_134_enterprise_industry_analysis_easy_easy064.md
@@ -0,0 +1,117 @@
+---
+id: task_134_enterprise_industry_analysis_easy_easy064
+name: enterprise_industry_analysis-easy-easy064
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy064.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is higher: the average year-on-year operating profit growth rate of Beijing's real estate industry or that of the nationwide information transmission, software, and IT services industry?
+
+Output guidelines:
+The answer must be either "Beijing Real Estate" or "Nationwide Information Transmission, Software and IT Services". If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Beijing Real Estate"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_135_enterprise_industry_analysis_easy_easy065.md b/tasks/task_135_enterprise_industry_analysis_easy_easy065.md
new file mode 100644
index 0000000000000000000000000000000000000000..302a9cb731b4fa7c3990d95bc66ba8f23a8f51fd
--- /dev/null
+++ b/tasks/task_135_enterprise_industry_analysis_easy_easy065.md
@@ -0,0 +1,117 @@
+---
+id: task_135_enterprise_industry_analysis_easy_easy065
+name: enterprise_industry_analysis-easy-easy065
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy065.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, what is the difference between the number of Shenzhen Stock Exchange-listed local state-owned enterprises in Beijing's information transmission, software, and IT services industry and the number of Shanghai Stock Exchange-listed foreign-funded enterprises in the nationwide real estate industry?
+
+Output guidelines:
+The answer must be a number with one decimal place. Output only the number without units, commas, or any text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-2.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_136_enterprise_industry_analysis_easy_easy066.md b/tasks/task_136_enterprise_industry_analysis_easy_easy066.md
new file mode 100644
index 0000000000000000000000000000000000000000..a13a373d287092f19e761bda8ea98442b9182a57
--- /dev/null
+++ b/tasks/task_136_enterprise_industry_analysis_easy_easy066.md
@@ -0,0 +1,117 @@
+---
+id: task_136_enterprise_industry_analysis_easy_easy066
+name: enterprise_industry_analysis-easy-easy066
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy066.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which value is higher: the minimum cumulative number of invalidated PCT invention patents in Beijing's comprehensive industry, or the same metric in the nationwide pharmaceutical manufacturing industry?
+
+Output guidelines:
+The answer must be "Equal", a company name, or "industry". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_137_enterprise_industry_analysis_easy_easy067.md b/tasks/task_137_enterprise_industry_analysis_easy_easy067.md
new file mode 100644
index 0000000000000000000000000000000000000000..d6a5c87f6752c501c820646ba1d0a334b66a7c32
--- /dev/null
+++ b/tasks/task_137_enterprise_industry_analysis_easy_easy067.md
@@ -0,0 +1,117 @@
+---
+id: task_137_enterprise_industry_analysis_easy_easy067
+name: enterprise_industry_analysis-easy-easy067
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy067.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is larger: the number of Shenzhen Stock Exchange-listed foreign-funded enterprises in Guangdong's scientific research and technical services industry, or the number of Shanghai Stock Exchange-listed state-owned institute enterprises in the same industry nationwide?
+
+Output guidelines:
+The answer must be "Equal", a company name, or "industry". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_138_enterprise_industry_analysis_easy_easy068.md b/tasks/task_138_enterprise_industry_analysis_easy_easy068.md
new file mode 100644
index 0000000000000000000000000000000000000000..130eb6e2933188bdd2977e572345893a82cf01c6
--- /dev/null
+++ b/tasks/task_138_enterprise_industry_analysis_easy_easy068.md
@@ -0,0 +1,117 @@
+---
+id: task_138_enterprise_industry_analysis_easy_easy068
+name: enterprise_industry_analysis-easy-easy068
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy068.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is higher: the average year-over-year change rate of R&D personnel in Guangdong Province's scientific research and technical services industry, or that of the same industry nationwide?
+
+Output guidelines:
+The answer must be "National" or "Guangdong". If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"National"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_139_enterprise_industry_analysis_easy_easy069.md b/tasks/task_139_enterprise_industry_analysis_easy_easy069.md
new file mode 100644
index 0000000000000000000000000000000000000000..99256ec44e76779594a360cf3ff1d6363e797fd7
--- /dev/null
+++ b/tasks/task_139_enterprise_industry_analysis_easy_easy069.md
@@ -0,0 +1,117 @@
+---
+id: task_139_enterprise_industry_analysis_easy_easy069
+name: enterprise_industry_analysis-easy-easy069
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy069.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, what percentage does the total number of employees in Jilin Province's comprehensive industry represent relative to the total number of employees in the same industry nationwide?
+
+Output guidelines:
+The answer must be "National" or "Jilin". If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_140_enterprise_industry_analysis_easy_easy070.md b/tasks/task_140_enterprise_industry_analysis_easy_easy070.md
new file mode 100644
index 0000000000000000000000000000000000000000..169cca0545feee21b399f9feb4201453d42928a0
--- /dev/null
+++ b/tasks/task_140_enterprise_industry_analysis_easy_easy070.md
@@ -0,0 +1,117 @@
+---
+id: task_140_enterprise_industry_analysis_easy_easy070
+name: enterprise_industry_analysis-easy-easy070
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy070.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, comparing the median number of participation in drafting industry standards for Jilin Province's comprehensive industry and the same indicator for the nationwide comprehensive industry, which one is lower?
+
+Output guidelines:
+The answer must be "Equal", a company name, or "industry". Output only one word or a company name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_141_enterprise_industry_analysis_easy_easy071.md b/tasks/task_141_enterprise_industry_analysis_easy_easy071.md
new file mode 100644
index 0000000000000000000000000000000000000000..e20dfbcc8ee29ca0b9f83355bcc86f8ee7be1ce8
--- /dev/null
+++ b/tasks/task_141_enterprise_industry_analysis_easy_easy071.md
@@ -0,0 +1,117 @@
+---
+id: task_141_enterprise_industry_analysis_easy_easy071
+name: enterprise_industry_analysis-easy-easy071
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy071.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is higher: the maximum asset-liability ratio in Jilin Province's commercial electrical machinery and equipment manufacturing industry, or the same indicator nationwide?
+
+Output guidelines:
+The answer must be "National" or "Jilin". If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"National"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_142_enterprise_industry_analysis_easy_easy072.md b/tasks/task_142_enterprise_industry_analysis_easy_easy072.md
new file mode 100644
index 0000000000000000000000000000000000000000..945fcfb45d7106ece0c9f1ac2b0d9f9599955895
--- /dev/null
+++ b/tasks/task_142_enterprise_industry_analysis_easy_easy072.md
@@ -0,0 +1,117 @@
+---
+id: task_142_enterprise_industry_analysis_easy_easy072
+name: enterprise_industry_analysis-easy-easy072
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy072.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is larger: the number of Shanghai Stock Exchange-listed Sino-foreign joint venture enterprises in Guangdong's semiconductor industry, or the number of Shenzhen Stock Exchange-listed state-owned institute enterprises in the nationwide semiconductor industry?
+
+Output guidelines:
+The answer must be "Equal" or the comparison conclusion term specified in the question. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_143_enterprise_industry_analysis_easy_easy073.md b/tasks/task_143_enterprise_industry_analysis_easy_easy073.md
new file mode 100644
index 0000000000000000000000000000000000000000..d7d93c31344d6c46997d5bc0f527573552c7f6aa
--- /dev/null
+++ b/tasks/task_143_enterprise_industry_analysis_easy_easy073.md
@@ -0,0 +1,117 @@
+---
+id: task_143_enterprise_industry_analysis_easy_easy073
+name: enterprise_industry_analysis-easy-easy073
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy073.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is larger: the total number of enterprises in Jilin's transportation, warehousing, and postal industry, or the number of Shenzhen Stock Exchange-listed central state-owned enterprises in the same industry nationwide?
+
+Output guidelines:
+The answer must be "Equal", a company name, or "industry". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Nationwide"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_144_enterprise_industry_analysis_easy_easy074.md b/tasks/task_144_enterprise_industry_analysis_easy_easy074.md
new file mode 100644
index 0000000000000000000000000000000000000000..5818c9f2b3d8e000c59e0faab8e7a52c5c209466
--- /dev/null
+++ b/tasks/task_144_enterprise_industry_analysis_easy_easy074.md
@@ -0,0 +1,117 @@
+---
+id: task_144_enterprise_industry_analysis_easy_easy074
+name: enterprise_industry_analysis-easy-easy074
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy074.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is higher: the maximum government award funding or subsidy value in Jilin's transportation, warehousing, and postal industry, or the same metric in the nationwide industry?
+
+Output guidelines:
+The answer must be "Nationwide" or "Jilin". If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Nationwide"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_145_enterprise_industry_analysis_easy_easy075.md b/tasks/task_145_enterprise_industry_analysis_easy_easy075.md
new file mode 100644
index 0000000000000000000000000000000000000000..ebb4af8a98b9293d648600881b9f212039ff3fed
--- /dev/null
+++ b/tasks/task_145_enterprise_industry_analysis_easy_easy075.md
@@ -0,0 +1,117 @@
+---
+id: task_145_enterprise_industry_analysis_easy_easy075
+name: enterprise_industry_analysis-easy-easy075
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy075.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is larger in the Tibet Autonomous Region: the total number of enterprises in the electricity, heat, gas, and water production and supply industry, or in the construction industry?
+
+Output guidelines:
+The answer must be "Equal", a company name, or "industry". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_146_enterprise_industry_analysis_easy_easy076.md b/tasks/task_146_enterprise_industry_analysis_easy_easy076.md
new file mode 100644
index 0000000000000000000000000000000000000000..d9db0d0853296081c37deb281b6d7f1f054fe400
--- /dev/null
+++ b/tasks/task_146_enterprise_industry_analysis_easy_easy076.md
@@ -0,0 +1,117 @@
+---
+id: task_146_enterprise_industry_analysis_easy_easy076
+name: enterprise_industry_analysis-easy-easy076
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy076.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is there any difference between the total number of enterprises in the Tibet Autonomous Region's electricity, heat, gas, and water production and supply industry and that in the construction industry?
+
+Output guidelines:
+The answer must be "Equal", a company name, or "industry". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_147_enterprise_industry_analysis_easy_easy077.md b/tasks/task_147_enterprise_industry_analysis_easy_easy077.md
new file mode 100644
index 0000000000000000000000000000000000000000..86ac62eaed52a2f2786c7ce6d93c0f4521b9f837
--- /dev/null
+++ b/tasks/task_147_enterprise_industry_analysis_easy_easy077.md
@@ -0,0 +1,117 @@
+---
+id: task_147_enterprise_industry_analysis_easy_easy077
+name: enterprise_industry_analysis-easy-easy077
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy077.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is larger in the Tibet Autonomous Region: the total number of enterprises in the electricity, heat, gas, and water production and supply industry, or in the leasing and business services industry?
+
+Output guidelines:
+The answer must be "Equal", a company name, or "industry". Output only one word or the company name, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_148_enterprise_industry_analysis_easy_easy078.md b/tasks/task_148_enterprise_industry_analysis_easy_easy078.md
new file mode 100644
index 0000000000000000000000000000000000000000..543d3170602cc19f474dac7f43d87e6ab2496dca
--- /dev/null
+++ b/tasks/task_148_enterprise_industry_analysis_easy_easy078.md
@@ -0,0 +1,117 @@
+---
+id: task_148_enterprise_industry_analysis_easy_easy078
+name: enterprise_industry_analysis-easy-easy078
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy078.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is there a difference between the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry and that in the commercial electrical machinery and equipment manufacturing industry?
+
+Output guidelines:
+The answer must be "Equal", a company name, or "industry". Output only one word or a company name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_149_enterprise_industry_analysis_easy_easy079.md b/tasks/task_149_enterprise_industry_analysis_easy_easy079.md
new file mode 100644
index 0000000000000000000000000000000000000000..0112da23c500b601197467d37f0817f86591e47c
--- /dev/null
+++ b/tasks/task_149_enterprise_industry_analysis_easy_easy079.md
@@ -0,0 +1,117 @@
+---
+id: task_149_enterprise_industry_analysis_easy_easy079
+name: enterprise_industry_analysis-easy-easy079
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy079.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry the same as that in the metal products industry?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_150_enterprise_industry_analysis_easy_easy080.md b/tasks/task_150_enterprise_industry_analysis_easy_easy080.md
new file mode 100644
index 0000000000000000000000000000000000000000..59a1961a934c5a5b59c570723fb33a0fc65f0f6c
--- /dev/null
+++ b/tasks/task_150_enterprise_industry_analysis_easy_easy080.md
@@ -0,0 +1,117 @@
+---
+id: task_150_enterprise_industry_analysis_easy_easy080
+name: enterprise_industry_analysis-easy-easy080
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy080.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, comparing the total number of enterprises in Tibet Autonomous Region's electricity, heat, gas and water production and supply industry with that in the general equipment manufacturing industry, which industry has more enterprises?
+
+Output guidelines:
+The answer must be "Equal", a company name, or "industry". Output only one word or a company name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_151_enterprise_industry_analysis_easy_easy081.md b/tasks/task_151_enterprise_industry_analysis_easy_easy081.md
new file mode 100644
index 0000000000000000000000000000000000000000..2dd35f61db210137d1d0d6991ce5961b81b092e0
--- /dev/null
+++ b/tasks/task_151_enterprise_industry_analysis_easy_easy081.md
@@ -0,0 +1,117 @@
+---
+id: task_151_enterprise_industry_analysis_easy_easy081
+name: enterprise_industry_analysis-easy-easy081
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy081.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is greater: the total number of enterprises in Jilin Province's chemical raw materials and chemical products manufacturing industry, or the number of SSE-listed enterprises in the same industry in Qinghai Province?
+
+Output guidelines:
+The answer must be "Equal", "Qinghai", or "Jilin" (as specified by the question). If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_152_enterprise_industry_analysis_easy_easy082.md b/tasks/task_152_enterprise_industry_analysis_easy_easy082.md
new file mode 100644
index 0000000000000000000000000000000000000000..125c72a4f6395c95bd0b579fea2bed9f25796562
--- /dev/null
+++ b/tasks/task_152_enterprise_industry_analysis_easy_easy082.md
@@ -0,0 +1,117 @@
+---
+id: task_152_enterprise_industry_analysis_easy_easy082
+name: enterprise_industry_analysis-easy-easy082
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy082.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is higher: the number of SZSE-listed private enterprises in Jilin's chemical raw materials and chemical products manufacturing industry, or the number of SSE-listed local state-owned enterprises in the same industry in Qinghai?
+
+Output guidelines:
+The answer must be "Equal", a company name, or "industry". Output only one word or company name without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_153_enterprise_industry_analysis_easy_easy083.md b/tasks/task_153_enterprise_industry_analysis_easy_easy083.md
new file mode 100644
index 0000000000000000000000000000000000000000..88eed661853eabd7893ffe5b4b49b8687fd8699d
--- /dev/null
+++ b/tasks/task_153_enterprise_industry_analysis_easy_easy083.md
@@ -0,0 +1,117 @@
+---
+id: task_153_enterprise_industry_analysis_easy_easy083
+name: enterprise_industry_analysis-easy-easy083
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy083.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is higher: the average capitalized R&D investment in Jilin's petroleum processing, coking, and nuclear fuel processing industry, or the same metric in Xinjiang?
+
+Output guidelines:
+The answer must be "Equal", a company name, or "industry". Output only one word or company name without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_154_enterprise_industry_analysis_easy_easy084.md b/tasks/task_154_enterprise_industry_analysis_easy_easy084.md
new file mode 100644
index 0000000000000000000000000000000000000000..dff61f320b580288a14a096dec38d13a09326299
--- /dev/null
+++ b/tasks/task_154_enterprise_industry_analysis_easy_easy084.md
@@ -0,0 +1,117 @@
+---
+id: task_154_enterprise_industry_analysis_easy_easy084
+name: enterprise_industry_analysis-easy-easy084
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy084.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is larger: the number of HKEX-listed central state-owned enterprises in China's petroleum processing, coking, and nuclear fuel processing industry, or the total number of enterprises in the textiles, footwear, and apparel industry?
+
+Output guidelines:
+The answer must be either "Petroleum Processing, Coking and Nuclear Fuel Processing" or "Textiles, Footwear and Apparel". If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Textiles, Footwear and Apparel"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_155_enterprise_industry_analysis_easy_easy085.md b/tasks/task_155_enterprise_industry_analysis_easy_easy085.md
new file mode 100644
index 0000000000000000000000000000000000000000..ecb26e684135c888bc12ece8c99f086cd1e9a6d7
--- /dev/null
+++ b/tasks/task_155_enterprise_industry_analysis_easy_easy085.md
@@ -0,0 +1,117 @@
+---
+id: task_155_enterprise_industry_analysis_easy_easy085
+name: enterprise_industry_analysis-easy-easy085
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy085.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is higher: the median number of National Technological Invention Awards in China's accommodation and catering industry, or in the real estate industry?
+
+Output guidelines:
+The answer must be "Equal", a company name, or "industry". Output only one word or company name without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_156_enterprise_industry_analysis_easy_easy086.md b/tasks/task_156_enterprise_industry_analysis_easy_easy086.md
new file mode 100644
index 0000000000000000000000000000000000000000..8bf0f6dfb63d3a4c40c7eb895b958f0422306856
--- /dev/null
+++ b/tasks/task_156_enterprise_industry_analysis_easy_easy086.md
@@ -0,0 +1,117 @@
+---
+id: task_156_enterprise_industry_analysis_easy_easy086
+name: enterprise_industry_analysis-easy-easy086
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy086.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is lower: the minimum cumulative number of Chinese invention patent applications in China's automobile manufacturing industry, or in the metal smelting and rolling processing industry?
+
+Output guidelines:
+The answer must be an industry name. Output only one term without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Metal Smelting and Rolling Processing"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_157_enterprise_industry_analysis_easy_easy088.md b/tasks/task_157_enterprise_industry_analysis_easy_easy088.md
new file mode 100644
index 0000000000000000000000000000000000000000..285bda7aeaaeff78130386b33274a826bdc5220c
--- /dev/null
+++ b/tasks/task_157_enterprise_industry_analysis_easy_easy088.md
@@ -0,0 +1,117 @@
+---
+id: task_157_enterprise_industry_analysis_easy_easy088
+name: enterprise_industry_analysis-easy-easy088
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy088.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Were the policy "Notice on Qualification Recognition Matters for Relevant R&D Institutions in Pudong New Area of Shanghai Applicable to Import Tax Policies" and the policy "Notice of the General Office of the Shanghai Municipal People's Government on Issuing the Action Plan for Cultivating the New Metaverse Track" issued by the same department?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_158_enterprise_industry_analysis_easy_easy090.md b/tasks/task_158_enterprise_industry_analysis_easy_easy090.md
new file mode 100644
index 0000000000000000000000000000000000000000..e9e34fa86a5012d67bd6d13f4270fb05ef777557
--- /dev/null
+++ b/tasks/task_158_enterprise_industry_analysis_easy_easy090.md
@@ -0,0 +1,117 @@
+---
+id: task_158_enterprise_industry_analysis_easy_easy090
+name: enterprise_industry_analysis-easy-easy090
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy090.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the minimum change in the R&D expenditure ratio of the Information Transmission, Software and IT Services industry and that of Other Manufacturing in China, which one is smaller?
+
+Output guidelines:
+The answer must be a single number with two decimal places. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Information Transmission, Software and IT Services"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_159_enterprise_industry_analysis_easy_easy091.md b/tasks/task_159_enterprise_industry_analysis_easy_easy091.md
new file mode 100644
index 0000000000000000000000000000000000000000..10f0b3b44ec6af623b9bf89ca2f49650a6a8d716
--- /dev/null
+++ b/tasks/task_159_enterprise_industry_analysis_easy_easy091.md
@@ -0,0 +1,117 @@
+---
+id: task_159_enterprise_industry_analysis_easy_easy091
+name: enterprise_industry_analysis-easy-easy091
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy091.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Are Wuli Changyuan Wholesale Company and Xinhua Yuantong Chain Company in a competitive relationship?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_160_enterprise_industry_analysis_easy_easy092.md b/tasks/task_160_enterprise_industry_analysis_easy_easy092.md
new file mode 100644
index 0000000000000000000000000000000000000000..bc37509630f01f7cad281642ad775ddd14e91f9f
--- /dev/null
+++ b/tasks/task_160_enterprise_industry_analysis_easy_easy092.md
@@ -0,0 +1,117 @@
+---
+id: task_160_enterprise_industry_analysis_easy_easy092
+name: enterprise_industry_analysis-easy-easy092
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy092.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Are Huadianeng Jin Hydropower Company and Huaneng Zeze New Energy Company in a competitive relationship?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_161_enterprise_industry_analysis_easy_easy093.md b/tasks/task_161_enterprise_industry_analysis_easy_easy093.md
new file mode 100644
index 0000000000000000000000000000000000000000..95188a5b899bb5cef153ca597f450e29abd8b8e8
--- /dev/null
+++ b/tasks/task_161_enterprise_industry_analysis_easy_easy093.md
@@ -0,0 +1,117 @@
+---
+id: task_161_enterprise_industry_analysis_easy_easy093
+name: enterprise_industry_analysis-easy-easy093
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy093.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Are Run Hui Shu Ke Technology Co., Ltd. and Hang Fa Tie Chuan Hang Kong Technology Co., Ltd. in the same industry?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_162_enterprise_industry_analysis_easy_easy094.md b/tasks/task_162_enterprise_industry_analysis_easy_easy094.md
new file mode 100644
index 0000000000000000000000000000000000000000..b076e3b47a7f484243707546d293d1369a256a27
--- /dev/null
+++ b/tasks/task_162_enterprise_industry_analysis_easy_easy094.md
@@ -0,0 +1,117 @@
+---
+id: task_162_enterprise_industry_analysis_easy_easy094
+name: enterprise_industry_analysis-easy-easy094
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy094.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Are Zhong Fang Chang Da Zhong Gong Co., Ltd. and Bao Xin Zhi Zhi Xi Tong Co., Ltd. in a competitive relationship?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_163_enterprise_industry_analysis_easy_easy095.md b/tasks/task_163_enterprise_industry_analysis_easy_easy095.md
new file mode 100644
index 0000000000000000000000000000000000000000..480a9517a60b2692a3f97a575055f4adaa534dca
--- /dev/null
+++ b/tasks/task_163_enterprise_industry_analysis_easy_easy095.md
@@ -0,0 +1,117 @@
+---
+id: task_163_enterprise_industry_analysis_easy_easy095
+name: enterprise_industry_analysis-easy-easy095
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy095.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Are Da Zu Jin Jing She Bei Co., Ltd. and Xi Fen Ye Jin Jin Shu Co., Ltd. competitors?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_164_enterprise_industry_analysis_easy_easy096.md b/tasks/task_164_enterprise_industry_analysis_easy_easy096.md
new file mode 100644
index 0000000000000000000000000000000000000000..3b95c3c8b44c4b53145d963aa5dae59392ad3307
--- /dev/null
+++ b/tasks/task_164_enterprise_industry_analysis_easy_easy096.md
@@ -0,0 +1,117 @@
+---
+id: task_164_enterprise_industry_analysis_easy_easy096
+name: enterprise_industry_analysis-easy-easy096
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy096.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Are Lv Tai Jie Xun Huan Bao Technology Co., Ltd. and Feng Huo Chuang Ze Wang Luo She Bei Co., Ltd. competitors?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_165_enterprise_industry_analysis_easy_easy097.md b/tasks/task_165_enterprise_industry_analysis_easy_easy097.md
new file mode 100644
index 0000000000000000000000000000000000000000..b1ef8b9fff9b4fa23408a40aab957af35c217780
--- /dev/null
+++ b/tasks/task_165_enterprise_industry_analysis_easy_easy097.md
@@ -0,0 +1,117 @@
+---
+id: task_165_enterprise_industry_analysis_easy_easy097
+name: enterprise_industry_analysis-easy-easy097
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy097.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Are Lv Shan Zhi Jin Real Estate Development Co., Ltd. and Huan Qiu Tai Jin Zhi Neng Dian Qi Co., Ltd. in the same industry?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_166_enterprise_industry_analysis_easy_easy098.md b/tasks/task_166_enterprise_industry_analysis_easy_easy098.md
new file mode 100644
index 0000000000000000000000000000000000000000..1dc969281cfc91f0eae8879035e3c376e8570190
--- /dev/null
+++ b/tasks/task_166_enterprise_industry_analysis_easy_easy098.md
@@ -0,0 +1,117 @@
+---
+id: task_166_enterprise_industry_analysis_easy_easy098
+name: enterprise_industry_analysis-easy-easy098
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy098.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Are Jingxin Ruihui Microelectronics Company and Ruixin Yaolan Integrated Circuit Company in the same industry?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_167_enterprise_industry_analysis_easy_easy099.md b/tasks/task_167_enterprise_industry_analysis_easy_easy099.md
new file mode 100644
index 0000000000000000000000000000000000000000..049c2209b6c8c9c6b4a5842a2b57126f8299b4b5
--- /dev/null
+++ b/tasks/task_167_enterprise_industry_analysis_easy_easy099.md
@@ -0,0 +1,117 @@
+---
+id: task_167_enterprise_industry_analysis_easy_easy099
+name: enterprise_industry_analysis-easy-easy099
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy099.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Will a downturn in the rubber and plastic products industry directly affect the operating conditions of Yao Shi Yuan Ze Sheng Wu Yi Yao Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_168_enterprise_industry_analysis_easy_easy100.md b/tasks/task_168_enterprise_industry_analysis_easy_easy100.md
new file mode 100644
index 0000000000000000000000000000000000000000..3a9eab0c1defda133647468ca549bedede462e41
--- /dev/null
+++ b/tasks/task_168_enterprise_industry_analysis_easy_easy100.md
@@ -0,0 +1,117 @@
+---
+id: task_168_enterprise_industry_analysis_easy_easy100
+name: enterprise_industry_analysis-easy-easy100
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy100.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Will a downturn in the Information Transmission, Software and IT Services industry directly affect the operating conditions of Zhong Ji Chang Yuan Gang Tie Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_169_enterprise_industry_analysis_easy_easy101.md b/tasks/task_169_enterprise_industry_analysis_easy_easy101.md
new file mode 100644
index 0000000000000000000000000000000000000000..53300708e0614116d33e2daf7e715dc01fd22bb3
--- /dev/null
+++ b/tasks/task_169_enterprise_industry_analysis_easy_easy101.md
@@ -0,0 +1,117 @@
+---
+id: task_169_enterprise_industry_analysis_easy_easy101
+name: enterprise_industry_analysis-easy-easy101
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy101.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Does the non-metallic mineral products industry include Lang Ji Lian Chuang Xin Xi Ji Shu Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_170_enterprise_industry_analysis_easy_easy102.md b/tasks/task_170_enterprise_industry_analysis_easy_easy102.md
new file mode 100644
index 0000000000000000000000000000000000000000..8e4e3305f27582277a815edda42bc4aa63245455
--- /dev/null
+++ b/tasks/task_170_enterprise_industry_analysis_easy_easy102.md
@@ -0,0 +1,117 @@
+---
+id: task_170_enterprise_industry_analysis_easy_easy102
+name: enterprise_industry_analysis-easy-easy102
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy102.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Does Lu An Fu Chang Mei Tan Co., Ltd. belong to the Comprehensive Utilization of Waste Resources industry?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_171_enterprise_industry_analysis_easy_easy103.md b/tasks/task_171_enterprise_industry_analysis_easy_easy103.md
new file mode 100644
index 0000000000000000000000000000000000000000..a7bee8997653da7f8d0452615f03b4c9e2e5ef4c
--- /dev/null
+++ b/tasks/task_171_enterprise_industry_analysis_easy_easy103.md
@@ -0,0 +1,117 @@
+---
+id: task_171_enterprise_industry_analysis_easy_easy103
+name: enterprise_industry_analysis-easy-easy103
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy103.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Does Heng Li Yun Chuang Xin Xi Ji Shu Co., Ltd. belong to the Automobile Manufacturing industry?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_172_enterprise_industry_analysis_easy_easy104.md b/tasks/task_172_enterprise_industry_analysis_easy_easy104.md
new file mode 100644
index 0000000000000000000000000000000000000000..de9981cecaa5c642728106c1f024c06f1598c976
--- /dev/null
+++ b/tasks/task_172_enterprise_industry_analysis_easy_easy104.md
@@ -0,0 +1,117 @@
+---
+id: task_172_enterprise_industry_analysis_easy_easy104
+name: enterprise_industry_analysis-easy-easy104
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy104.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Does Wan Hui Sheng Zhi Construction Development Co., Ltd. belong to the Real Estate industry?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_173_enterprise_industry_analysis_easy_easy105.md b/tasks/task_173_enterprise_industry_analysis_easy_easy105.md
new file mode 100644
index 0000000000000000000000000000000000000000..bf461b29b46fc54d4d02af2518b8174b9747ad52
--- /dev/null
+++ b/tasks/task_173_enterprise_industry_analysis_easy_easy105.md
@@ -0,0 +1,117 @@
+---
+id: task_173_enterprise_industry_analysis_easy_easy105
+name: enterprise_industry_analysis-easy-easy105
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy105.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Does Hang Fa Yuan Jin Hang Kong Technology Co., Ltd. belong to Education?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_174_enterprise_industry_analysis_easy_easy106.md b/tasks/task_174_enterprise_industry_analysis_easy_easy106.md
new file mode 100644
index 0000000000000000000000000000000000000000..64048df5cad3773f0cc5f0be3090769b1ba8cea4
--- /dev/null
+++ b/tasks/task_174_enterprise_industry_analysis_easy_easy106.md
@@ -0,0 +1,117 @@
+---
+id: task_174_enterprise_industry_analysis_easy_easy106
+name: enterprise_industry_analysis-easy-easy106
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy106.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Will a downturn in the Water Conservancy, Environment and Public Facilities Management industry directly affect the operating conditions of Hua Lu Rong Rong Hua Xue Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_175_enterprise_industry_analysis_easy_easy107.md b/tasks/task_175_enterprise_industry_analysis_easy_easy107.md
new file mode 100644
index 0000000000000000000000000000000000000000..44abbe72c3376cf72a1609a63ec50eed1921b7bd
--- /dev/null
+++ b/tasks/task_175_enterprise_industry_analysis_easy_easy107.md
@@ -0,0 +1,117 @@
+---
+id: task_175_enterprise_industry_analysis_easy_easy107
+name: enterprise_industry_analysis-easy-easy107
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy107.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Does San San Gong Ji Technology Co., Ltd. belong to the Non-metallic Mineral Products industry?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_176_enterprise_industry_analysis_easy_easy108.md b/tasks/task_176_enterprise_industry_analysis_easy_easy108.md
new file mode 100644
index 0000000000000000000000000000000000000000..4ee48a0b4601f3af41023f07d5a7e96e2b097a2d
--- /dev/null
+++ b/tasks/task_176_enterprise_industry_analysis_easy_easy108.md
@@ -0,0 +1,117 @@
+---
+id: task_176_enterprise_industry_analysis_easy_easy108
+name: enterprise_industry_analysis-easy-easy108
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy108.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is Zhong Ju Yue Yin Shi Pin Co., Ltd. registered in Chongqing Municipality?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_177_enterprise_industry_analysis_easy_easy109.md b/tasks/task_177_enterprise_industry_analysis_easy_easy109.md
new file mode 100644
index 0000000000000000000000000000000000000000..9c1a54fc4b7cc312e3e50814240309e1525a6fda
--- /dev/null
+++ b/tasks/task_177_enterprise_industry_analysis_easy_easy109.md
@@ -0,0 +1,117 @@
+---
+id: task_177_enterprise_industry_analysis_easy_easy109
+name: enterprise_industry_analysis-easy-easy109
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy109.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Does Hua Cheng Jin Jin Zong He Kai Fa Co., Ltd. contribute to the development of Zhejiang Province?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_178_enterprise_industry_analysis_easy_easy110.md b/tasks/task_178_enterprise_industry_analysis_easy_easy110.md
new file mode 100644
index 0000000000000000000000000000000000000000..abf3394e831d1c5e8e7aac3b5341693c67391166
--- /dev/null
+++ b/tasks/task_178_enterprise_industry_analysis_easy_easy110.md
@@ -0,0 +1,117 @@
+---
+id: task_178_enterprise_industry_analysis_easy_easy110
+name: enterprise_industry_analysis-easy-easy110
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy110.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Does Guangdong Province have the enterprise Lv Shan Chan Jin Zhi Ye Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_179_enterprise_industry_analysis_easy_easy111.md b/tasks/task_179_enterprise_industry_analysis_easy_easy111.md
new file mode 100644
index 0000000000000000000000000000000000000000..c44987d6a1ed6bce043b66cbb641be3cb315c5c6
--- /dev/null
+++ b/tasks/task_179_enterprise_industry_analysis_easy_easy111.md
@@ -0,0 +1,117 @@
+---
+id: task_179_enterprise_industry_analysis_easy_easy111
+name: enterprise_industry_analysis-easy-easy111
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy111.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is Hua Xin Ze Chang Xin Cai Liao Co., Ltd. registered in Hebei Province?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_180_enterprise_industry_analysis_easy_easy112.md b/tasks/task_180_enterprise_industry_analysis_easy_easy112.md
new file mode 100644
index 0000000000000000000000000000000000000000..39d3618ed2bdbad42b85cb6f3d3cc8fa587399b6
--- /dev/null
+++ b/tasks/task_180_enterprise_industry_analysis_easy_easy112.md
@@ -0,0 +1,117 @@
+---
+id: task_180_enterprise_industry_analysis_easy_easy112
+name: enterprise_industry_analysis-easy-easy112
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy112.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Does Wan Hui Jin Sheng Real Estate Development Co., Ltd. contribute to the development of Guangdong Province?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_181_enterprise_industry_analysis_easy_easy113.md b/tasks/task_181_enterprise_industry_analysis_easy_easy113.md
new file mode 100644
index 0000000000000000000000000000000000000000..f28205e249e188a947ef83734f51a85db914861a
--- /dev/null
+++ b/tasks/task_181_enterprise_industry_analysis_easy_easy113.md
@@ -0,0 +1,117 @@
+---
+id: task_181_enterprise_industry_analysis_easy_easy113
+name: enterprise_industry_analysis-easy-easy113
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy113.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Will changes in Anhui Province's economic environment affect Bao Jin Jin Chang Tong Ye Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_182_enterprise_industry_analysis_easy_easy114.md b/tasks/task_182_enterprise_industry_analysis_easy_easy114.md
new file mode 100644
index 0000000000000000000000000000000000000000..cd40bdc537e33a90e48b593c99a95111a513eb72
--- /dev/null
+++ b/tasks/task_182_enterprise_industry_analysis_easy_easy114.md
@@ -0,0 +1,117 @@
+---
+id: task_182_enterprise_industry_analysis_easy_easy114
+name: enterprise_industry_analysis-easy-easy114
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy114.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Will changes in Guizhou Province's economic environment affect Zhong You Zheng Da Jin Yun Shu Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_183_enterprise_industry_analysis_easy_easy115.md b/tasks/task_183_enterprise_industry_analysis_easy_easy115.md
new file mode 100644
index 0000000000000000000000000000000000000000..6fbeff98d86207a6bb5e4c546763f99be9299a7f
--- /dev/null
+++ b/tasks/task_183_enterprise_industry_analysis_easy_easy115.md
@@ -0,0 +1,117 @@
+---
+id: task_183_enterprise_industry_analysis_easy_easy115
+name: enterprise_industry_analysis-easy-easy115
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy115.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Does Huan Xing Jin Ya Apparel Co., Ltd. contribute to the development of Jiangxi Province?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_184_enterprise_industry_analysis_easy_easy116.md b/tasks/task_184_enterprise_industry_analysis_easy_easy116.md
new file mode 100644
index 0000000000000000000000000000000000000000..ae86a9d25e31f31955d2a2f62abdcb6315825ac5
--- /dev/null
+++ b/tasks/task_184_enterprise_industry_analysis_easy_easy116.md
@@ -0,0 +1,117 @@
+---
+id: task_184_enterprise_industry_analysis_easy_easy116
+name: enterprise_industry_analysis-easy-easy116
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy116.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Would changes in Guangdong Province's economic environment affect Hua Dian Neng Jin Hydropower Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_185_enterprise_industry_analysis_easy_easy117.md b/tasks/task_185_enterprise_industry_analysis_easy_easy117.md
new file mode 100644
index 0000000000000000000000000000000000000000..4b3d977388163fccd78e71705f1ba88bbd02ba81
--- /dev/null
+++ b/tasks/task_185_enterprise_industry_analysis_easy_easy117.md
@@ -0,0 +1,117 @@
+---
+id: task_185_enterprise_industry_analysis_easy_easy117
+name: enterprise_industry_analysis-easy-easy117
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/easy117.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is Zhongke Zhiyun Data Services Co., Ltd. registered in Guangdong Province?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_186_enterprise_industry_analysis_medium_medium001.md b/tasks/task_186_enterprise_industry_analysis_medium_medium001.md
new file mode 100644
index 0000000000000000000000000000000000000000..c2cfafbf3c32e5230761935bd03cad6fc236059b
--- /dev/null
+++ b/tasks/task_186_enterprise_industry_analysis_medium_medium001.md
@@ -0,0 +1,117 @@
+---
+id: task_186_enterprise_industry_analysis_medium_medium001
+name: enterprise_industry_analysis-medium-medium001
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium001.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, what is the difference between the year-over-year employee change rate of Kangsheng Kangjian Pharmaceutical Co., Ltd. and the minimum level of its industry?
+
+Output guidelines:
+The answer must be a number with two decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`94.35`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_187_enterprise_industry_analysis_medium_medium002.md b/tasks/task_187_enterprise_industry_analysis_medium_medium002.md
new file mode 100644
index 0000000000000000000000000000000000000000..b60a7528867343ce1620c39fbc30e738dedb1f63
--- /dev/null
+++ b/tasks/task_187_enterprise_industry_analysis_medium_medium002.md
@@ -0,0 +1,117 @@
+---
+id: task_187_enterprise_industry_analysis_medium_medium002
+name: enterprise_industry_analysis-medium-medium002
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium002.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, what is the difference between the total number of employees of Kangsheng Kangjian Pharmaceutical Co., Ltd. and the industry average?
+
+Output guidelines:
+The answer must be an exact number and keep all significant decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-958.60986547085`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_188_enterprise_industry_analysis_medium_medium003.md b/tasks/task_188_enterprise_industry_analysis_medium_medium003.md
new file mode 100644
index 0000000000000000000000000000000000000000..1e6f6dbd5894092c57dd952b46491a75f4735272
--- /dev/null
+++ b/tasks/task_188_enterprise_industry_analysis_medium_medium003.md
@@ -0,0 +1,117 @@
+---
+id: task_188_enterprise_industry_analysis_medium_medium003
+name: enterprise_industry_analysis-medium-medium003
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium003.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, what is the difference between the total number of employees of Ling You Se Ye Da Zi Yuan Co., Ltd. and the industry maximum?
+
+Output guidelines:
+The answer must be a number with one decimal place. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-65010.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_189_enterprise_industry_analysis_medium_medium004.md b/tasks/task_189_enterprise_industry_analysis_medium_medium004.md
new file mode 100644
index 0000000000000000000000000000000000000000..e5cc80d910e97f007667729c766bbc2099cbd9d2
--- /dev/null
+++ b/tasks/task_189_enterprise_industry_analysis_medium_medium004.md
@@ -0,0 +1,117 @@
+---
+id: task_189_enterprise_industry_analysis_medium_medium004
+name: enterprise_industry_analysis-medium-medium004
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium004.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, what is the difference between the year-over-year net profit change rate of Ling You Se Ye Da Zi Yuan Co., Ltd. and the minimum value of this indicator in its industry?
+
+Output guidelines:
+The answer must be a number with two decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`2110.42`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_190_enterprise_industry_analysis_medium_medium005.md b/tasks/task_190_enterprise_industry_analysis_medium_medium005.md
new file mode 100644
index 0000000000000000000000000000000000000000..c20e2378e3701c0f076f02f123da2dc0c1885416
--- /dev/null
+++ b/tasks/task_190_enterprise_industry_analysis_medium_medium005.md
@@ -0,0 +1,117 @@
+---
+id: task_190_enterprise_industry_analysis_medium_medium005
+name: enterprise_industry_analysis-medium-medium005
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium005.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is higher: the year-over-year employee change rate of Yong Feng Xin Chuang Ke Ji Co., Ltd. or the maximum value of this indicator in its industry?
+
+Output guidelines:
+The answer must be either "industry" or the company name. Output only one word or the company name, without any explanation, analysis, or descriptive text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_191_enterprise_industry_analysis_medium_medium006.md b/tasks/task_191_enterprise_industry_analysis_medium_medium006.md
new file mode 100644
index 0000000000000000000000000000000000000000..c9d26558069b6e0b617489e7d663a6aacb500a34
--- /dev/null
+++ b/tasks/task_191_enterprise_industry_analysis_medium_medium006.md
@@ -0,0 +1,117 @@
+---
+id: task_191_enterprise_industry_analysis_medium_medium006
+name: enterprise_industry_analysis-medium-medium006
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium006.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the year-over-year net profit change rate of Yong Feng Xin Chuang Ke Ji Co., Ltd. higher than the median of this indicator in its industry?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_192_enterprise_industry_analysis_medium_medium007.md b/tasks/task_192_enterprise_industry_analysis_medium_medium007.md
new file mode 100644
index 0000000000000000000000000000000000000000..231c495f937b1418b3e7624c99ba7e174acbf41d
--- /dev/null
+++ b/tasks/task_192_enterprise_industry_analysis_medium_medium007.md
@@ -0,0 +1,117 @@
+---
+id: task_192_enterprise_industry_analysis_medium_medium007
+name: enterprise_industry_analysis-medium-medium007
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium007.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, what is the difference between the operating profit amount of Lian Ji Chuang Ji Ji Chuang Co., Ltd. and the total operating profit amount of the same industry in its province?
+
+Output guidelines:
+The answer must be a number with two decimal places. Output only the number without units, commas, or any additional text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-1738758524.92`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_193_enterprise_industry_analysis_medium_medium008.md b/tasks/task_193_enterprise_industry_analysis_medium_medium008.md
new file mode 100644
index 0000000000000000000000000000000000000000..4fad1399b90bafa37e3b572272a066dc370f7f86
--- /dev/null
+++ b/tasks/task_193_enterprise_industry_analysis_medium_medium008.md
@@ -0,0 +1,117 @@
+---
+id: task_193_enterprise_industry_analysis_medium_medium008
+name: enterprise_industry_analysis-medium-medium008
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium008.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, compared with the average level of the same industry in its province, which is higher: the number of R&D personnel of Lianji Chuangji Machine Tool Company or the industry average?
+
+Output guidelines:
+The answer must be either the company name or the word "industry". Output only the name, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Lianji Chuangji Machine Tool Company"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_194_enterprise_industry_analysis_medium_medium009.md b/tasks/task_194_enterprise_industry_analysis_medium_medium009.md
new file mode 100644
index 0000000000000000000000000000000000000000..e228a6d680e062a7d6c53537d725c00f79ac2e13
--- /dev/null
+++ b/tasks/task_194_enterprise_industry_analysis_medium_medium009.md
@@ -0,0 +1,117 @@
+---
+id: task_194_enterprise_industry_analysis_medium_medium009
+name: enterprise_industry_analysis-medium-medium009
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium009.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is higher: the operating profit amount of Run Hui Shu Zhi Xi Tong Co., Ltd. or the total operating profit amount of the same industry in its province?
+
+Output guidelines:
+The answer must be either the company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Run Hui Shu Zhi Xi Tong Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_195_enterprise_industry_analysis_medium_medium010.md b/tasks/task_195_enterprise_industry_analysis_medium_medium010.md
new file mode 100644
index 0000000000000000000000000000000000000000..85da826e989d3676255d5d1b13c8d672a28f9764
--- /dev/null
+++ b/tasks/task_195_enterprise_industry_analysis_medium_medium010.md
@@ -0,0 +1,117 @@
+---
+id: task_195_enterprise_industry_analysis_medium_medium010
+name: enterprise_industry_analysis-medium-medium010
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium010.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the operating revenue amount of Run Hui Shu Zhi Xi Tong Co., Ltd. higher than the total operating revenue amount of the corresponding industry in its province?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_196_enterprise_industry_analysis_medium_medium011.md b/tasks/task_196_enterprise_industry_analysis_medium_medium011.md
new file mode 100644
index 0000000000000000000000000000000000000000..3e7b239e0c928208fe4f6b8468ba96929530623a
--- /dev/null
+++ b/tasks/task_196_enterprise_industry_analysis_medium_medium011.md
@@ -0,0 +1,117 @@
+---
+id: task_196_enterprise_industry_analysis_medium_medium011
+name: enterprise_industry_analysis-medium-medium011
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium011.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, which is higher: the cumulative number of PCT invention patent applications of Zhong Ji Da Chang Tong Ye Co., Ltd. or the minimum value of the same indicator in the same industry in its province?
+
+Output guidelines:
+The answer must be "equal", the company name, or "industry". Output only one word or the company name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_197_enterprise_industry_analysis_medium_medium012.md b/tasks/task_197_enterprise_industry_analysis_medium_medium012.md
new file mode 100644
index 0000000000000000000000000000000000000000..a76ce3d9828a0b0728a1388cdf5156b0cb7c7376
--- /dev/null
+++ b/tasks/task_197_enterprise_industry_analysis_medium_medium012.md
@@ -0,0 +1,117 @@
+---
+id: task_197_enterprise_industry_analysis_medium_medium012
+name: enterprise_industry_analysis-medium-medium012
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium012.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the debt-to-asset ratio of Zhong Ji Da Chang Tong Ye Co., Ltd. higher than the minimum debt-to-asset ratio of the corresponding industry in its province?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_198_enterprise_industry_analysis_medium_medium013.md b/tasks/task_198_enterprise_industry_analysis_medium_medium013.md
new file mode 100644
index 0000000000000000000000000000000000000000..fc1e49a91eb9af0abd6b6069aa68535a7d960ced
--- /dev/null
+++ b/tasks/task_198_enterprise_industry_analysis_medium_medium013.md
@@ -0,0 +1,117 @@
+---
+id: task_198_enterprise_industry_analysis_medium_medium013
+name: enterprise_industry_analysis-medium-medium013
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium013.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, is the market capitalization of Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. lower than the operating revenue of Long He Zhi Jin Zhi Ye Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_199_enterprise_industry_analysis_medium_medium014.md b/tasks/task_199_enterprise_industry_analysis_medium_medium014.md
new file mode 100644
index 0000000000000000000000000000000000000000..f77fd5be00ed059a15c595599bc4c3e0395b2733
--- /dev/null
+++ b/tasks/task_199_enterprise_industry_analysis_medium_medium014.md
@@ -0,0 +1,117 @@
+---
+id: task_199_enterprise_industry_analysis_medium_medium014
+name: enterprise_industry_analysis-medium-medium014
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium014.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Comparing the number of SSE-listed state-owned enterprise institutes in the industry of Huijin Jinrui Wealth Management Co., Ltd. with the number of HKEX-listed sino-foreign joint ventures in the industry of Zhongke Zhiyun Data Services Co., Ltd., which value is larger?
+
+Output guidelines:
+The answer must be a company name, "industry", or "Same". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Same"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_200_enterprise_industry_analysis_medium_medium015.md b/tasks/task_200_enterprise_industry_analysis_medium_medium015.md
new file mode 100644
index 0000000000000000000000000000000000000000..6cb144b1358a13f2b0500749655554c5d97c31fb
--- /dev/null
+++ b/tasks/task_200_enterprise_industry_analysis_medium_medium015.md
@@ -0,0 +1,117 @@
+---
+id: task_200_enterprise_industry_analysis_medium_medium015
+name: enterprise_industry_analysis-medium-medium015
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium015.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is the minimum R&D personnel ratio in the industry of Huijin Jinrui Wealth Management Co., Ltd. lower than the minimum R&D personnel ratio in the industry of Zhongke Zhiyun Data Services Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_201_enterprise_industry_analysis_medium_medium016.md b/tasks/task_201_enterprise_industry_analysis_medium_medium016.md
new file mode 100644
index 0000000000000000000000000000000000000000..0f06951190d565610d978d92f9d0e7ddaa29b377
--- /dev/null
+++ b/tasks/task_201_enterprise_industry_analysis_medium_medium016.md
@@ -0,0 +1,117 @@
+---
+id: task_201_enterprise_industry_analysis_medium_medium016
+name: enterprise_industry_analysis-medium-medium016
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium016.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the median annual number of Chinese invention patent applications in the industry of Changqiao Jinchuang Technology Co., Ltd. and the corresponding metric for the industry in Zhejiang Province where Wuli Huida Chain Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`16.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_202_enterprise_industry_analysis_medium_medium017.md b/tasks/task_202_enterprise_industry_analysis_medium_medium017.md
new file mode 100644
index 0000000000000000000000000000000000000000..bbb071400afe20c1799ea30b25ab8fbad502e1bc
--- /dev/null
+++ b/tasks/task_202_enterprise_industry_analysis_medium_medium017.md
@@ -0,0 +1,117 @@
+---
+id: task_202_enterprise_industry_analysis_medium_medium017
+name: enterprise_industry_analysis-medium-medium017
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium017.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the minimum R&D personnel ratio in the industry of Changqiao Jinchuang Technology Co., Ltd. and the corresponding metric for the industry in Zhejiang Province where Wuli Huida Chain Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a number rounded to two decimal places. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`1.04`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_203_enterprise_industry_analysis_medium_medium018.md b/tasks/task_203_enterprise_industry_analysis_medium_medium018.md
new file mode 100644
index 0000000000000000000000000000000000000000..0ba5ef73f8e5f7b74768295ef0bf532cb1a37072
--- /dev/null
+++ b/tasks/task_203_enterprise_industry_analysis_medium_medium018.md
@@ -0,0 +1,117 @@
+---
+id: task_203_enterprise_industry_analysis_medium_medium018
+name: enterprise_industry_analysis-medium-medium018
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium018.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of SSE-listed local state-owned enterprises in the corresponding industry of the province where Zhong Ke Shu Ruan Software Co., Ltd. is located, or the number of SZSE-listed foreign-funded enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_204_enterprise_industry_analysis_medium_medium019.md b/tasks/task_204_enterprise_industry_analysis_medium_medium019.md
new file mode 100644
index 0000000000000000000000000000000000000000..12ee8e86d3310c7f2ecf86fb5bd2e1dcd9350c47
--- /dev/null
+++ b/tasks/task_204_enterprise_industry_analysis_medium_medium019.md
@@ -0,0 +1,117 @@
+---
+id: task_204_enterprise_industry_analysis_medium_medium019
+name: enterprise_industry_analysis-medium-medium019
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium019.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of HKEX-listed foreign-funded enterprises in the corresponding industry of the province where Zhong Ke Shu Ruan Software Co., Ltd. is located, or the number of HKEX-listed enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_205_enterprise_industry_analysis_medium_medium020.md b/tasks/task_205_enterprise_industry_analysis_medium_medium020.md
new file mode 100644
index 0000000000000000000000000000000000000000..a9a26ce19f92b2afd8d58d8d13e584efde77e306
--- /dev/null
+++ b/tasks/task_205_enterprise_industry_analysis_medium_medium020.md
@@ -0,0 +1,117 @@
+---
+id: task_205_enterprise_industry_analysis_medium_medium020
+name: enterprise_industry_analysis-medium-medium020
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium020.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of SZSE-listed central state-owned enterprises in the industry of Bi Yuan Zhi Ze Urban Development Co., Ltd., or the number of SSE-listed local state-owned enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_206_enterprise_industry_analysis_medium_medium021.md b/tasks/task_206_enterprise_industry_analysis_medium_medium021.md
new file mode 100644
index 0000000000000000000000000000000000000000..9c190c65ed291c9659ee8e32a8f80837ebe55497
--- /dev/null
+++ b/tasks/task_206_enterprise_industry_analysis_medium_medium021.md
@@ -0,0 +1,117 @@
+---
+id: task_206_enterprise_industry_analysis_medium_medium021
+name: enterprise_industry_analysis-medium-medium021
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium021.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the median operating profit amount of the industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd. and that of the industry of Tong Tong Ze Hong Securities Co., Ltd.?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-880561639.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_207_enterprise_industry_analysis_medium_medium022.md b/tasks/task_207_enterprise_industry_analysis_medium_medium022.md
new file mode 100644
index 0000000000000000000000000000000000000000..68dc983fd26ad8c337900f89b2646ec7080a2171
--- /dev/null
+++ b/tasks/task_207_enterprise_industry_analysis_medium_medium022.md
@@ -0,0 +1,117 @@
+---
+id: task_207_enterprise_industry_analysis_medium_medium022
+name: enterprise_industry_analysis-medium-medium022
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium022.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Comparing the number of SZSE-listed central state-owned enterprises in the industry of Zhaoye Huachang Real Estate Development Co., Ltd. with the number of SZSE-listed enterprises in the industry of Tongtong Zehong Securities Co., Ltd., which is larger?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Tongtong Zehong Securities Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_208_enterprise_industry_analysis_medium_medium023.md b/tasks/task_208_enterprise_industry_analysis_medium_medium023.md
new file mode 100644
index 0000000000000000000000000000000000000000..4ac87e23c93724eccd3ff0238f31de9d3f55abb8
--- /dev/null
+++ b/tasks/task_208_enterprise_industry_analysis_medium_medium023.md
@@ -0,0 +1,117 @@
+---
+id: task_208_enterprise_industry_analysis_medium_medium023
+name: enterprise_industry_analysis-medium-medium023
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium023.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Comparing the number of SZSE-listed local state-owned enterprises in the industry of Aijian Yikang Fuzhongxin Co., Ltd. with the number of SZSE-listed sino-foreign joint ventures in the industry of Zhongke Zhiyun Data Services Co., Ltd., which value is larger?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Zhongke Zhiyun Data Services Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_209_enterprise_industry_analysis_medium_medium024.md b/tasks/task_209_enterprise_industry_analysis_medium_medium024.md
new file mode 100644
index 0000000000000000000000000000000000000000..24f23d2d50c0159e58918f80cc1e5acb910547a5
--- /dev/null
+++ b/tasks/task_209_enterprise_industry_analysis_medium_medium024.md
@@ -0,0 +1,117 @@
+---
+id: task_209_enterprise_industry_analysis_medium_medium024
+name: enterprise_industry_analysis-medium-medium024
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium024.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the minimum value of provincial enterprise technology innovation awards in the industry of Aijian Yikang Fuzhongxin Co., Ltd. and that in the industry of Zhongke Zhiyun Data Services Co., Ltd.?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_210_enterprise_industry_analysis_medium_medium025.md b/tasks/task_210_enterprise_industry_analysis_medium_medium025.md
new file mode 100644
index 0000000000000000000000000000000000000000..35587a79a41055b974fd691826bd5fdb55d8d028
--- /dev/null
+++ b/tasks/task_210_enterprise_industry_analysis_medium_medium025.md
@@ -0,0 +1,117 @@
+---
+id: task_210_enterprise_industry_analysis_medium_medium025
+name: enterprise_industry_analysis-medium-medium025
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium025.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is the average number of provincial or ministerial natural science awards in the industry of Biyuan Shenghua Construction Development Co., Ltd. lower than that in the industry of Baoxin Huihui Network Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_211_enterprise_industry_analysis_medium_medium026.md b/tasks/task_211_enterprise_industry_analysis_medium_medium026.md
new file mode 100644
index 0000000000000000000000000000000000000000..31080d24d2d73f0cd6e49c518bd7afa30ff550c2
--- /dev/null
+++ b/tasks/task_211_enterprise_industry_analysis_medium_medium026.md
@@ -0,0 +1,117 @@
+---
+id: task_211_enterprise_industry_analysis_medium_medium026
+name: enterprise_industry_analysis-medium-medium026
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium026.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is the median cumulative citation count of all patents in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. lower than the same metric in the industry of Bao Xin Hui Hui Wang Luo Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_212_enterprise_industry_analysis_medium_medium027.md b/tasks/task_212_enterprise_industry_analysis_medium_medium027.md
new file mode 100644
index 0000000000000000000000000000000000000000..6ff67b377291bb4bd5754080e916b6f8bd9d3dec
--- /dev/null
+++ b/tasks/task_212_enterprise_industry_analysis_medium_medium027.md
@@ -0,0 +1,117 @@
+---
+id: task_212_enterprise_industry_analysis_medium_medium027
+name: enterprise_industry_analysis-medium-medium027
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium027.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is the number of SSE-listed enterprises in the industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. equal to the number of SSE-listed state-owned institute enterprises in the industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_213_enterprise_industry_analysis_medium_medium028.md b/tasks/task_213_enterprise_industry_analysis_medium_medium028.md
new file mode 100644
index 0000000000000000000000000000000000000000..e7ef02a43585aeb1c33f4b2e050eead2f5456cad
--- /dev/null
+++ b/tasks/task_213_enterprise_industry_analysis_medium_medium028.md
@@ -0,0 +1,117 @@
+---
+id: task_213_enterprise_industry_analysis_medium_medium028
+name: enterprise_industry_analysis-medium-medium028
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium028.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Are the maximum values of the State Technological Invention Award metric the same between the industry of Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. and the industry of Zhong Che Yuan Ze Shipbuilding Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_214_enterprise_industry_analysis_medium_medium029.md b/tasks/task_214_enterprise_industry_analysis_medium_medium029.md
new file mode 100644
index 0000000000000000000000000000000000000000..14f4da4217d41edab6533eb56e2a1b397dfb24af
--- /dev/null
+++ b/tasks/task_214_enterprise_industry_analysis_medium_medium029.md
@@ -0,0 +1,117 @@
+---
+id: task_214_enterprise_industry_analysis_medium_medium029
+name: enterprise_industry_analysis-medium-medium029
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium029.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of SSE-listed central state-owned enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd., or the number of SSE-listed enterprises in the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_215_enterprise_industry_analysis_medium_medium030.md b/tasks/task_215_enterprise_industry_analysis_medium_medium030.md
new file mode 100644
index 0000000000000000000000000000000000000000..6062f7eaaee34ac08fc02773ced15c1768f6cea7
--- /dev/null
+++ b/tasks/task_215_enterprise_industry_analysis_medium_medium030.md
@@ -0,0 +1,117 @@
+---
+id: task_215_enterprise_industry_analysis_medium_medium030
+name: enterprise_industry_analysis-medium-medium030
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium030.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is lower: the minimum total liabilities value in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd., or the same metric in the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_216_enterprise_industry_analysis_medium_medium031.md b/tasks/task_216_enterprise_industry_analysis_medium_medium031.md
new file mode 100644
index 0000000000000000000000000000000000000000..51e18249545f18c3cf49b9f5228e61692e4b0824
--- /dev/null
+++ b/tasks/task_216_enterprise_industry_analysis_medium_medium031.md
@@ -0,0 +1,117 @@
+---
+id: task_216_enterprise_industry_analysis_medium_medium031
+name: enterprise_industry_analysis-medium-medium031
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium031.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of SSE-listed state-owned institute enterprises in the industry of Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd., or the number of SZSE-listed private enterprises in the industry of Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_217_enterprise_industry_analysis_medium_medium032.md b/tasks/task_217_enterprise_industry_analysis_medium_medium032.md
new file mode 100644
index 0000000000000000000000000000000000000000..d726a1e0112b5665aa3c01627add8f69c68d731b
--- /dev/null
+++ b/tasks/task_217_enterprise_industry_analysis_medium_medium032.md
@@ -0,0 +1,117 @@
+---
+id: task_217_enterprise_industry_analysis_medium_medium032
+name: enterprise_industry_analysis-medium-medium032
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium032.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Comparing the number of SZSE-listed private enterprises in the industry of Jinzhi Hongsheng Asset Management Co., Ltd. with the number of SSE-listed private enterprises in the industry of Biyuan Zhize Urban Development Co., Ltd., which is larger?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Jinzhi Hongsheng Asset Management Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_218_enterprise_industry_analysis_medium_medium033.md b/tasks/task_218_enterprise_industry_analysis_medium_medium033.md
new file mode 100644
index 0000000000000000000000000000000000000000..82f03a313a714b55e1b598efbd381fc1e0645b33
--- /dev/null
+++ b/tasks/task_218_enterprise_industry_analysis_medium_medium033.md
@@ -0,0 +1,117 @@
+---
+id: task_218_enterprise_industry_analysis_medium_medium033
+name: enterprise_industry_analysis-medium-medium033
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium033.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Comparing the number of enterprises in Health and Social Work in the industry of Jianfan Ningze Elderly Care Services Co., Ltd. with the number of SSE-listed central state-owned enterprises in the industry of Zhongche Yuanze Shipbuilding Co., Ltd., which value is larger?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Jianfan Ningze Elderly Care Services Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_219_enterprise_industry_analysis_medium_medium034.md b/tasks/task_219_enterprise_industry_analysis_medium_medium034.md
new file mode 100644
index 0000000000000000000000000000000000000000..68ac8981a7b8f9b1bd57c0b0c714dd416c3dffce
--- /dev/null
+++ b/tasks/task_219_enterprise_industry_analysis_medium_medium034.md
@@ -0,0 +1,117 @@
+---
+id: task_219_enterprise_industry_analysis_medium_medium034
+name: enterprise_industry_analysis-medium-medium034
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium034.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is the minimum market capitalization in the industry of Jianfan Ningze Elderly Care Services Co., Ltd. lower than the minimum market capitalization in the industry of Zhongche Yuanze Shipbuilding Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or notes. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_220_enterprise_industry_analysis_medium_medium035.md b/tasks/task_220_enterprise_industry_analysis_medium_medium035.md
new file mode 100644
index 0000000000000000000000000000000000000000..174fa15f9b6cfe95beb6533fcf81e4db618fca5f
--- /dev/null
+++ b/tasks/task_220_enterprise_industry_analysis_medium_medium035.md
@@ -0,0 +1,117 @@
+---
+id: task_220_enterprise_industry_analysis_medium_medium035
+name: enterprise_industry_analysis-medium-medium035
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium035.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Comparing the number of HKEX-listed sino-foreign joint ventures in the industry of Yihai Changjin Business Co., Ltd. with the number of SSE-listed local state-owned enterprises in the corresponding industry of Shanghai where Jinzhi Hongsheng Asset Management Co., Ltd. is located, which is larger?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Jinzhi Hongsheng Asset Management Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_221_enterprise_industry_analysis_medium_medium036.md b/tasks/task_221_enterprise_industry_analysis_medium_medium036.md
new file mode 100644
index 0000000000000000000000000000000000000000..966d48c7219552890818bad2d546824e4d604a48
--- /dev/null
+++ b/tasks/task_221_enterprise_industry_analysis_medium_medium036.md
@@ -0,0 +1,117 @@
+---
+id: task_221_enterprise_industry_analysis_medium_medium036
+name: enterprise_industry_analysis-medium-medium036
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium036.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is the mean annual number of Chinese invention patent applications in the industry of Yi Hai Chang Jin Shang Wu Co., Ltd. greater than the corresponding industry metric in Shanghai, where Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. is located?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_222_enterprise_industry_analysis_medium_medium037.md b/tasks/task_222_enterprise_industry_analysis_medium_medium037.md
new file mode 100644
index 0000000000000000000000000000000000000000..417f01d638be4c415bbb068d651ca0b35c45dd0d
--- /dev/null
+++ b/tasks/task_222_enterprise_industry_analysis_medium_medium037.md
@@ -0,0 +1,117 @@
+---
+id: task_222_enterprise_industry_analysis_medium_medium037
+name: enterprise_industry_analysis-medium-medium037
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium037.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of SZSE-listed local state-owned enterprises in the industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd., or the number of HKEX-listed enterprises in the corresponding industry of Guangdong Province where Gao Yin Ze Tong Pi Fa Co., Ltd. is located?
+
+Output guidelines:
+The answer must be either "Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. industry local state-owned enterprise SZSE-listed count" or "HKEX-listed enterprise count". Output only the answer text without explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. industry local state-owned enterprise SZSE-listed count"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_223_enterprise_industry_analysis_medium_medium038.md b/tasks/task_223_enterprise_industry_analysis_medium_medium038.md
new file mode 100644
index 0000000000000000000000000000000000000000..30a42e1d0e3571ea4389f1e44a9985e667a74acb
--- /dev/null
+++ b/tasks/task_223_enterprise_industry_analysis_medium_medium038.md
@@ -0,0 +1,117 @@
+---
+id: task_223_enterprise_industry_analysis_medium_medium038
+name: enterprise_industry_analysis-medium-medium038
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium038.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the maximum capitalized R&D expenditure of the industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. and the corresponding industry metric in Guangdong Province where Gao Yin Ze Tong Pi Fa Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a number rounded to two decimal places. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`4026077173.89`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_224_enterprise_industry_analysis_medium_medium039.md b/tasks/task_224_enterprise_industry_analysis_medium_medium039.md
new file mode 100644
index 0000000000000000000000000000000000000000..e031776913a873b1e3ad8f121bb3b57f0106d0d7
--- /dev/null
+++ b/tasks/task_224_enterprise_industry_analysis_medium_medium039.md
@@ -0,0 +1,117 @@
+---
+id: task_224_enterprise_industry_analysis_medium_medium039
+name: enterprise_industry_analysis-medium-medium039
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium039.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is there any difference between the minimum State Technological Invention Award value in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd. and the corresponding industry metric in Shanghai where Lang Ji Hui Ruan Technology Co., Ltd. is located?
+
+Output guidelines:
+The answer must be "No difference" or the other entity mentioned in the question. Output only the entity text without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No difference"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_225_enterprise_industry_analysis_medium_medium040.md b/tasks/task_225_enterprise_industry_analysis_medium_medium040.md
new file mode 100644
index 0000000000000000000000000000000000000000..5fd11779d631df8e4b762f23c2c85715438420c9
--- /dev/null
+++ b/tasks/task_225_enterprise_industry_analysis_medium_medium040.md
@@ -0,0 +1,117 @@
+---
+id: task_225_enterprise_industry_analysis_medium_medium040
+name: enterprise_industry_analysis-medium-medium040
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium040.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of HKEX-listed foreign-funded enterprises in the industry of Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd., or the number of SSE-listed enterprises in the corresponding industry of Shanghai where Lang Ji Hui Ruan Technology Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Bi Yuan Sheng Hua Jian She Kai Fa Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_226_enterprise_industry_analysis_medium_medium041.md b/tasks/task_226_enterprise_industry_analysis_medium_medium041.md
new file mode 100644
index 0000000000000000000000000000000000000000..93b84c52274d41773a75de57b815e6afa091af3f
--- /dev/null
+++ b/tasks/task_226_enterprise_industry_analysis_medium_medium041.md
@@ -0,0 +1,117 @@
+---
+id: task_226_enterprise_industry_analysis_medium_medium041
+name: enterprise_industry_analysis-medium-medium041
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium041.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of SSE-listed local state-owned enterprises in the industry of Hua Xin Yuan Shi New Materials Co., Ltd., or the number of BSE-listed enterprises in the corresponding industry of Guangdong where Zhong Ke Ke Shu Software Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Hua Xin Yuan Shi New Materials Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_227_enterprise_industry_analysis_medium_medium042.md b/tasks/task_227_enterprise_industry_analysis_medium_medium042.md
new file mode 100644
index 0000000000000000000000000000000000000000..78442d4554c6b2a1d678f4b87d5ea613981b0cb8
--- /dev/null
+++ b/tasks/task_227_enterprise_industry_analysis_medium_medium042.md
@@ -0,0 +1,117 @@
+---
+id: task_227_enterprise_industry_analysis_medium_medium042
+name: enterprise_industry_analysis-medium-medium042
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium042.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Comparing the minimum cumulative citation count of core patents in the industry of Huijin Jinrui Wealth Management Co., Ltd. with the same indicator in the corresponding industry of the province where the company is located, which value is larger?
+
+Output guidelines:
+The answer must be either "minimum cumulative citation count of core patents in the industry of Huijin Jinrui Wealth Management Co., Ltd." or "the same indicator in the corresponding industry of the province where the company is located". Output only the province or region name without any explanation, analysis, or descriptive text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_228_enterprise_industry_analysis_medium_medium043.md b/tasks/task_228_enterprise_industry_analysis_medium_medium043.md
new file mode 100644
index 0000000000000000000000000000000000000000..72a8943a5b9a71d7fe10f32362d166cdc20a4347
--- /dev/null
+++ b/tasks/task_228_enterprise_industry_analysis_medium_medium043.md
@@ -0,0 +1,117 @@
+---
+id: task_228_enterprise_industry_analysis_medium_medium043
+name: enterprise_industry_analysis-medium-medium043
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium043.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of SZSE-listed local state-owned enterprises in the industry of Huijin Jinrui Wealth Management Co., Ltd. and the number of SZSE-listed enterprises in the corresponding industry of the province where the company is located?
+
+Output guidelines:
+The answer must be a number with one decimal place. Output only the number without units, commas, or any explanatory text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`10.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_229_enterprise_industry_analysis_medium_medium044.md b/tasks/task_229_enterprise_industry_analysis_medium_medium044.md
new file mode 100644
index 0000000000000000000000000000000000000000..4105ea73722b14a7dea92359ef026ad3c5a0c58f
--- /dev/null
+++ b/tasks/task_229_enterprise_industry_analysis_medium_medium044.md
@@ -0,0 +1,117 @@
+---
+id: task_229_enterprise_industry_analysis_medium_medium044
+name: enterprise_industry_analysis-medium-medium044
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium044.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is higher: the number of SSE-listed enterprises in the corresponding industry of the province where Biyuan Shenghua Construction Development Co., Ltd. is located, or the number of HKEX-listed local state-owned enterprises in the industry of Huaying Taisheng Wealth Management Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Huaying Taisheng Wealth Management Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_230_enterprise_industry_analysis_medium_medium045.md b/tasks/task_230_enterprise_industry_analysis_medium_medium045.md
new file mode 100644
index 0000000000000000000000000000000000000000..4a295b8d8cecf79c1056a89b6c3683a052af6a9b
--- /dev/null
+++ b/tasks/task_230_enterprise_industry_analysis_medium_medium045.md
@@ -0,0 +1,117 @@
+---
+id: task_230_enterprise_industry_analysis_medium_medium045
+name: enterprise_industry_analysis-medium-medium045
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium045.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Comparing the number of HKEX-listed enterprises in the province where Huatu Wenjiao Online Education Co., Ltd. is located with the number of SZSE-listed enterprises in the industry of Yihai Changjin Business Co., Ltd., which value is larger?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yihai Changjin Business Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_231_enterprise_industry_analysis_medium_medium046.md b/tasks/task_231_enterprise_industry_analysis_medium_medium046.md
new file mode 100644
index 0000000000000000000000000000000000000000..40c669766cc572d09f5e7d17b9195b033fd67186
--- /dev/null
+++ b/tasks/task_231_enterprise_industry_analysis_medium_medium046.md
@@ -0,0 +1,117 @@
+---
+id: task_231_enterprise_industry_analysis_medium_medium046
+name: enterprise_industry_analysis-medium-medium046
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium046.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is higher: the median operating profit of the corresponding industry in Shanghai, where Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. is located, or the median operating profit of the industry of Yi Hai Chang Jin Shang Wu Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_232_enterprise_industry_analysis_medium_medium047.md b/tasks/task_232_enterprise_industry_analysis_medium_medium047.md
new file mode 100644
index 0000000000000000000000000000000000000000..b206b59f665b12e252cba45676153785773fd5b4
--- /dev/null
+++ b/tasks/task_232_enterprise_industry_analysis_medium_medium047.md
@@ -0,0 +1,117 @@
+---
+id: task_232_enterprise_industry_analysis_medium_medium047
+name: enterprise_industry_analysis-medium-medium047
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium047.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is higher: the maximum cumulative citations of all patents in the corresponding industry in Shanghai, where Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. is located, or the same metric in the industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_233_enterprise_industry_analysis_medium_medium048.md b/tasks/task_233_enterprise_industry_analysis_medium_medium048.md
new file mode 100644
index 0000000000000000000000000000000000000000..1eafb5cfde9c9c2146c10d53da612eb4dd45cb41
--- /dev/null
+++ b/tasks/task_233_enterprise_industry_analysis_medium_medium048.md
@@ -0,0 +1,117 @@
+---
+id: task_233_enterprise_industry_analysis_medium_medium048
+name: enterprise_industry_analysis-medium-medium048
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium048.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of HKEX-listed central state-owned enterprises in the corresponding industry in Shanghai where Jin Zhi Hong Sheng Zi Chan Guan Li Co., Ltd. is located, or the number of SSE-listed local state-owned enterprises in the industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_234_enterprise_industry_analysis_medium_medium049.md b/tasks/task_234_enterprise_industry_analysis_medium_medium049.md
new file mode 100644
index 0000000000000000000000000000000000000000..2f0aa5ede96c8ea3657e2a4782de03e925e3d89b
--- /dev/null
+++ b/tasks/task_234_enterprise_industry_analysis_medium_medium049.md
@@ -0,0 +1,117 @@
+---
+id: task_234_enterprise_industry_analysis_medium_medium049
+name: enterprise_industry_analysis-medium-medium049
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium049.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the total number of enterprises in the corresponding industry of Tianjin, where Bi Yuan Chan Jin Bu Dong Chan Co., Ltd. is located, and the number of SZSE-listed state-owned institute enterprises in the industry of Zhang Qiao Jin Chuang Technology Co., Ltd.?
+
+Output guidelines:
+The answer must be a number with one decimal place. Output only the number, without units, commas, or any text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`6.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_235_enterprise_industry_analysis_medium_medium050.md b/tasks/task_235_enterprise_industry_analysis_medium_medium050.md
new file mode 100644
index 0000000000000000000000000000000000000000..91ef54ddc71f0e0e5f5abe7db223d403dfca607b
--- /dev/null
+++ b/tasks/task_235_enterprise_industry_analysis_medium_medium050.md
@@ -0,0 +1,117 @@
+---
+id: task_235_enterprise_industry_analysis_medium_medium050
+name: enterprise_industry_analysis-medium-medium050
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium050.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of SZSE-listed local state-owned enterprises in the corresponding industry of Tianjin, where Bi Yuan Chan Jin Bu Dong Chan Co., Ltd. is located, and the number of HKEX-listed private enterprises in the industry of Zhang Qiao Jin Chuang Technology Co., Ltd.?
+
+Output guidelines:
+The answer must be a number with one decimal place. Output only the number, without units, commas, or any text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-43.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_236_enterprise_industry_analysis_medium_medium051.md b/tasks/task_236_enterprise_industry_analysis_medium_medium051.md
new file mode 100644
index 0000000000000000000000000000000000000000..3dae021f2c77e4dda3453ea033c40b7f611923d7
--- /dev/null
+++ b/tasks/task_236_enterprise_industry_analysis_medium_medium051.md
@@ -0,0 +1,117 @@
+---
+id: task_236_enterprise_industry_analysis_medium_medium051
+name: enterprise_industry_analysis-medium-medium051
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium051.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the gap between the number of SZSE-listed foreign-funded enterprises in the corresponding industry of Guangdong, where Gao Yin Ze Tong Pi Fa Co., Ltd. is located, and the number of SSE-listed enterprises in the industry of Lang Ji Hui Ruan Technology Co., Ltd.?
+
+Output guidelines:
+The answer must be a number with one decimal place. Output only the number, without units, commas, or any text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-139.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_237_enterprise_industry_analysis_medium_medium052.md b/tasks/task_237_enterprise_industry_analysis_medium_medium052.md
new file mode 100644
index 0000000000000000000000000000000000000000..582978a345bad47d5d83181b3c9e5ed467390f3c
--- /dev/null
+++ b/tasks/task_237_enterprise_industry_analysis_medium_medium052.md
@@ -0,0 +1,117 @@
+---
+id: task_237_enterprise_industry_analysis_medium_medium052
+name: enterprise_industry_analysis-medium-medium052
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium052.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Comparing the minimum value of State Natural Science Awards in the corresponding industry of the province where Gaoyin Zetong Wholesale Co., Ltd. is located with the minimum value of State Natural Science Awards in the industry of Langji Huiruan Technology Co., Ltd., which value is larger?
+
+Output guidelines:
+The answer must be a company name, "industry", or "Equal". Output only the name or "Equal" without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_238_enterprise_industry_analysis_medium_medium053.md b/tasks/task_238_enterprise_industry_analysis_medium_medium053.md
new file mode 100644
index 0000000000000000000000000000000000000000..34e6e775ddb70e23f09b89ef76adf78490b1355f
--- /dev/null
+++ b/tasks/task_238_enterprise_industry_analysis_medium_medium053.md
@@ -0,0 +1,117 @@
+---
+id: task_238_enterprise_industry_analysis_medium_medium053
+name: enterprise_industry_analysis-medium-medium053
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium053.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of HKEX-listed private enterprises in the corresponding industry of the province where Zhangqiao Jinchuang Technology Co., Ltd. is located and the number of HKEX-listed enterprises in the corresponding industry of the province where Zhaoye Huachang Real Estate Development Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-11.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_239_enterprise_industry_analysis_medium_medium054.md b/tasks/task_239_enterprise_industry_analysis_medium_medium054.md
new file mode 100644
index 0000000000000000000000000000000000000000..ec0b4979817def1326e678ec5a2ae210d1869fa8
--- /dev/null
+++ b/tasks/task_239_enterprise_industry_analysis_medium_medium054.md
@@ -0,0 +1,117 @@
+---
+id: task_239_enterprise_industry_analysis_medium_medium054
+name: enterprise_industry_analysis-medium-medium054
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium054.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is higher: the minimum R&D expenditure ratio in the corresponding industry of the province where Zhangqiao Jinchuang Technology Co., Ltd. is located, or the minimum R&D expenditure ratio in the corresponding industry of the province where Zhaoye Huachang Real Estate Development Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Zhangqiao Jinchuang Technology Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_240_enterprise_industry_analysis_medium_medium055.md b/tasks/task_240_enterprise_industry_analysis_medium_medium055.md
new file mode 100644
index 0000000000000000000000000000000000000000..4fc87d3f46129dcb22bd11a8227b75f76c7036fe
--- /dev/null
+++ b/tasks/task_240_enterprise_industry_analysis_medium_medium055.md
@@ -0,0 +1,117 @@
+---
+id: task_240_enterprise_industry_analysis_medium_medium055
+name: enterprise_industry_analysis-medium-medium055
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium055.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the median annual number of PCT invention patent applications for the corresponding industry in the province where Jinzhi Hongsheng Asset Management Company is located and that in the province where Zhonghai Gongchangjin Architectural Design Company is located, which is higher?
+
+Output guidelines:
+The answer must be either a company name or the word "industry". Output only the name, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Jinzhi Hongsheng Asset Management Company"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_241_enterprise_industry_analysis_medium_medium056.md b/tasks/task_241_enterprise_industry_analysis_medium_medium056.md
new file mode 100644
index 0000000000000000000000000000000000000000..9c3f097788cf277ebb5fb070359eea03924dce68
--- /dev/null
+++ b/tasks/task_241_enterprise_industry_analysis_medium_medium056.md
@@ -0,0 +1,117 @@
+---
+id: task_241_enterprise_industry_analysis_medium_medium056
+name: enterprise_industry_analysis-medium-medium056
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium056.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is the total liabilities (total) for the industry corresponding to the province where Jin Zhi Hong Sheng Zi Chan Management Co., Ltd. is located higher than the total liabilities (total) for the industry corresponding to the province where Zhong Hai Gong Chang Jin Jian Zhu She Ji Co., Ltd. is located?
+
+Output guidelines:
+The answer must be "Yes" or "No", output only one word, without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_242_enterprise_industry_analysis_medium_medium057.md b/tasks/task_242_enterprise_industry_analysis_medium_medium057.md
new file mode 100644
index 0000000000000000000000000000000000000000..d9cbf3b1ed30c820c25358d75d5a886971963fa5
--- /dev/null
+++ b/tasks/task_242_enterprise_industry_analysis_medium_medium057.md
@@ -0,0 +1,117 @@
+---
+id: task_242_enterprise_industry_analysis_medium_medium057
+name: enterprise_industry_analysis-medium-medium057
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium057.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is the maximum State Science and Technology Progress Award value in the corresponding industry of Shanghai, where Lang Ji Hui Ruan Technology Co., Ltd. is located, the same as the corresponding value in Beijing, where Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. is located?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_243_enterprise_industry_analysis_medium_medium058.md b/tasks/task_243_enterprise_industry_analysis_medium_medium058.md
new file mode 100644
index 0000000000000000000000000000000000000000..0b8d8e8c56b9f17f3c8e7160843e5756484053b5
--- /dev/null
+++ b/tasks/task_243_enterprise_industry_analysis_medium_medium058.md
@@ -0,0 +1,117 @@
+---
+id: task_243_enterprise_industry_analysis_medium_medium058
+name: enterprise_industry_analysis-medium-medium058
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium058.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+By how much is the mean capitalized R&D expenditure in the corresponding industry of Shanghai, where Lang Ji Hui Ruan Technology Co., Ltd. is located, higher than the corresponding value in Beijing, where Ai Jian Yi Kang Fu Zhong Xin Co., Ltd. is located?
+
+Output guidelines:
+The answer must be an exact number, preserving all significant decimal places. Output only the number without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`85678136.92375`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_244_enterprise_industry_analysis_medium_medium059.md b/tasks/task_244_enterprise_industry_analysis_medium_medium059.md
new file mode 100644
index 0000000000000000000000000000000000000000..a79b0dce5ff25bdb4c454fb990779c572dfd2646
--- /dev/null
+++ b/tasks/task_244_enterprise_industry_analysis_medium_medium059.md
@@ -0,0 +1,117 @@
+---
+id: task_244_enterprise_industry_analysis_medium_medium059
+name: enterprise_industry_analysis-medium-medium059
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium059.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the HKEX-listed private enterprise count in the corresponding industry of Guangdong, where Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. is located, or the total HKEX-listed enterprise count in the corresponding industry of Guangdong, where Zhao Ye Ze Jin Di Chan Holdings Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Zhao Ye Ze Jin Di Chan Holdings Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_245_enterprise_industry_analysis_medium_medium060.md b/tasks/task_245_enterprise_industry_analysis_medium_medium060.md
new file mode 100644
index 0000000000000000000000000000000000000000..3482ab81befbd7a3e251e78520f1989e7277e740
--- /dev/null
+++ b/tasks/task_245_enterprise_industry_analysis_medium_medium060.md
@@ -0,0 +1,117 @@
+---
+id: task_245_enterprise_industry_analysis_medium_medium060
+name: enterprise_industry_analysis-medium-medium060
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium060.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Are the minimum values of the Provincial or Ministerial Science and Technology Progress Award metric the same between the corresponding industries in Guangdong for Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. and Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_246_enterprise_industry_analysis_medium_medium061.md b/tasks/task_246_enterprise_industry_analysis_medium_medium061.md
new file mode 100644
index 0000000000000000000000000000000000000000..17061090396f3763722fabd8d1730413f912e30e
--- /dev/null
+++ b/tasks/task_246_enterprise_industry_analysis_medium_medium061.md
@@ -0,0 +1,117 @@
+---
+id: task_246_enterprise_industry_analysis_medium_medium061
+name: enterprise_industry_analysis-medium-medium061
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium061.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Are the total State Natural Science Award values the same between the corresponding industry in Hong Kong, where Rui Xing Jian Kang Zhi Yao Co., Ltd. is located, and the corresponding industry in Zhejiang, where Wu Li Hui Da Chain Co., Ltd. is located?
+
+Output guidelines:
+The answer must be "Yes" or "No". Output only one word without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_247_enterprise_industry_analysis_medium_medium062.md b/tasks/task_247_enterprise_industry_analysis_medium_medium062.md
new file mode 100644
index 0000000000000000000000000000000000000000..a63561e9510f7d688e94614627406e9e92f38c24
--- /dev/null
+++ b/tasks/task_247_enterprise_industry_analysis_medium_medium062.md
@@ -0,0 +1,117 @@
+---
+id: task_247_enterprise_industry_analysis_medium_medium062
+name: enterprise_industry_analysis-medium-medium062
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium062.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Rui Xing Jian Kang Zhi Yao Co., Ltd.industry in its province R&D headcount YoY change (minimum) and Wu Li Hui Da Chain Co., Ltd.industry in its province R&D headcount YoY change (minimum)compared with difference how much？
+
+Output guidelines:
+The answer must units, .Output only, without units, commas, or explanations.Wu, answer"No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-19.19`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_248_enterprise_industry_analysis_medium_medium063.md b/tasks/task_248_enterprise_industry_analysis_medium_medium063.md
new file mode 100644
index 0000000000000000000000000000000000000000..ee73699d0e8d08446aa3dd120d8a0f5d9262a031
--- /dev/null
+++ b/tasks/task_248_enterprise_industry_analysis_medium_medium063.md
@@ -0,0 +1,117 @@
+---
+id: task_248_enterprise_industry_analysis_medium_medium063
+name: enterprise_industry_analysis-medium-medium063
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium063.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of central state-owned enterprises listed on the Shenzhen Stock Exchange in the corresponding industry of the province where Baoxin Huihui Network Company is located and the number of central state-owned enterprises in the metal smelting and rolling industry listed on the Hong Kong Stock Exchange?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-2.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_249_enterprise_industry_analysis_medium_medium064.md b/tasks/task_249_enterprise_industry_analysis_medium_medium064.md
new file mode 100644
index 0000000000000000000000000000000000000000..3937ffa95b302ac0737b54b6f2a61b66975aa9b8
--- /dev/null
+++ b/tasks/task_249_enterprise_industry_analysis_medium_medium064.md
@@ -0,0 +1,117 @@
+---
+id: task_249_enterprise_industry_analysis_medium_medium064
+name: enterprise_industry_analysis-medium-medium064
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium064.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the median R&D expenditure ratio of the corresponding industry in the province where Baoxin Huihui Network Company is located and that of the metal smelting and rolling processing industry, which is higher?
+
+Output guidelines:
+The answer must be either a company name or the word "industry". Output only the name, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Baoxin Huihui Network Company"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_250_enterprise_industry_analysis_medium_medium065.md b/tasks/task_250_enterprise_industry_analysis_medium_medium065.md
new file mode 100644
index 0000000000000000000000000000000000000000..925d6740c5cb53bb0ea80aa789e43aa1451e9903
--- /dev/null
+++ b/tasks/task_250_enterprise_industry_analysis_medium_medium065.md
@@ -0,0 +1,117 @@
+---
+id: task_250_enterprise_industry_analysis_medium_medium065
+name: enterprise_industry_analysis-medium-medium065
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium065.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of Shanghai Stock Exchange-listed private enterprises in the corresponding industry of the province where Beikong Zejing Water Company is located and the number of Beijing Stock Exchange-listed enterprises in the Information Transmission, Software and IT Services industry, which is larger?
+
+Output guidelines:
+The answer must be either "the number of SSE-listed private enterprises in the corresponding industry of the province where Beikong Zejing Water Company is located" or "the number of BSE-listed enterprises in the Information Transmission, Software and IT Services industry". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"the number of BSE-listed enterprises in the Information Transmission, Software and IT Services industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_251_enterprise_industry_analysis_medium_medium066.md b/tasks/task_251_enterprise_industry_analysis_medium_medium066.md
new file mode 100644
index 0000000000000000000000000000000000000000..39be46e3c2b59794e51af2431e505736c6e3e60e
--- /dev/null
+++ b/tasks/task_251_enterprise_industry_analysis_medium_medium066.md
@@ -0,0 +1,117 @@
+---
+id: task_251_enterprise_industry_analysis_medium_medium066
+name: enterprise_industry_analysis-medium-medium066
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium066.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the total assets of the corresponding industry in the province where Beikong Zejing Water Company is located, or the total assets of the Information Transmission, Software and IT Services industry?
+
+Output guidelines:
+The answer must be either "the total assets of the corresponding industry in the province where Beikong Zejing Water Company is located" or "the Information Transmission, Software and IT Services industry". Output only the name, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"the Information Transmission, Software and IT Services industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_252_enterprise_industry_analysis_medium_medium067.md b/tasks/task_252_enterprise_industry_analysis_medium_medium067.md
new file mode 100644
index 0000000000000000000000000000000000000000..f97834fb95ecce611c17c5400f8e503b0d4cf2aa
--- /dev/null
+++ b/tasks/task_252_enterprise_industry_analysis_medium_medium067.md
@@ -0,0 +1,117 @@
+---
+id: task_252_enterprise_industry_analysis_medium_medium067
+name: enterprise_industry_analysis-medium-medium067
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium067.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of HKEX-listed foreign-funded enterprises in the corresponding industry of the province where Aijian Yikang Fuzhongxin Company is located and the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry, which value is larger?
+
+Output guidelines:
+The answer must be either "the number of HKEX-listed foreign-funded enterprises in the corresponding industry of the province where Aijian Yikang Fuzhongxin Company is located" or "the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"the number of SSE-listed central state-owned enterprises in the pharmaceutical manufacturing industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_253_enterprise_industry_analysis_medium_medium068.md b/tasks/task_253_enterprise_industry_analysis_medium_medium068.md
new file mode 100644
index 0000000000000000000000000000000000000000..665af19dfeb823b389de7e2623923a8f64fbc210
--- /dev/null
+++ b/tasks/task_253_enterprise_industry_analysis_medium_medium068.md
@@ -0,0 +1,117 @@
+---
+id: task_253_enterprise_industry_analysis_medium_medium068
+name: enterprise_industry_analysis-medium-medium068
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium068.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the maximum annual number of China patent applications for the corresponding industry in the province where Aijian Yikang Fuzhongxin Company is located and the maximum annual number of China patent applications for the pharmaceutical manufacturing industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-329.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_254_enterprise_industry_analysis_medium_medium069.md b/tasks/task_254_enterprise_industry_analysis_medium_medium069.md
new file mode 100644
index 0000000000000000000000000000000000000000..2d537796c8aba7b437b5576723d6b4de7be03630
--- /dev/null
+++ b/tasks/task_254_enterprise_industry_analysis_medium_medium069.md
@@ -0,0 +1,117 @@
+---
+id: task_254_enterprise_industry_analysis_medium_medium069
+name: enterprise_industry_analysis-medium-medium069
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium069.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the maximum capitalized R&D expenditure of the corresponding industry in the province where Zhongche Yuanze Shipbuilding Company is located and the maximum capitalized R&D expenditure of the Electricity, Heat, Gas and Water Production and Supply industry, which is higher?
+
+Output guidelines:
+The answer must be either the company name or "industry". Output only the name, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_255_enterprise_industry_analysis_medium_medium070.md b/tasks/task_255_enterprise_industry_analysis_medium_medium070.md
new file mode 100644
index 0000000000000000000000000000000000000000..e54f2d98ad2ac95334e9d6e553a11286b441c701
--- /dev/null
+++ b/tasks/task_255_enterprise_industry_analysis_medium_medium070.md
@@ -0,0 +1,117 @@
+---
+id: task_255_enterprise_industry_analysis_medium_medium070
+name: enterprise_industry_analysis-medium-medium070
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium070.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of local state-owned enterprises listed on the Shanghai Stock Exchange in the corresponding industry in Beijing Municipality and the number of private enterprises listed on the Shanghai Stock Exchange in the Electricity, Heat, Gas and Water Production and Supply industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-18.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_256_enterprise_industry_analysis_medium_medium071.md b/tasks/task_256_enterprise_industry_analysis_medium_medium071.md
new file mode 100644
index 0000000000000000000000000000000000000000..ecdb84217d9b6c45d35d1eee2e6322396d4e9616
--- /dev/null
+++ b/tasks/task_256_enterprise_industry_analysis_medium_medium071.md
@@ -0,0 +1,117 @@
+---
+id: task_256_enterprise_industry_analysis_medium_medium071
+name: enterprise_industry_analysis-medium-medium071
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium071.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of SSE-listed enterprises in the province where Biyuan Chanjin Real Estate Company is located, or the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry?
+
+Output guidelines:
+The answer must be either "the number of SSE-listed enterprises in the province where Biyuan Chanjin Real Estate Company is located" or "the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"the number of SZSE-listed local state-owned enterprises in the Transportation, Storage and Postal Services industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_257_enterprise_industry_analysis_medium_medium072.md b/tasks/task_257_enterprise_industry_analysis_medium_medium072.md
new file mode 100644
index 0000000000000000000000000000000000000000..286ee530c99a8d902b671a0a60fb030417920d7c
--- /dev/null
+++ b/tasks/task_257_enterprise_industry_analysis_medium_medium072.md
@@ -0,0 +1,117 @@
+---
+id: task_257_enterprise_industry_analysis_medium_medium072
+name: enterprise_industry_analysis-medium-medium072
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium072.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which value is larger: the total number of real estate enterprises in the province where Biyuan Chanjin Real Estate Company is located, or the number of SZSE-listed central state-owned enterprises in the Transportation, Storage and Postal Services industry?
+
+Output guidelines:
+The answer must be either "the total number of real estate enterprises in the province where Biyuan Chanjin Real Estate Company is located" or "the number of SZSE-listed central state-owned enterprises in the Transportation, Storage and Postal Services industry". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"the total number of real estate enterprises in the province where Biyuan Chanjin Real Estate Company is located"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_258_enterprise_industry_analysis_medium_medium073.md b/tasks/task_258_enterprise_industry_analysis_medium_medium073.md
new file mode 100644
index 0000000000000000000000000000000000000000..2f802abb1bf10da0bb3b246ed6bbc754adb23c45
--- /dev/null
+++ b/tasks/task_258_enterprise_industry_analysis_medium_medium073.md
@@ -0,0 +1,117 @@
+---
+id: task_258_enterprise_industry_analysis_medium_medium073
+name: enterprise_industry_analysis-medium-medium073
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium073.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of SZSE-listed foreign-funded enterprises in the industry where Zhongke Keshu Software Company operates and the number of BSE-listed private enterprises in the construction industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`2.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_259_enterprise_industry_analysis_medium_medium074.md b/tasks/task_259_enterprise_industry_analysis_medium_medium074.md
new file mode 100644
index 0000000000000000000000000000000000000000..b89020d60dca8016817bea93b9e554f4f00336f3
--- /dev/null
+++ b/tasks/task_259_enterprise_industry_analysis_medium_medium074.md
@@ -0,0 +1,117 @@
+---
+id: task_259_enterprise_industry_analysis_medium_medium074
+name: enterprise_industry_analysis-medium-medium074
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium074.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of HKEX-listed local state-owned enterprises in the industry where Zhongke Keshu Software Company operates and the number of HKEX-listed foreign-funded enterprises in the construction industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`3.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_260_enterprise_industry_analysis_medium_medium075.md b/tasks/task_260_enterprise_industry_analysis_medium_medium075.md
new file mode 100644
index 0000000000000000000000000000000000000000..5571e02c5ecc436726e234f878aa068d5cfadb32
--- /dev/null
+++ b/tasks/task_260_enterprise_industry_analysis_medium_medium075.md
@@ -0,0 +1,117 @@
+---
+id: task_260_enterprise_industry_analysis_medium_medium075
+name: enterprise_industry_analysis-medium-medium075
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium075.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of SSE-listed foreign-funded enterprises in the industry where Hengli Kezhi Software Company operates and the number of SSE-listed private enterprises in the conglomerates industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`2.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_261_enterprise_industry_analysis_medium_medium076.md b/tasks/task_261_enterprise_industry_analysis_medium_medium076.md
new file mode 100644
index 0000000000000000000000000000000000000000..07cb7412f3792c773b681f70c22750e065906f24
--- /dev/null
+++ b/tasks/task_261_enterprise_industry_analysis_medium_medium076.md
@@ -0,0 +1,117 @@
+---
+id: task_261_enterprise_industry_analysis_medium_medium076
+name: enterprise_industry_analysis-medium-medium076
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium076.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the median number of industry standards participated in drafting in the industry where Hengli Kezhi Software Company operates and that of the conglomerates industry, which value is larger?
+
+Output guidelines:
+The answer must be either the company name or "industry". Output only the name, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Hengli Kezhi Software Company"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_262_enterprise_industry_analysis_medium_medium077.md b/tasks/task_262_enterprise_industry_analysis_medium_medium077.md
new file mode 100644
index 0000000000000000000000000000000000000000..e80e7958d4b57f723311355bbc8f9869cff61d55
--- /dev/null
+++ b/tasks/task_262_enterprise_industry_analysis_medium_medium077.md
@@ -0,0 +1,117 @@
+---
+id: task_262_enterprise_industry_analysis_medium_medium077
+name: enterprise_industry_analysis-medium-medium077
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium077.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of BSE-listed enterprises in the industry of Lang Ji Hui Ruan Technology Co., Ltd., or the number of enterprises in chemical raw materials and chemical products manufacturing?
+
+Output guidelines:
+The answer must be either "Number of BSE-listed enterprises in the industry of Lang Ji Hui Ruan Technology Co., Ltd." or "Number of enterprises in chemical raw materials and chemical products manufacturing"; output only the designated answer text, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Chemical Raw Materials and Chemical Products Manufacturing"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_263_enterprise_industry_analysis_medium_medium078.md b/tasks/task_263_enterprise_industry_analysis_medium_medium078.md
new file mode 100644
index 0000000000000000000000000000000000000000..8dafb9d114a3796e0328d0a93451157fbc44183b
--- /dev/null
+++ b/tasks/task_263_enterprise_industry_analysis_medium_medium078.md
@@ -0,0 +1,117 @@
+---
+id: task_263_enterprise_industry_analysis_medium_medium078
+name: enterprise_industry_analysis-medium-medium078
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium078.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the mean change in R&D expenditure ratio for the industry where Langji Huiruan Technology Company operates and the same indicator for the Chemical Raw Materials and Chemical Products Manufacturing industry, which is higher?
+
+Output guidelines:
+The answer must be either the company name or "industry". Output only the name, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Langji Huiruan Technology Company"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_264_enterprise_industry_analysis_medium_medium079.md b/tasks/task_264_enterprise_industry_analysis_medium_medium079.md
new file mode 100644
index 0000000000000000000000000000000000000000..02fc958fe9b09341e12d2d3a1507b0301407b399
--- /dev/null
+++ b/tasks/task_264_enterprise_industry_analysis_medium_medium079.md
@@ -0,0 +1,117 @@
+---
+id: task_264_enterprise_industry_analysis_medium_medium079
+name: enterprise_industry_analysis-medium-medium079
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium079.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of SZSE-listed central state-owned enterprises in the industry where Huijin Jinrui Wealth Management Company operates and the number of HKEX-listed central state-owned enterprises in the Electrical Machinery and Equipment Manufacturing industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`2.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_265_enterprise_industry_analysis_medium_medium080.md b/tasks/task_265_enterprise_industry_analysis_medium_medium080.md
new file mode 100644
index 0000000000000000000000000000000000000000..47590191f0d4c789736554da0cd24f80d03b06ca
--- /dev/null
+++ b/tasks/task_265_enterprise_industry_analysis_medium_medium080.md
@@ -0,0 +1,117 @@
+---
+id: task_265_enterprise_industry_analysis_medium_medium080
+name: enterprise_industry_analysis-medium-medium080
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium080.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the median net profit amount of Huijin Jinrui Wealth Management Company and that of the Electrical Machinery and Equipment Manufacturing industry, which is higher?
+
+Output guidelines:
+The answer must be either the company name or "industry". Output only the name, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Huijin Jinrui Wealth Management Company"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_266_enterprise_industry_analysis_medium_medium081.md b/tasks/task_266_enterprise_industry_analysis_medium_medium081.md
new file mode 100644
index 0000000000000000000000000000000000000000..02f78694c2233b8211bd9b2b525b3e3d3e6aa403
--- /dev/null
+++ b/tasks/task_266_enterprise_industry_analysis_medium_medium081.md
@@ -0,0 +1,117 @@
+---
+id: task_266_enterprise_industry_analysis_medium_medium081
+name: enterprise_industry_analysis-medium-medium081
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium081.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of SZSE-listed enterprises in the industry where Zhaoye Huachang Real Estate Development Company operates and the number of SZSE-listed private enterprises in the General Equipment Manufacturing industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-35.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_267_enterprise_industry_analysis_medium_medium082.md b/tasks/task_267_enterprise_industry_analysis_medium_medium082.md
new file mode 100644
index 0000000000000000000000000000000000000000..4552ae9eebfcda4b512ac3b76def811ae6e2e236
--- /dev/null
+++ b/tasks/task_267_enterprise_industry_analysis_medium_medium082.md
@@ -0,0 +1,117 @@
+---
+id: task_267_enterprise_industry_analysis_medium_medium082
+name: enterprise_industry_analysis-medium-medium082
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium082.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the maximum operating revenue amount of the industry where Zhaoye Huachang Real Estate Development Company operates and that of the General Equipment Manufacturing industry, which is higher?
+
+Output guidelines:
+The answer must be either the company name or "industry". Output only the name, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Zhaoye Huachang Real Estate Development Company"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_268_enterprise_industry_analysis_medium_medium083.md b/tasks/task_268_enterprise_industry_analysis_medium_medium083.md
new file mode 100644
index 0000000000000000000000000000000000000000..3ef63f5476989cee7632710dbcbe0d7194f29744
--- /dev/null
+++ b/tasks/task_268_enterprise_industry_analysis_medium_medium083.md
@@ -0,0 +1,117 @@
+---
+id: task_268_enterprise_industry_analysis_medium_medium083
+name: enterprise_industry_analysis-medium-medium083
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium083.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of SZSE-listed enterprises in the industry where Yihai Changjin Business Company operates and the number of SZSE-listed central state-owned enterprises in the Communication Transmission Equipment industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`41.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_269_enterprise_industry_analysis_medium_medium084.md b/tasks/task_269_enterprise_industry_analysis_medium_medium084.md
new file mode 100644
index 0000000000000000000000000000000000000000..451c02c2703e6b9c6adb9ce17d3212d397490b70
--- /dev/null
+++ b/tasks/task_269_enterprise_industry_analysis_medium_medium084.md
@@ -0,0 +1,117 @@
+---
+id: task_269_enterprise_industry_analysis_medium_medium084
+name: enterprise_industry_analysis-medium-medium084
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium084.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of SSE-listed central state-owned enterprises in the industry where Yihai Changjin Business Company operates, or the total number of enterprises in the Communication Transmission Equipment industry?
+
+Output guidelines:
+The answer must be either "the number of SSE-listed central state-owned enterprises in the industry where Yihai Changjin Business Company operates" or "the total number of enterprises in the Communication Transmission Equipment industry". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"the total number of enterprises in the Communication Transmission Equipment industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_270_enterprise_industry_analysis_medium_medium085.md b/tasks/task_270_enterprise_industry_analysis_medium_medium085.md
new file mode 100644
index 0000000000000000000000000000000000000000..f6894f47516f5bcdd8edbb0ea55c876db630a24a
--- /dev/null
+++ b/tasks/task_270_enterprise_industry_analysis_medium_medium085.md
@@ -0,0 +1,117 @@
+---
+id: task_270_enterprise_industry_analysis_medium_medium085
+name: enterprise_industry_analysis-medium-medium085
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium085.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of HKEX-listed local state-owned enterprises in the industry where Zhongke Zhiyun Data Services Company operates, or the number of BSE-listed enterprises in the Transportation, Storage and Postal Services industry?
+
+Output guidelines:
+The answer must be either "Zhongke Zhiyun Data Services Company" or "the number of BSE-listed enterprises in the Transportation, Storage and Postal Services industry". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Zhongke Zhiyun Data Services Company"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_271_enterprise_industry_analysis_medium_medium086.md b/tasks/task_271_enterprise_industry_analysis_medium_medium086.md
new file mode 100644
index 0000000000000000000000000000000000000000..1ef2715fb8853b5f78b66b2166de524cd0a4eadd
--- /dev/null
+++ b/tasks/task_271_enterprise_industry_analysis_medium_medium086.md
@@ -0,0 +1,117 @@
+---
+id: task_271_enterprise_industry_analysis_medium_medium086
+name: enterprise_industry_analysis-medium-medium086
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium086.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of SSE-listed central state-owned enterprises in the industry where Zhongke Zhiyun Data Services Company operates, or the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry?
+
+Output guidelines:
+The answer must be either "the number of SSE-listed central state-owned enterprises in the industry where Zhongke Zhiyun Data Services Company operates" or "the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"the number of SSE-listed enterprises in the Transportation, Storage and Postal Services industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_272_enterprise_industry_analysis_medium_medium087.md b/tasks/task_272_enterprise_industry_analysis_medium_medium087.md
new file mode 100644
index 0000000000000000000000000000000000000000..6266c3507d36b16cdafcd7536eead31e80f1b828
--- /dev/null
+++ b/tasks/task_272_enterprise_industry_analysis_medium_medium087.md
@@ -0,0 +1,117 @@
+---
+id: task_272_enterprise_industry_analysis_medium_medium087
+name: enterprise_industry_analysis-medium-medium087
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium087.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of SZSE-listed central state-owned enterprises in the industry where Wuli Huida Chain Company operates, or the number of HKEX-listed foreign-funded enterprises in the Communication Transmission Equipment industry?
+
+Output guidelines:
+The answer must be either the company name or "industry". Output only the name, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Wuli Huida Chain Company"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_273_enterprise_industry_analysis_medium_medium088.md b/tasks/task_273_enterprise_industry_analysis_medium_medium088.md
new file mode 100644
index 0000000000000000000000000000000000000000..26c0596d78c93e49e85ad5015b3893c0cab48703
--- /dev/null
+++ b/tasks/task_273_enterprise_industry_analysis_medium_medium088.md
@@ -0,0 +1,117 @@
+---
+id: task_273_enterprise_industry_analysis_medium_medium088
+name: enterprise_industry_analysis-medium-medium088
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium088.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is higher: the total capitalized R&D expenditure of the industry where Wuli Huida Chain Company operates, or that of the Communication Transmission Equipment industry?
+
+Output guidelines:
+The answer must be either "the total capitalized R&D expenditure of the industry where Wuli Huida Chain Company operates" or "the Communication Transmission Equipment industry". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"the Communication Transmission Equipment industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_274_enterprise_industry_analysis_medium_medium089.md b/tasks/task_274_enterprise_industry_analysis_medium_medium089.md
new file mode 100644
index 0000000000000000000000000000000000000000..7db84a596d30b575420618bf94077e863af2e8ac
--- /dev/null
+++ b/tasks/task_274_enterprise_industry_analysis_medium_medium089.md
@@ -0,0 +1,117 @@
+---
+id: task_274_enterprise_industry_analysis_medium_medium089
+name: enterprise_industry_analysis-medium-medium089
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium089.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the minimum cumulative citation count of core patents in the industry where Huaxin Yuanshi New Materials Company operates and the same metric in the Scientific Research and Technical Services industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_275_enterprise_industry_analysis_medium_medium090.md b/tasks/task_275_enterprise_industry_analysis_medium_medium090.md
new file mode 100644
index 0000000000000000000000000000000000000000..c7eb181109dd7a4c4a64519c50f516bf1b1f61c4
--- /dev/null
+++ b/tasks/task_275_enterprise_industry_analysis_medium_medium090.md
@@ -0,0 +1,117 @@
+---
+id: task_275_enterprise_industry_analysis_medium_medium090
+name: enterprise_industry_analysis-medium-medium090
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium090.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Compared with the same indicator in the Scientific Research and Technical Services industry, what is the difference in the mean cumulative number of PCT invention patent applications for the industry where Huaxin Yuanshi New Materials Company operates?
+
+Output guidelines:
+The answer must be an exact number, preserving all meaningful decimal places. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-21.9787234042554`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_276_enterprise_industry_analysis_medium_medium091.md b/tasks/task_276_enterprise_industry_analysis_medium_medium091.md
new file mode 100644
index 0000000000000000000000000000000000000000..dcdafc1e50827c48aff7c0667378dfa7c5a5ba6d
--- /dev/null
+++ b/tasks/task_276_enterprise_industry_analysis_medium_medium091.md
@@ -0,0 +1,117 @@
+---
+id: task_276_enterprise_industry_analysis_medium_medium091
+name: enterprise_industry_analysis-medium-medium091
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium091.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the maximum number of international industry awards in the industry where Aijian Yikang Fuzhongxin Company operates and the maximum number of international industry awards in the Chemical Fiber Manufacturing industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_277_enterprise_industry_analysis_medium_medium092.md b/tasks/task_277_enterprise_industry_analysis_medium_medium092.md
new file mode 100644
index 0000000000000000000000000000000000000000..9d8732ad19a36ee6f1e8742923e6d512f21e3dd0
--- /dev/null
+++ b/tasks/task_277_enterprise_industry_analysis_medium_medium092.md
@@ -0,0 +1,117 @@
+---
+id: task_277_enterprise_industry_analysis_medium_medium092
+name: enterprise_industry_analysis-medium-medium092
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium092.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the median year-on-year change in operating profit for the industry where Aijian Yikang Fuzhongxin Company operates and that of the Chemical Fiber Manufacturing industry?
+
+Output guidelines:
+The answer must be a single number with three decimal places. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`16.675`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_278_enterprise_industry_analysis_medium_medium093.md b/tasks/task_278_enterprise_industry_analysis_medium_medium093.md
new file mode 100644
index 0000000000000000000000000000000000000000..a7ba5f57fc0e8aec687e264848543a95a2e13449
--- /dev/null
+++ b/tasks/task_278_enterprise_industry_analysis_medium_medium093.md
@@ -0,0 +1,117 @@
+---
+id: task_278_enterprise_industry_analysis_medium_medium093
+name: enterprise_industry_analysis-medium-medium093
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium093.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of SSE-listed private enterprises in the industry where Zhongke Keshu Software Company operates and the number of SZSE-listed enterprises in Other Manufacturing in China?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`71.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_279_enterprise_industry_analysis_medium_medium094.md b/tasks/task_279_enterprise_industry_analysis_medium_medium094.md
new file mode 100644
index 0000000000000000000000000000000000000000..01c944a71a6a89016cc22f5c8ac5a5c997c3a7fa
--- /dev/null
+++ b/tasks/task_279_enterprise_industry_analysis_medium_medium094.md
@@ -0,0 +1,117 @@
+---
+id: task_279_enterprise_industry_analysis_medium_medium094
+name: enterprise_industry_analysis-medium-medium094
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium094.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of SZSE-listed private enterprises in the industry where Yihai Changjin Business Company operates and that in China's Transportation, Storage and Postal Services industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`14.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_280_enterprise_industry_analysis_medium_medium095.md b/tasks/task_280_enterprise_industry_analysis_medium_medium095.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b808e8b4930833d4b5f615a8f14070231ba8c9d
--- /dev/null
+++ b/tasks/task_280_enterprise_industry_analysis_medium_medium095.md
@@ -0,0 +1,117 @@
+---
+id: task_280_enterprise_industry_analysis_medium_medium095
+name: enterprise_industry_analysis-medium-medium095
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium095.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of SSE-listed private enterprises in the industry where Yihai Changjin Business Company operates, or the number of BSE-listed private enterprises in China's Transportation, Storage and Postal Services industry?
+
+Output guidelines:
+The answer must be either the company name or "industry". Output only the name, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yihai Changjin Business Company"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_281_enterprise_industry_analysis_medium_medium096.md b/tasks/task_281_enterprise_industry_analysis_medium_medium096.md
new file mode 100644
index 0000000000000000000000000000000000000000..3494332499497728ed164c18a7317507a5033744
--- /dev/null
+++ b/tasks/task_281_enterprise_industry_analysis_medium_medium096.md
@@ -0,0 +1,117 @@
+---
+id: task_281_enterprise_industry_analysis_medium_medium096
+name: enterprise_industry_analysis-medium-medium096
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium096.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of HKEX-listed central state-owned enterprises in the industry where Zhongche Yuanze Shipbuilding Company operates, or the number of HKEX-listed private enterprises in China's Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing industry?
+
+Output guidelines:
+The answer must be either the company name or "industry". Output only the name, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_282_enterprise_industry_analysis_medium_medium097.md b/tasks/task_282_enterprise_industry_analysis_medium_medium097.md
new file mode 100644
index 0000000000000000000000000000000000000000..2ddb26ecea1e91a520b7b8948be2ca87a599933b
--- /dev/null
+++ b/tasks/task_282_enterprise_industry_analysis_medium_medium097.md
@@ -0,0 +1,117 @@
+---
+id: task_282_enterprise_industry_analysis_medium_medium097
+name: enterprise_industry_analysis-medium-medium097
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium097.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of HKEX-listed private enterprises in the industry where Zhongche Yuanze Shipbuilding Company operates, or the number of SSE-listed private enterprises in China's Cultural, Educational, Industrial Aesthetics, Sports and Entertainment Goods Manufacturing industry?
+
+Output guidelines:
+The answer must be either "the number of HKEX-listed private enterprises in the industry where Zhongche Yuanze Shipbuilding Company operates", "China", or "Equal". Output only the answer itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_283_enterprise_industry_analysis_medium_medium098.md b/tasks/task_283_enterprise_industry_analysis_medium_medium098.md
new file mode 100644
index 0000000000000000000000000000000000000000..01fcfa9fadc4507e28442345119d400198b73d7c
--- /dev/null
+++ b/tasks/task_283_enterprise_industry_analysis_medium_medium098.md
@@ -0,0 +1,117 @@
+---
+id: task_283_enterprise_industry_analysis_medium_medium098
+name: enterprise_industry_analysis-medium-medium098
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium098.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of SZSE-listed enterprises in the industry where Yihai Changjin Business Company operates, or the number of SZSE-listed private enterprises in China's Conglomerates industry?
+
+Output guidelines:
+The answer must be either the company name or "industry". Output only the name, without any explanation or description. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yihai Changjin Business Company"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_284_enterprise_industry_analysis_medium_medium099.md b/tasks/task_284_enterprise_industry_analysis_medium_medium099.md
new file mode 100644
index 0000000000000000000000000000000000000000..ab0c7c6d44a1c4be9707b93f2dcd2b7756a0f4f6
--- /dev/null
+++ b/tasks/task_284_enterprise_industry_analysis_medium_medium099.md
@@ -0,0 +1,117 @@
+---
+id: task_284_enterprise_industry_analysis_medium_medium099
+name: enterprise_industry_analysis-medium-medium099
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium099.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Yi Hai Chang Jin Shang Wu Co., Ltd.industry's Number of SZSE-listed enterprises and China ConglomeratesindustryPrivate enterpriseShenzhen Stock Exchange countcompared with which unitsgreater？
+
+Output guidelines:
+The answer must a company name or "industry", Output onlyname, without any explanation or description.Wu, answer"No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yi Hai Chang Jin Shang Wu Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_285_enterprise_industry_analysis_medium_medium100.md b/tasks/task_285_enterprise_industry_analysis_medium_medium100.md
new file mode 100644
index 0000000000000000000000000000000000000000..f2f7334f255657ac38594bf52484fa802e09be5f
--- /dev/null
+++ b/tasks/task_285_enterprise_industry_analysis_medium_medium100.md
@@ -0,0 +1,117 @@
+---
+id: task_285_enterprise_industry_analysis_medium_medium100
+name: enterprise_industry_analysis-medium-medium100
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium100.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the median year-on-year change in R&D expenditure for the industry where Biyuan Zhize Urban Development Company operates and that of the Pharmaceutical Manufacturing industry in China?
+
+Output guidelines:
+The answer must be a single number with two decimal places. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-9.87`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_286_enterprise_industry_analysis_medium_medium101.md b/tasks/task_286_enterprise_industry_analysis_medium_medium101.md
new file mode 100644
index 0000000000000000000000000000000000000000..1ec27571700070dda50ec5a461b39d87783ecf42
--- /dev/null
+++ b/tasks/task_286_enterprise_industry_analysis_medium_medium101.md
@@ -0,0 +1,117 @@
+---
+id: task_286_enterprise_industry_analysis_medium_medium101
+name: enterprise_industry_analysis-medium-medium101
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium101.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of SSE-listed foreign-funded enterprises in the industry corresponding to Biyuan Zhize Urban Development Company and the number of HKEX-listed state-owned research institute enterprises in China's Pharmaceutical Manufacturing industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`5.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_287_enterprise_industry_analysis_medium_medium102.md b/tasks/task_287_enterprise_industry_analysis_medium_medium102.md
new file mode 100644
index 0000000000000000000000000000000000000000..658e7ac0823b83e758fb90a75c8dd504d06efce5
--- /dev/null
+++ b/tasks/task_287_enterprise_industry_analysis_medium_medium102.md
@@ -0,0 +1,117 @@
+---
+id: task_287_enterprise_industry_analysis_medium_medium102
+name: enterprise_industry_analysis-medium-medium102
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium102.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the total number of Consumer Electronics and Electrical industry enterprises in the province where Changqiao Jinchuang Technology Company is located, or the number of HKEX-listed foreign-funded enterprises in China's Conglomerates industry?
+
+Output guidelines:
+The answer must be either "the total number of Consumer Electronics and Electrical industry enterprises in the province where Changqiao Jinchuang Technology Company is located" or "the number of HKEX-listed foreign-funded enterprises in China's Conglomerates industry". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"the number of HKEX-listed foreign-funded enterprises in China's Conglomerates industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_288_enterprise_industry_analysis_medium_medium103.md b/tasks/task_288_enterprise_industry_analysis_medium_medium103.md
new file mode 100644
index 0000000000000000000000000000000000000000..c799b213eb5a933f9cd835824d9e86e9594ebf89
--- /dev/null
+++ b/tasks/task_288_enterprise_industry_analysis_medium_medium103.md
@@ -0,0 +1,117 @@
+---
+id: task_288_enterprise_industry_analysis_medium_medium103
+name: enterprise_industry_analysis-medium-medium103
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium103.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the total number of Consumer Electronics and Electrical industry enterprises in the province where Changqiao Jinchuang Technology Company is located, or the number of HKEX-listed private enterprises in China's Conglomerates industry?
+
+Output guidelines:
+The answer must be either "the total number of Consumer Electronics and Electrical industry enterprises in the province where Changqiao Jinchuang Technology Company is located" or "the number of HKEX-listed private enterprises in China's Conglomerates industry". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"the number of HKEX-listed private enterprises in China's Conglomerates industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_289_enterprise_industry_analysis_medium_medium104.md b/tasks/task_289_enterprise_industry_analysis_medium_medium104.md
new file mode 100644
index 0000000000000000000000000000000000000000..0ea3a8a5a9ad4448792797cc1c2bb39c5ba516c3
--- /dev/null
+++ b/tasks/task_289_enterprise_industry_analysis_medium_medium104.md
@@ -0,0 +1,117 @@
+---
+id: task_289_enterprise_industry_analysis_medium_medium104
+name: enterprise_industry_analysis-medium-medium104
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium104.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which value is larger: the number of HKEX-listed foreign-funded enterprises in the province where Jinzhi Hongsheng Asset Management Company is located, or the number of SSE-listed central state-owned enterprises in China's Culture, Sports and Entertainment industry?
+
+Output guidelines:
+The answer must be either "the number of HKEX-listed foreign-funded enterprises in the province where Jinzhi Hongsheng Asset Management Company is located", "the number of SSE-listed central state-owned enterprises in China's Culture, Sports and Entertainment industry", or "Equal". Output only the answer itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_290_enterprise_industry_analysis_medium_medium105.md b/tasks/task_290_enterprise_industry_analysis_medium_medium105.md
new file mode 100644
index 0000000000000000000000000000000000000000..5cb541bc85bc9643b238b9b2d5108e08b469c3a4
--- /dev/null
+++ b/tasks/task_290_enterprise_industry_analysis_medium_medium105.md
@@ -0,0 +1,117 @@
+---
+id: task_290_enterprise_industry_analysis_medium_medium105
+name: enterprise_industry_analysis-medium-medium105
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium105.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the mean number of provincial or ministerial Science and Technology Progress Awards in the province where Jinzhi Hongsheng Asset Management Company is located and the mean number of the same awards in China's Culture, Sports and Entertainment industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`3.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_291_enterprise_industry_analysis_medium_medium106.md b/tasks/task_291_enterprise_industry_analysis_medium_medium106.md
new file mode 100644
index 0000000000000000000000000000000000000000..c7162386262a2d098205464044becfd163ea7173
--- /dev/null
+++ b/tasks/task_291_enterprise_industry_analysis_medium_medium106.md
@@ -0,0 +1,117 @@
+---
+id: task_291_enterprise_industry_analysis_medium_medium106
+name: enterprise_industry_analysis-medium-medium106
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium106.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the total number of Consumer Electronics and Electrical industry enterprises in the province where Shiyang Jinjin Electrical Appliances Company is located and the number of SZSE-listed enterprises in China's Rubber and Plastic Products industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-68.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_292_enterprise_industry_analysis_medium_medium107.md b/tasks/task_292_enterprise_industry_analysis_medium_medium107.md
new file mode 100644
index 0000000000000000000000000000000000000000..82b1093db517fcce552059d81bf45f8b6b8331b1
--- /dev/null
+++ b/tasks/task_292_enterprise_industry_analysis_medium_medium107.md
@@ -0,0 +1,117 @@
+---
+id: task_292_enterprise_industry_analysis_medium_medium107
+name: enterprise_industry_analysis-medium-medium107
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium107.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the total number of Consumer Electronics and Electrical industry enterprises in the province where Shiyang Jinjin Electrical Appliances Company is located and the number of enterprises in China's Rubber and Plastic Products industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-107.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_293_enterprise_industry_analysis_medium_medium108.md b/tasks/task_293_enterprise_industry_analysis_medium_medium108.md
new file mode 100644
index 0000000000000000000000000000000000000000..b49be87ed72b264b32483af7007282545920e5b9
--- /dev/null
+++ b/tasks/task_293_enterprise_industry_analysis_medium_medium108.md
@@ -0,0 +1,117 @@
+---
+id: task_293_enterprise_industry_analysis_medium_medium108
+name: enterprise_industry_analysis-medium-medium108
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium108.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which value is larger: the total number of industry enterprises in the province where Zhongke Keshu Software Company is located, or the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry?
+
+Output guidelines:
+The answer must be either "the total number of industry enterprises in the province where Zhongke Keshu Software Company is located" or "the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_294_enterprise_industry_analysis_medium_medium109.md b/tasks/task_294_enterprise_industry_analysis_medium_medium109.md
new file mode 100644
index 0000000000000000000000000000000000000000..a7950c01f5d21cb6c420d81d14a2eafe9b4a68c0
--- /dev/null
+++ b/tasks/task_294_enterprise_industry_analysis_medium_medium109.md
@@ -0,0 +1,117 @@
+---
+id: task_294_enterprise_industry_analysis_medium_medium109
+name: enterprise_industry_analysis-medium-medium109
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium109.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of SSE-listed private enterprises in the Information Transmission, Software and IT Services industry in the province where Zhongke Keshu Software Company is located, or the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry?
+
+Output guidelines:
+The answer must be either "the number of SSE-listed private enterprises in the Information Transmission, Software and IT Services industry in the province where Zhongke Keshu Software Company is located" or "the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry". Output only the phrase itself, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"the number of SZSE-listed private enterprises in China's Communication Transmission Equipment industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_295_enterprise_industry_analysis_medium_medium110.md b/tasks/task_295_enterprise_industry_analysis_medium_medium110.md
new file mode 100644
index 0000000000000000000000000000000000000000..01f888f5dc296e26804e58318519c8850e0e8bdd
--- /dev/null
+++ b/tasks/task_295_enterprise_industry_analysis_medium_medium110.md
@@ -0,0 +1,117 @@
+---
+id: task_295_enterprise_industry_analysis_medium_medium110
+name: enterprise_industry_analysis-medium-medium110
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium110.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the total annual number of China patent grants in the province where Xingkuwen Arts and Crafts Company is located and that in China's Chemical Fiber Manufacturing industry?
+
+Output guidelines:
+The answer must be a single number rounded to one decimal place. Output only the number, without units, commas, or any explanatory text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-1314.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_296_enterprise_industry_analysis_medium_medium111.md b/tasks/task_296_enterprise_industry_analysis_medium_medium111.md
new file mode 100644
index 0000000000000000000000000000000000000000..3ffbd1d16bb9628d00024e70a7d947829c287cbf
--- /dev/null
+++ b/tasks/task_296_enterprise_industry_analysis_medium_medium111.md
@@ -0,0 +1,117 @@
+---
+id: task_296_enterprise_industry_analysis_medium_medium111
+name: enterprise_industry_analysis-medium-medium111
+category: enterprise_industry_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_analysis/medium111.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of SZSE-listed enterprises in the province where Xingkuwen Arts and Crafts Company is located, or the number of SSE-listed private enterprises in China's Chemical Fiber Manufacturing industry?
+
+Output guidelines:
+The answer must be either "the number of SZSE-listed enterprises in the province where Xingkuwen Arts and Crafts Company is located" or "the number of SSE-listed private enterprises in China's Chemical Fiber Manufacturing industry". Output only one of these, without any explanation, analysis, or descriptive text. If no relevant data can be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"the number of SSE-listed private enterprises in China's Chemical Fiber Manufacturing industry"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_297_enterprise_industry_policy_analysis_easy_easy067.md b/tasks/task_297_enterprise_industry_policy_analysis_easy_easy067.md
new file mode 100644
index 0000000000000000000000000000000000000000..fb745690a72c1382775f7fb6349cf1ca23350709
--- /dev/null
+++ b/tasks/task_297_enterprise_industry_policy_analysis_easy_easy067.md
@@ -0,0 +1,117 @@
+---
+id: task_297_enterprise_industry_policy_analysis_easy_easy067
+name: enterprise_industry_policy_analysis-easy-easy067
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/easy067.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+“江西省人民政府印发关于做优做强我省锂电新能源产业若干政策措施的通知”对众白昌锦商贸公司的行业是否适用？
+
+Output guidelines:
+答案必须是"是"或"否"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答"未查询到相关数据"。
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"否"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_298_enterprise_industry_policy_analysis_easy_easy068.md b/tasks/task_298_enterprise_industry_policy_analysis_easy_easy068.md
new file mode 100644
index 0000000000000000000000000000000000000000..f8de888ca0fd5307f847814b983dacb51df39c11
--- /dev/null
+++ b/tasks/task_298_enterprise_industry_policy_analysis_easy_easy068.md
@@ -0,0 +1,117 @@
+---
+id: task_298_enterprise_industry_policy_analysis_easy_easy068
+name: enterprise_industry_policy_analysis-easy-easy068
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/easy068.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+大花表仪医疗科技公司所在行业是否会受到“合肥市促进“两强一增”行动若干政策”的影响？
+
+Output guidelines:
+答案必须是"是"或"否"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答"未查询到相关数据"。
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"否"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_299_enterprise_industry_policy_analysis_easy_easy069.md b/tasks/task_299_enterprise_industry_policy_analysis_easy_easy069.md
new file mode 100644
index 0000000000000000000000000000000000000000..2fdf22f603f3097dac546718eb57d05722b39948
--- /dev/null
+++ b/tasks/task_299_enterprise_industry_policy_analysis_easy_easy069.md
@@ -0,0 +1,117 @@
+---
+id: task_299_enterprise_industry_policy_analysis_easy_easy069
+name: enterprise_industry_policy_analysis-easy-easy069
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/easy069.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+丽群汇通零售公司所在行业是否会受到“教育部办公厅　工业和信息化部办公厅　国家知识产权局办公室关于组织开展“千校万企”协同创新伙伴行动的通知”的影响？
+
+Output guidelines:
+答案必须是"是"或"否"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答"未查询到相关数据"。
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"否"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_300_enterprise_industry_policy_analysis_easy_easy070.md b/tasks/task_300_enterprise_industry_policy_analysis_easy_easy070.md
new file mode 100644
index 0000000000000000000000000000000000000000..e443e20fae565d9eaab2d065e9263293d16d93ae
--- /dev/null
+++ b/tasks/task_300_enterprise_industry_policy_analysis_easy_easy070.md
@@ -0,0 +1,117 @@
+---
+id: task_300_enterprise_industry_policy_analysis_easy_easy070
+name: enterprise_industry_policy_analysis-easy-easy070
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/easy070.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+“关于印发重庆市促进大中小企业融通发展工作方案（2022—2025年）的通知”对乐动乐博娱乐用品公司的行业是否适用？
+
+Output guidelines:
+答案必须是"是"或"否"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答"未查询到相关数据"。
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"否"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_301_enterprise_industry_policy_analysis_easy_easy071.md b/tasks/task_301_enterprise_industry_policy_analysis_easy_easy071.md
new file mode 100644
index 0000000000000000000000000000000000000000..f6480423bcdfe534317d5568b06ce81416eb17a6
--- /dev/null
+++ b/tasks/task_301_enterprise_industry_policy_analysis_easy_easy071.md
@@ -0,0 +1,117 @@
+---
+id: task_301_enterprise_industry_policy_analysis_easy_easy071
+name: enterprise_industry_policy_analysis-easy-easy071
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/easy071.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+“科技部等九部门关于印发《“十四五” 东西部科技合作实施方案》的通知”对亚玮工泽机床公司的行业是否适用？
+
+Output guidelines:
+答案必须是"是"或"否"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答"未查询到相关数据"。
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"否"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_302_enterprise_industry_policy_analysis_easy_easy072.md b/tasks/task_302_enterprise_industry_policy_analysis_easy_easy072.md
new file mode 100644
index 0000000000000000000000000000000000000000..0d0cc458ac1efdfb7d626a19adc60f4300457c65
--- /dev/null
+++ b/tasks/task_302_enterprise_industry_policy_analysis_easy_easy072.md
@@ -0,0 +1,117 @@
+---
+id: task_302_enterprise_industry_policy_analysis_easy_easy072
+name: enterprise_industry_policy_analysis-easy-easy072
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/easy072.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+果融泽鸿资产管理公司所在行业是否受益于“自治区发展改革委关于印发《宁夏回族自治区氢能产业发展规划》的通知”？
+
+Output guidelines:
+答案必须是"是"或"否"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答"未查询到相关数据"。
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"否"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_303_enterprise_industry_policy_analysis_easy_easy073.md b/tasks/task_303_enterprise_industry_policy_analysis_easy_easy073.md
new file mode 100644
index 0000000000000000000000000000000000000000..adbcb4d335dbf9191560f0afabc440c2f7dbd409
--- /dev/null
+++ b/tasks/task_303_enterprise_industry_policy_analysis_easy_easy073.md
@@ -0,0 +1,117 @@
+---
+id: task_303_enterprise_industry_policy_analysis_easy_easy073
+name: enterprise_industry_policy_analysis-easy-easy073
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/easy073.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+“商务部等14部门关于开展内外贸一体化试点的通知”是否可以推进物丽昌源批发公司所属行业发展？
+
+Output guidelines:
+答案必须是"是"或"否"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答"未查询到相关数据"。
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"是"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_304_enterprise_industry_policy_analysis_easy_easy074.md b/tasks/task_304_enterprise_industry_policy_analysis_easy_easy074.md
new file mode 100644
index 0000000000000000000000000000000000000000..618fcf9019a77e92c1f80411c4e3021951f2b4b7
--- /dev/null
+++ b/tasks/task_304_enterprise_industry_policy_analysis_easy_easy074.md
@@ -0,0 +1,117 @@
+---
+id: task_304_enterprise_industry_policy_analysis_easy_easy074
+name: enterprise_industry_policy_analysis-easy-easy074
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/easy074.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+“广东省人民政府关于印发中国（韶关）等8个  跨境电子商务综合试验区实施方案的通知”对恒通达达信息技术公司的行业是否适用？
+
+Output guidelines:
+答案必须是"是"或"否"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答"未查询到相关数据"。
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"是"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_305_enterprise_industry_policy_analysis_easy_easy075.md b/tasks/task_305_enterprise_industry_policy_analysis_easy_easy075.md
new file mode 100644
index 0000000000000000000000000000000000000000..61db344161d00ddaaccb725ffb2350360229bcff
--- /dev/null
+++ b/tasks/task_305_enterprise_industry_policy_analysis_easy_easy075.md
@@ -0,0 +1,117 @@
+---
+id: task_305_enterprise_industry_policy_analysis_easy_easy075
+name: enterprise_industry_policy_analysis-easy-easy075
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/easy075.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+众课科数软件公司所在行业是否受益于“民政部、中央政法委、中央网信办、发展改革委、工业和信息化部、公安部、财政部、住房城乡建设部、农业农村部印发《关于深入推进智慧社区建设的意见》的通知”？
+
+Output guidelines:
+答案必须是"是"或"否"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答"未查询到相关数据"。
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"是"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_306_enterprise_industry_policy_analysis_easy_easy076.md b/tasks/task_306_enterprise_industry_policy_analysis_easy_easy076.md
new file mode 100644
index 0000000000000000000000000000000000000000..e3538fbf3a7af7499cb6c8178f940e2957380a9c
--- /dev/null
+++ b/tasks/task_306_enterprise_industry_policy_analysis_easy_easy076.md
@@ -0,0 +1,117 @@
+---
+id: task_306_enterprise_industry_policy_analysis_easy_easy076
+name: enterprise_industry_policy_analysis-easy-easy076
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/easy076.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+“四部门关于公布农业、建筑、医疗、矿山领域机器人典型应用场景名单的通知”对以山泽辰医疗器械公司的行业是否适用？
+
+Output guidelines:
+答案必须是"是"或"否"，仅输出一个字，不要添加任何解释或说明。如果无法找到相关数据，请回答"未查询到相关数据"。
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"否"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_307_enterprise_industry_policy_analysis_medium_medium001.md b/tasks/task_307_enterprise_industry_policy_analysis_medium_medium001.md
new file mode 100644
index 0000000000000000000000000000000000000000..5f266813b5d7008742c36209e53c9e44988880e1
--- /dev/null
+++ b/tasks/task_307_enterprise_industry_policy_analysis_medium_medium001.md
@@ -0,0 +1,117 @@
+---
+id: task_307_enterprise_industry_policy_analysis_medium_medium001
+name: enterprise_industry_policy_analysis-medium-medium001
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium001.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the gap between the number of central ministry/agency policies issued by the Ministry of Commerce in the ministerial policies for the industry of Wu Li Hui Da Lian Suo Co., Ltd. and the number of local policies issued by the Hainan Province Department of Industry and Information Technology for the industry of Xing Ku Wen Gong Yi Mei Shu Pin Co., Ltd.?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`2.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_308_enterprise_industry_policy_analysis_medium_medium002.md b/tasks/task_308_enterprise_industry_policy_analysis_medium_medium002.md
new file mode 100644
index 0000000000000000000000000000000000000000..f5da464a4b7591fd783c52a8266f7e9e21e82d85
--- /dev/null
+++ b/tasks/task_308_enterprise_industry_policy_analysis_medium_medium002.md
@@ -0,0 +1,117 @@
+---
+id: task_308_enterprise_industry_policy_analysis_medium_medium002
+name: enterprise_industry_policy_analysis-medium-medium002
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium002.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of policies issued by the Gansu Province General Office of the People's Government in the local policies for the Wholesale and Retail industry and the number of policies issued by the Ministry of Transport in the ministerial policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without any unit, comma, or explanatory text. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_309_enterprise_industry_policy_analysis_medium_medium003.md b/tasks/task_309_enterprise_industry_policy_analysis_medium_medium003.md
new file mode 100644
index 0000000000000000000000000000000000000000..19117dd0da9cb78fa93a4f7887874f34b351bfca
--- /dev/null
+++ b/tasks/task_309_enterprise_industry_policy_analysis_medium_medium003.md
@@ -0,0 +1,117 @@
+---
+id: task_309_enterprise_industry_policy_analysis_medium_medium003
+name: enterprise_industry_policy_analysis-medium-medium003
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium003.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of policies issued by the General Administration of Customs in the ministerial policies for the industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd., or the number of policies issued by the General Office of the State Council in the State Council policies for the industry of Bei Kong Ze Jing Water Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Bei Kong Ze Jing Water Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_310_enterprise_industry_policy_analysis_medium_medium004.md b/tasks/task_310_enterprise_industry_policy_analysis_medium_medium004.md
new file mode 100644
index 0000000000000000000000000000000000000000..1847bb9cc5fc716b7817118b32fb5df14b3e9f70
--- /dev/null
+++ b/tasks/task_310_enterprise_industry_policy_analysis_medium_medium004.md
@@ -0,0 +1,117 @@
+---
+id: task_310_enterprise_industry_policy_analysis_medium_medium004
+name: enterprise_industry_policy_analysis-medium-medium004
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium004.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is larger: the number of central ministry/agency policies issued by the Ministry of Agriculture and Rural Affairs in the ministerial policies for the Financial Services industry, or the number of central ministry/agency policies issued by the National Health Commission in the ministerial policies for the Water Conservancy, Environment and Public Facilities Management industry?
+
+Output guidelines:
+The answer must be a company name; output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Bei Kong Ze Jing Water Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_311_enterprise_industry_policy_analysis_medium_medium005.md b/tasks/task_311_enterprise_industry_policy_analysis_medium_medium005.md
new file mode 100644
index 0000000000000000000000000000000000000000..a794a419a889034b51c19dff9ca2f16b7f1c0e62
--- /dev/null
+++ b/tasks/task_311_enterprise_industry_policy_analysis_medium_medium005.md
@@ -0,0 +1,117 @@
+---
+id: task_311_enterprise_industry_policy_analysis_medium_medium005
+name: enterprise_industry_policy_analysis-medium-medium005
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium005.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is the number of central ministry/agency policies issued by the Development and Reform Commission in the ministerial policies for the industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd. higher than the number of local policies issued by the Shenzhen Municipality Bureau of Commerce in the local policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd.?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_312_enterprise_industry_policy_analysis_medium_medium006.md b/tasks/task_312_enterprise_industry_policy_analysis_medium_medium006.md
new file mode 100644
index 0000000000000000000000000000000000000000..88149bf316e5213057a1fd71be6567cc228ae44f
--- /dev/null
+++ b/tasks/task_312_enterprise_industry_policy_analysis_medium_medium006.md
@@ -0,0 +1,117 @@
+---
+id: task_312_enterprise_industry_policy_analysis_medium_medium006
+name: enterprise_industry_policy_analysis-medium-medium006
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium006.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of local policies issued by the Yunnan Province Communications Administration in the local policies for the industry of Zhong Ke Zhi Yun Shu Ju Fu Wu Co., Ltd., or the number of central ministry/agency policies issued by the Development and Reform Commission in the ministerial policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd.?
+
+Output guidelines:
+The answer must be either a company name or "Equal"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_313_enterprise_industry_policy_analysis_medium_medium007.md b/tasks/task_313_enterprise_industry_policy_analysis_medium_medium007.md
new file mode 100644
index 0000000000000000000000000000000000000000..d91c765c865b125c38689cbe29e8197dd23691ca
--- /dev/null
+++ b/tasks/task_313_enterprise_industry_policy_analysis_medium_medium007.md
@@ -0,0 +1,117 @@
+---
+id: task_313_enterprise_industry_policy_analysis_medium_medium007
+name: enterprise_industry_policy_analysis-medium-medium007
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium007.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of policies issued by the Ministry of Industry and Information Technology in the ministerial policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd. and the number of local policies issued by the Guangdong Provincial People's Government for the industry of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`1.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_314_enterprise_industry_policy_analysis_medium_medium008.md b/tasks/task_314_enterprise_industry_policy_analysis_medium_medium008.md
new file mode 100644
index 0000000000000000000000000000000000000000..a1c1cb9684d2004a3c156e2bbcb780fb1567a4df
--- /dev/null
+++ b/tasks/task_314_enterprise_industry_policy_analysis_medium_medium008.md
@@ -0,0 +1,117 @@
+---
+id: task_314_enterprise_industry_policy_analysis_medium_medium008
+name: enterprise_industry_policy_analysis-medium-medium008
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium008.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of local policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd., or the number of policies issued by the General Office of the State Council in the State Council policies for the industry of Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "industry". Output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yi Hai Chang Jin Shang Wu Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_315_enterprise_industry_policy_analysis_medium_medium009.md b/tasks/task_315_enterprise_industry_policy_analysis_medium_medium009.md
new file mode 100644
index 0000000000000000000000000000000000000000..f451059460f3e1019048413748c2b042df90ceb8
--- /dev/null
+++ b/tasks/task_315_enterprise_industry_policy_analysis_medium_medium009.md
@@ -0,0 +1,117 @@
+---
+id: task_315_enterprise_industry_policy_analysis_medium_medium009
+name: enterprise_industry_policy_analysis-medium-medium009
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium009.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of local policies issued by the Guangdong Province General Office of the People's Government in the local policies for the industry of Zhang Qiao Jin Chuang Technology Co., Ltd., or the number of central ministry/agency policies issued by the National Cryptography Administration in the ministerial policies for the industry of Heng Li Ke Zhi Ruan Jian Co., Ltd.?
+
+Output guidelines:
+The answer must be either "Number of local policies issued by the Guangdong Province General Office of the People's Government for the industry of Zhang Qiao Jin Chuang Technology Co., Ltd." or "Number of central ministry/agency policies issued by the National Cryptography Administration for the industry of Heng Li Ke Zhi Ruan Jian Co., Ltd." or "Equal". Output only the selected answer text, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_316_enterprise_industry_policy_analysis_medium_medium010.md b/tasks/task_316_enterprise_industry_policy_analysis_medium_medium010.md
new file mode 100644
index 0000000000000000000000000000000000000000..3939a6e09872f57df7702a129d814ef14e5b1277
--- /dev/null
+++ b/tasks/task_316_enterprise_industry_policy_analysis_medium_medium010.md
@@ -0,0 +1,117 @@
+---
+id: task_316_enterprise_industry_policy_analysis_medium_medium010
+name: enterprise_industry_policy_analysis-medium-medium010
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium010.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of local policies issued by the Hainan Province Department of Finance for the industry of Zhang Qiao Jin Chuang Technology Co., Ltd., or the number of central ministry/agency policies issued by the National Health Commission for the industry of Heng Li Ke Zhi Ruan Jian Co., Ltd.?
+
+Output guidelines:
+The answer must be a company name or "Equal". Output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_317_enterprise_industry_policy_analysis_medium_medium011.md b/tasks/task_317_enterprise_industry_policy_analysis_medium_medium011.md
new file mode 100644
index 0000000000000000000000000000000000000000..057035bdcb70772c2f7f5c4b4436eacd55de867c
--- /dev/null
+++ b/tasks/task_317_enterprise_industry_policy_analysis_medium_medium011.md
@@ -0,0 +1,117 @@
+---
+id: task_317_enterprise_industry_policy_analysis_medium_medium011
+name: enterprise_industry_policy_analysis-medium-medium011
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium011.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of central ministry/agency policies issued by the State Administration of Foreign Exchange in the ministerial policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. and the number of local policies issued by the Guangdong Provincial Committee of the CPC for the Financial Services industry in Guangdong Province where Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_318_enterprise_industry_policy_analysis_medium_medium012.md b/tasks/task_318_enterprise_industry_policy_analysis_medium_medium012.md
new file mode 100644
index 0000000000000000000000000000000000000000..9e49a6c6572aa04ef46a4c5110ea428ac8420959
--- /dev/null
+++ b/tasks/task_318_enterprise_industry_policy_analysis_medium_medium012.md
@@ -0,0 +1,117 @@
+---
+id: task_318_enterprise_industry_policy_analysis_medium_medium012
+name: enterprise_industry_policy_analysis-medium-medium012
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium012.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of central ministry/agency policies issued by the People's Bank of China in the ministerial policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd., or the number of local policies issued by the Guangdong Provincial Committee of the CPC for the Financial Services industry in Guangdong Province where Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a company name or "Equal"; output only the selected answer text, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_319_enterprise_industry_policy_analysis_medium_medium013.md b/tasks/task_319_enterprise_industry_policy_analysis_medium_medium013.md
new file mode 100644
index 0000000000000000000000000000000000000000..175482cd0d8f8e28dd0cca9ee757bcb7d915b68b
--- /dev/null
+++ b/tasks/task_319_enterprise_industry_policy_analysis_medium_medium013.md
@@ -0,0 +1,117 @@
+---
+id: task_319_enterprise_industry_policy_analysis_medium_medium013
+name: enterprise_industry_policy_analysis-medium-medium013
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium013.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of local policies issued by the Yunnan Province Development and Reform Commission for the industry of Bei Kong Ze Jing Water Co., Ltd., or the number of local policies issued by the Shanghai Municipality Science and Technology Commission for the Health and Social Work industry in Shanghai Municipality where Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a company name or "Equal"; output only the selected answer text, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Bei Kong Ze Jing Water Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_320_enterprise_industry_policy_analysis_medium_medium014.md b/tasks/task_320_enterprise_industry_policy_analysis_medium_medium014.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ffd15337d052ecde83a6d090dedf2e1a0060004
--- /dev/null
+++ b/tasks/task_320_enterprise_industry_policy_analysis_medium_medium014.md
@@ -0,0 +1,117 @@
+---
+id: task_320_enterprise_industry_policy_analysis_medium_medium014
+name: enterprise_industry_policy_analysis-medium-medium014
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium014.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is higher: the number of central ministry/agency policies issued by the Ministry of Agriculture and Rural Affairs in the ministerial policies for the industry of Bei Kong Ze Jing Water Co., Ltd., or the number of local policies issued by the Shanghai Municipality Health Commission for the Health and Social Work industry in Shanghai Municipality where Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a company name or "industry"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Bei Kong Ze Jing Water Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_321_enterprise_industry_policy_analysis_medium_medium015.md b/tasks/task_321_enterprise_industry_policy_analysis_medium_medium015.md
new file mode 100644
index 0000000000000000000000000000000000000000..7d7dc967bfe05394e459a4ae84be7ea88e592413
--- /dev/null
+++ b/tasks/task_321_enterprise_industry_policy_analysis_medium_medium015.md
@@ -0,0 +1,117 @@
+---
+id: task_321_enterprise_industry_policy_analysis_medium_medium015
+name: enterprise_industry_policy_analysis-medium-medium015
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium015.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of central ministry/agency policies issued by the State Administration of Foreign Exchange in the ministerial policies for the industry of Wu Li Hui Da Lian Suo Co., Ltd., or the number of local policies for the Water Conservancy, Environment and Public Facilities Management industry in Guangdong Province where Bei Kong Ze Jing Water Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a company name or "Equal"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_322_enterprise_industry_policy_analysis_medium_medium016.md b/tasks/task_322_enterprise_industry_policy_analysis_medium_medium016.md
new file mode 100644
index 0000000000000000000000000000000000000000..c67c1ddb9f07f1edece8dfe8e5913be815dd17fb
--- /dev/null
+++ b/tasks/task_322_enterprise_industry_policy_analysis_medium_medium016.md
@@ -0,0 +1,117 @@
+---
+id: task_322_enterprise_industry_policy_analysis_medium_medium016
+name: enterprise_industry_policy_analysis-medium-medium016
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium016.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of local policies issued by the Shenzhen Municipal People's Government for the industry of Wu Li Hui Da Lian Suo Co., Ltd., or the total number of policies for the Water Conservancy, Environment and Public Facilities Management industry in Guangdong Province where Bei Kong Ze Jing Water Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a company name or "Equal"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_323_enterprise_industry_policy_analysis_medium_medium017.md b/tasks/task_323_enterprise_industry_policy_analysis_medium_medium017.md
new file mode 100644
index 0000000000000000000000000000000000000000..6edf7825cea9b8395ddb8ea3835ae4a1db2ba05c
--- /dev/null
+++ b/tasks/task_323_enterprise_industry_policy_analysis_medium_medium017.md
@@ -0,0 +1,117 @@
+---
+id: task_323_enterprise_industry_policy_analysis_medium_medium017
+name: enterprise_industry_policy_analysis-medium-medium017
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium017.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of central ministry/agency policies issued by the National Health Commission in the ministerial policies for the industry of Bei Kong Ze Jing Water Co., Ltd., or the number of local policies issued by the Guangdong Provincial Committee of the CPC for the Financial Services industry in Guangdong Province where Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a company name or "industry"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Bei Kong Ze Jing Water Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_324_enterprise_industry_policy_analysis_medium_medium018.md b/tasks/task_324_enterprise_industry_policy_analysis_medium_medium018.md
new file mode 100644
index 0000000000000000000000000000000000000000..8f3f92c7b8100e4da2b56c0ae10bec7d83e69377
--- /dev/null
+++ b/tasks/task_324_enterprise_industry_policy_analysis_medium_medium018.md
@@ -0,0 +1,117 @@
+---
+id: task_324_enterprise_industry_policy_analysis_medium_medium018
+name: enterprise_industry_policy_analysis-medium-medium018
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium018.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of local policies issued by the Liaoning Province People's Government for the industry of Bei Kong Ze Jing Water Co., Ltd., or the number of local policies issued by the Guangdong Province General Office of the People's Government for the Financial Services industry in Guangdong Province where Hui Jin Jin Rui Cai Fu Management Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a company name or "industry"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Hui Jin Jin Rui Cai Fu Management Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_325_enterprise_industry_policy_analysis_medium_medium019.md b/tasks/task_325_enterprise_industry_policy_analysis_medium_medium019.md
new file mode 100644
index 0000000000000000000000000000000000000000..5faa5c2c1669be0f54a8a96e32ce0d7a65140e43
--- /dev/null
+++ b/tasks/task_325_enterprise_industry_policy_analysis_medium_medium019.md
@@ -0,0 +1,117 @@
+---
+id: task_325_enterprise_industry_policy_analysis_medium_medium019
+name: enterprise_industry_policy_analysis-medium-medium019
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium019.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of local policies issued by the Shaanxi Province Development and Reform Commission for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. and the total number of policies for the Conglomerates industry in Guangdong Province where Hua Cheng Sheng Yuan Integrated Development Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-1.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_326_enterprise_industry_policy_analysis_medium_medium020.md b/tasks/task_326_enterprise_industry_policy_analysis_medium_medium020.md
new file mode 100644
index 0000000000000000000000000000000000000000..437f7348a923359347b30bfcf30eb2da60bbe986
--- /dev/null
+++ b/tasks/task_326_enterprise_industry_policy_analysis_medium_medium020.md
@@ -0,0 +1,117 @@
+---
+id: task_326_enterprise_industry_policy_analysis_medium_medium020
+name: enterprise_industry_policy_analysis-medium-medium020
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium020.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of central ministry/agency policies issued by the Ministry of Housing and Urban-Rural Development in the ministerial policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd., or the number of local policies for the Conglomerates industry in Guangdong Province where Hua Cheng Sheng Yuan Zong He Kai Fa Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a company name or "Equal"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_327_enterprise_industry_policy_analysis_medium_medium021.md b/tasks/task_327_enterprise_industry_policy_analysis_medium_medium021.md
new file mode 100644
index 0000000000000000000000000000000000000000..37bf9353d33e7f92ea4a833b2fa1b3859180038f
--- /dev/null
+++ b/tasks/task_327_enterprise_industry_policy_analysis_medium_medium021.md
@@ -0,0 +1,117 @@
+---
+id: task_327_enterprise_industry_policy_analysis_medium_medium021
+name: enterprise_industry_policy_analysis-medium-medium021
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium021.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of specific local policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd., or the number of similar local policies for the industry of Bao Xin Hui Hui Wang Luo Co., Ltd. in its province?
+
+Output guidelines:
+The answer must be a company name or "industry"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Bao Xin Hui Hui Wang Luo Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_328_enterprise_industry_policy_analysis_medium_medium022.md b/tasks/task_328_enterprise_industry_policy_analysis_medium_medium022.md
new file mode 100644
index 0000000000000000000000000000000000000000..20b2124d2da87ab24e95c590bdd5a2791bbc1d60
--- /dev/null
+++ b/tasks/task_328_enterprise_industry_policy_analysis_medium_medium022.md
@@ -0,0 +1,117 @@
+---
+id: task_328_enterprise_industry_policy_analysis_medium_medium022
+name: enterprise_industry_policy_analysis-medium-medium022
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium022.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between Zhong Ke Ke Shu Ruan Jian Co., Ltd. and Bao Xin Hui Hui Wang Luo Co., Ltd., which obtains a greater number of local policy supports?
+
+Output guidelines:
+The answer must be a company name or "Equal"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_329_enterprise_industry_policy_analysis_medium_medium023.md b/tasks/task_329_enterprise_industry_policy_analysis_medium_medium023.md
new file mode 100644
index 0000000000000000000000000000000000000000..a2103cccf738bc7475c682a3163394607df2626b
--- /dev/null
+++ b/tasks/task_329_enterprise_industry_policy_analysis_medium_medium023.md
@@ -0,0 +1,117 @@
+---
+id: task_329_enterprise_industry_policy_analysis_medium_medium023
+name: enterprise_industry_policy_analysis-medium-medium023
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium023.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of local policies issued by the Hunan Province General Office of the People's Government for the industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd. and the total number of policies for the Consumer Electronics and Electrical Equipment industry in Guangdong Province where Zhang Qiao Jin Chuang Technology Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_330_enterprise_industry_policy_analysis_medium_medium024.md b/tasks/task_330_enterprise_industry_policy_analysis_medium_medium024.md
new file mode 100644
index 0000000000000000000000000000000000000000..afb9ecc09159916e45fc438aed67418747ac2d71
--- /dev/null
+++ b/tasks/task_330_enterprise_industry_policy_analysis_medium_medium024.md
@@ -0,0 +1,117 @@
+---
+id: task_330_enterprise_industry_policy_analysis_medium_medium024
+name: enterprise_industry_policy_analysis-medium-medium024
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium024.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Which is greater: the number of central ministry/agency policies issued by the Ministry of Science and Technology in the ministerial policies for the industry of Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd., or the number of local policies for the Consumer Electronics and Electrical Equipment industry in Guangdong Province where Zhang Qiao Jin Chuang Technology Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a company name or "industry"; output only the answer text, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Zhao Ye Hua Chang Fang Di Chan Kai Fa Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_331_enterprise_industry_policy_analysis_medium_medium025.md b/tasks/task_331_enterprise_industry_policy_analysis_medium_medium025.md
new file mode 100644
index 0000000000000000000000000000000000000000..8e376f0d7e0b2b23b210867e00d6bc46bc30c68f
--- /dev/null
+++ b/tasks/task_331_enterprise_industry_policy_analysis_medium_medium025.md
@@ -0,0 +1,117 @@
+---
+id: task_331_enterprise_industry_policy_analysis_medium_medium025
+name: enterprise_industry_policy_analysis-medium-medium025
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium025.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of local policies issued by the Anhui Province People's Government for the industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd. and the total number of local policies for the Consumer Electronics and Electrical Equipment industry in Guangdong Province where Zhang Qiao Jin Chuang Technology Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_332_enterprise_industry_policy_analysis_medium_medium026.md b/tasks/task_332_enterprise_industry_policy_analysis_medium_medium026.md
new file mode 100644
index 0000000000000000000000000000000000000000..f045b8aae635f17c2ee0ec0ae2868849b8127b46
--- /dev/null
+++ b/tasks/task_332_enterprise_industry_policy_analysis_medium_medium026.md
@@ -0,0 +1,117 @@
+---
+id: task_332_enterprise_industry_policy_analysis_medium_medium026
+name: enterprise_industry_policy_analysis-medium-medium026
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium026.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of central ministry/agency policies issued by the General Office of the China Banking and Insurance Regulatory Commission in the ministerial policies for the industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd. and the number of local policies for the Consumer Electronics and Electrical Equipment industry in Guangdong Province where Zhang Qiao Jin Chuang Technology Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_333_enterprise_industry_policy_analysis_medium_medium027.md b/tasks/task_333_enterprise_industry_policy_analysis_medium_medium027.md
new file mode 100644
index 0000000000000000000000000000000000000000..87bac380a0c95b7c9aa0bdeacb400d1a96c15109
--- /dev/null
+++ b/tasks/task_333_enterprise_industry_policy_analysis_medium_medium027.md
@@ -0,0 +1,117 @@
+---
+id: task_333_enterprise_industry_policy_analysis_medium_medium027
+name: enterprise_industry_policy_analysis-medium-medium027
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium027.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of local policies issued by the General Office of the People's Government of Henan Province in the local policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd. and the number of local policies issued by the Shanghai Municipality Finance Bureau in the local policies for the industry of Lang Ji Hui Ruan Technology Co., Ltd. in Shanghai, which is greater?
+
+Output guidelines:
+The answer must be a company name or the word "industry"; output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Lang Ji Hui Ruan Technology Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_334_enterprise_industry_policy_analysis_medium_medium028.md b/tasks/task_334_enterprise_industry_policy_analysis_medium_medium028.md
new file mode 100644
index 0000000000000000000000000000000000000000..fd22393f0dd6a85fec5e56d4e1bfe5a0ba9be393
--- /dev/null
+++ b/tasks/task_334_enterprise_industry_policy_analysis_medium_medium028.md
@@ -0,0 +1,117 @@
+---
+id: task_334_enterprise_industry_policy_analysis_medium_medium028
+name: enterprise_industry_policy_analysis-medium-medium028
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium028.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of central ministry/agency policies issued by the China Federation of Logistics & Purchasing in the ministerial policies for the industry of Yi Hai Chang Jin Shang Wu Co., Ltd. and the number of local policies for the Information Transmission, Software and IT Services industry in Shanghai Municipality where Lang Ji Hui Ruan Technology Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-14.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_335_enterprise_industry_policy_analysis_medium_medium029.md b/tasks/task_335_enterprise_industry_policy_analysis_medium_medium029.md
new file mode 100644
index 0000000000000000000000000000000000000000..fd3adc2afe231b772b5e2413d2e04bc152ce498a
--- /dev/null
+++ b/tasks/task_335_enterprise_industry_policy_analysis_medium_medium029.md
@@ -0,0 +1,117 @@
+---
+id: task_335_enterprise_industry_policy_analysis_medium_medium029
+name: enterprise_industry_policy_analysis-medium-medium029
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium029.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of central ministry/agency policies in the ministerial policies for the industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd. and the number of local policies in the local policies for the industry of Hua Cheng Sheng Yuan Integrated Development Co., Ltd. in Guangdong Province, which is greater?
+
+Output guidelines:
+The answer must be a company name or the word "industry"; output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Zhao Ye Hua Chang Real Estate Development Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_336_enterprise_industry_policy_analysis_medium_medium030.md b/tasks/task_336_enterprise_industry_policy_analysis_medium_medium030.md
new file mode 100644
index 0000000000000000000000000000000000000000..7575324b1002025dea0b4cbdd8da9084a222f89a
--- /dev/null
+++ b/tasks/task_336_enterprise_industry_policy_analysis_medium_medium030.md
@@ -0,0 +1,117 @@
+---
+id: task_336_enterprise_industry_policy_analysis_medium_medium030
+name: enterprise_industry_policy_analysis-medium-medium030
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium030.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of central ministry/agency policies issued by the Ministry of Housing and Urban-Rural Development in the ministerial policies for the industry of Zhao Ye Hua Chang Real Estate Development Co., Ltd. and the number of policies for the Integrated industry in Guangdong Province where Hua Cheng Sheng Yuan Integrated Development Co., Ltd. is located?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-1.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_337_enterprise_industry_policy_analysis_medium_medium031.md b/tasks/task_337_enterprise_industry_policy_analysis_medium_medium031.md
new file mode 100644
index 0000000000000000000000000000000000000000..e41128fa5ded970e750eaada2b9d5be218cc74d4
--- /dev/null
+++ b/tasks/task_337_enterprise_industry_policy_analysis_medium_medium031.md
@@ -0,0 +1,117 @@
+---
+id: task_337_enterprise_industry_policy_analysis_medium_medium031
+name: enterprise_industry_policy_analysis-medium-medium031
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium031.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of central ministry/agency policies issued by the Ministry of Culture and Tourism in the ministerial policies for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the number of local policies issued by the Chengdu Municipality Bureau of Economy and Information Technology in the local policies for the Commercial Electrical Machinery and Equipment Manufacturing industry?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_338_enterprise_industry_policy_analysis_medium_medium032.md b/tasks/task_338_enterprise_industry_policy_analysis_medium_medium032.md
new file mode 100644
index 0000000000000000000000000000000000000000..d5bdd101ffa0adbc55f8071d06274c3434abf290
--- /dev/null
+++ b/tasks/task_338_enterprise_industry_policy_analysis_medium_medium032.md
@@ -0,0 +1,117 @@
+---
+id: task_338_enterprise_industry_policy_analysis_medium_medium032
+name: enterprise_industry_policy_analysis-medium-medium032
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium032.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of local policies related to Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the corresponding number of local policies for the Commercial Electrical Machinery and Equipment Manufacturing industry, which is greater?
+
+Output guidelines:
+The answer must be a company name or the word "industry"; output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Hui Jin Jin Rui Cai Fu Management Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_339_enterprise_industry_policy_analysis_medium_medium033.md b/tasks/task_339_enterprise_industry_policy_analysis_medium_medium033.md
new file mode 100644
index 0000000000000000000000000000000000000000..074be64676bae3056f7365a9b31eeaa825ab7037
--- /dev/null
+++ b/tasks/task_339_enterprise_industry_policy_analysis_medium_medium033.md
@@ -0,0 +1,117 @@
+---
+id: task_339_enterprise_industry_policy_analysis_medium_medium033
+name: enterprise_industry_policy_analysis-medium-medium033
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium033.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of local policies issued by the General Office of the Guangxi Zhuang Autonomous Regional People's Government in the local policies for the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. and the number of local policies issued by the Guizhou Province Department of Housing and Urban-Rural Development in the local policies for the Information Transmission, Software and IT Services industry, which is greater?
+
+Output guidelines:
+The answer must be "Equal", a company name, or the word "industry"; output only one word or company name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_340_enterprise_industry_policy_analysis_medium_medium034.md b/tasks/task_340_enterprise_industry_policy_analysis_medium_medium034.md
new file mode 100644
index 0000000000000000000000000000000000000000..73f45a997f2e3c181907e0664efa05640e69e100
--- /dev/null
+++ b/tasks/task_340_enterprise_industry_policy_analysis_medium_medium034.md
@@ -0,0 +1,117 @@
+---
+id: task_340_enterprise_industry_policy_analysis_medium_medium034
+name: enterprise_industry_policy_analysis-medium-medium034
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium034.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of policies issued by the Yunnan Province Department of Industry and Information Technology in the local policies for the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. and the number of policies issued by the Provincial General Office of the People's Government in the local policies for the Information Transmission, Software and IT Services industry, which is greater?
+
+Output guidelines:
+The answer must be "Number of policies for the industry of Hua Tu Wen Jiao Zai Xian Jiao Yu Co., Ltd. issued by the Yunnan Province Department of Industry and Information Technology", "Number of policies for the Information Transmission, Software and IT Services industry issued by the Provincial General Office of the People's Government", or "Equal". Output only the specified answer text, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_341_enterprise_industry_policy_analysis_medium_medium035.md b/tasks/task_341_enterprise_industry_policy_analysis_medium_medium035.md
new file mode 100644
index 0000000000000000000000000000000000000000..20b26b0661812b973ffe5f287fc003ad1a4d6ce6
--- /dev/null
+++ b/tasks/task_341_enterprise_industry_policy_analysis_medium_medium035.md
@@ -0,0 +1,117 @@
+---
+id: task_341_enterprise_industry_policy_analysis_medium_medium035
+name: enterprise_industry_policy_analysis-medium-medium035
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium035.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of local policies issued by the Shanghai Municipality Commission of Economy and Information Technology in the local policies for the industry of Rui Xing Jian Kang Zhi Yao Co., Ltd. and the number of local policies issued by the Liaoning Province People's Government in the local policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`2.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_342_enterprise_industry_policy_analysis_medium_medium036.md b/tasks/task_342_enterprise_industry_policy_analysis_medium_medium036.md
new file mode 100644
index 0000000000000000000000000000000000000000..9c6810ad893a2cb985b830942552e21a1e483a13
--- /dev/null
+++ b/tasks/task_342_enterprise_industry_policy_analysis_medium_medium036.md
@@ -0,0 +1,117 @@
+---
+id: task_342_enterprise_industry_policy_analysis_medium_medium036
+name: enterprise_industry_policy_analysis-medium-medium036
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium036.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of local policies issued by the Shanghai Municipality Commission of Economy and Information Technology in the local policies for the industry of Rui Xing Jian Kang Zhi Yao Co., Ltd. and the number of central ministry/agency policies issued by the Ministry of Agriculture and Rural Affairs in the ministerial policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry, which is greater?
+
+Output guidelines:
+The answer must be a company name or the word "industry"; output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Rui Xing Jian Kang Zhi Yao Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_343_enterprise_industry_policy_analysis_medium_medium037.md b/tasks/task_343_enterprise_industry_policy_analysis_medium_medium037.md
new file mode 100644
index 0000000000000000000000000000000000000000..ed5031cd89be9fccbd521181cc57a90f2d4f72db
--- /dev/null
+++ b/tasks/task_343_enterprise_industry_policy_analysis_medium_medium037.md
@@ -0,0 +1,117 @@
+---
+id: task_343_enterprise_industry_policy_analysis_medium_medium037
+name: enterprise_industry_policy_analysis-medium-medium037
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium037.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of local policies issued by the Secretariat of the Office of Nansha Development Zone Administrative Committee, Guangzhou in the local policies for the industry of Bao Xin Hui Hui Wang Luo Co., Ltd. and the number of local policies issued by the Shanghai Municipality Finance Bureau in the local policies for the General Equipment Manufacturing industry, which is greater?
+
+Output guidelines:
+The answer must be "Number of local policies for the industry of Bao Xin Hui Hui Wang Luo Co., Ltd. issued by the Secretariat of the Office of Nansha Development Zone Administrative Committee, Guangzhou" or "Number of local policies for the General Equipment Manufacturing industry issued by the Shanghai Municipality Finance Bureau". Output only the policy title, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"General Equipment Manufacturing"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_344_enterprise_industry_policy_analysis_medium_medium038.md b/tasks/task_344_enterprise_industry_policy_analysis_medium_medium038.md
new file mode 100644
index 0000000000000000000000000000000000000000..a8c118bbc3bbde23f4e010b2cf54ae91800df283
--- /dev/null
+++ b/tasks/task_344_enterprise_industry_policy_analysis_medium_medium038.md
@@ -0,0 +1,117 @@
+---
+id: task_344_enterprise_industry_policy_analysis_medium_medium038
+name: enterprise_industry_policy_analysis-medium-medium038
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium038.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of central ministry/agency policies issued by the General Office of the China National Intellectual Property Administration in the ministerial policies for the industry of Bao Xin Hui Hui Wang Luo Co., Ltd. and the number of local policies issued by the Sichuan Province People's Government in the local policies for the General Equipment Manufacturing industry?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`1.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_345_enterprise_industry_policy_analysis_medium_medium039.md b/tasks/task_345_enterprise_industry_policy_analysis_medium_medium039.md
new file mode 100644
index 0000000000000000000000000000000000000000..6de98e51bc88ed1bbd3a6164c0fab53ba34a59c9
--- /dev/null
+++ b/tasks/task_345_enterprise_industry_policy_analysis_medium_medium039.md
@@ -0,0 +1,117 @@
+---
+id: task_345_enterprise_industry_policy_analysis_medium_medium039
+name: enterprise_industry_policy_analysis-medium-medium039
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium039.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of local policies issued by the Hefei Municipality Office of the People's Government in the local policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. and the number of local policies issued by the Shandong Province People's Government in the local policies for the Scientific Research and Technical Services industry, which is greater?
+
+Output guidelines:
+The answer must be "Number of local policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. issued by the Hefei Municipality Office of the People's Government" or "Number of local policies for the Scientific Research and Technical Services industry issued by the Shandong Province People's Government". Output only the policy title, without any explanation, analysis, or descriptive wording. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Scientific Research and Technical Services"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_346_enterprise_industry_policy_analysis_medium_medium040.md b/tasks/task_346_enterprise_industry_policy_analysis_medium_medium040.md
new file mode 100644
index 0000000000000000000000000000000000000000..0e2b6d0c1649ab0d64a7a6e70bc77960f20f3fb0
--- /dev/null
+++ b/tasks/task_346_enterprise_industry_policy_analysis_medium_medium040.md
@@ -0,0 +1,117 @@
+---
+id: task_346_enterprise_industry_policy_analysis_medium_medium040
+name: enterprise_industry_policy_analysis-medium-medium040
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium040.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of local policies issued by the Guangzhou Development District Bureau of Economy and Information Technology in the local policies for the industry of Zhong Ke Ke Shu Ruan Jian Co., Ltd. and the number of local policies issued by the Yunnan Province General Office of the People's Government in the local policies for the Scientific Research and Technical Services industry?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_347_enterprise_industry_policy_analysis_medium_medium041.md b/tasks/task_347_enterprise_industry_policy_analysis_medium_medium041.md
new file mode 100644
index 0000000000000000000000000000000000000000..b773f288d8afd387b6c752607b853ceea558beb7
--- /dev/null
+++ b/tasks/task_347_enterprise_industry_policy_analysis_medium_medium041.md
@@ -0,0 +1,117 @@
+---
+id: task_347_enterprise_industry_policy_analysis_medium_medium041
+name: enterprise_industry_policy_analysis-medium-medium041
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium041.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of central ministry/agency policies issued by the Ministry of Public Security for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the number of local policies for the Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing industry in Anhui Province, which is greater?
+
+Output guidelines:
+The answer must be a company name or the word "industry"; output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Hui Jin Jin Rui Cai Fu Management Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_348_enterprise_industry_policy_analysis_medium_medium042.md b/tasks/task_348_enterprise_industry_policy_analysis_medium_medium042.md
new file mode 100644
index 0000000000000000000000000000000000000000..213b8d56aa36d84e943a5d0ed338e56012694733
--- /dev/null
+++ b/tasks/task_348_enterprise_industry_policy_analysis_medium_medium042.md
@@ -0,0 +1,117 @@
+---
+id: task_348_enterprise_industry_policy_analysis_medium_medium042
+name: enterprise_industry_policy_analysis-medium-medium042
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium042.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of central ministry/agency policies issued by the General Administration of Sport of China in the ministerial policies for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the total number of policies for the Railway, Ship, Aerospace and Other Transportation Equipment Manufacturing industry in Anhui Province, which is greater?
+
+Output guidelines:
+The answer must be a company name or the word "industry"; output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Hui Jin Jin Rui Cai Fu Management Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_349_enterprise_industry_policy_analysis_medium_medium043.md b/tasks/task_349_enterprise_industry_policy_analysis_medium_medium043.md
new file mode 100644
index 0000000000000000000000000000000000000000..7987e3111c14d62b9021472663898c175b960251
--- /dev/null
+++ b/tasks/task_349_enterprise_industry_policy_analysis_medium_medium043.md
@@ -0,0 +1,117 @@
+---
+id: task_349_enterprise_industry_policy_analysis_medium_medium043
+name: enterprise_industry_policy_analysis-medium-medium043
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium043.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of local policies issued by the Hunan Province People's Government in the local policies for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the number of policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry in Guangdong Province, which is greater?
+
+Output guidelines:
+The answer must be a company name or the word "industry"; output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Hui Jin Jin Rui Cai Fu Management Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_350_enterprise_industry_policy_analysis_medium_medium044.md b/tasks/task_350_enterprise_industry_policy_analysis_medium_medium044.md
new file mode 100644
index 0000000000000000000000000000000000000000..83b4f4d8baf3b108b5bbd76f1500c2a1d5306246
--- /dev/null
+++ b/tasks/task_350_enterprise_industry_policy_analysis_medium_medium044.md
@@ -0,0 +1,117 @@
+---
+id: task_350_enterprise_industry_policy_analysis_medium_medium044
+name: enterprise_industry_policy_analysis-medium-medium044
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium044.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of central ministry/agency policies issued by the State Administration of Foreign Exchange in the ministerial policies for the industry of Hui Jin Jin Rui Cai Fu Management Co., Ltd. and the number of local policies issued by the Shenzhen Municipality People's Government in the local policies for the Cultural, Arts, Sports and Entertainment Goods Manufacturing industry in Guangdong Province, which is greater?
+
+Output guidelines:
+The answer must be a company name or the word "industry"; output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Hui Jin Jin Rui Cai Fu Management Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_351_enterprise_industry_policy_analysis_medium_medium045.md b/tasks/task_351_enterprise_industry_policy_analysis_medium_medium045.md
new file mode 100644
index 0000000000000000000000000000000000000000..0e45537d9bff6fd9fc9a4c95b81c1882a5794860
--- /dev/null
+++ b/tasks/task_351_enterprise_industry_policy_analysis_medium_medium045.md
@@ -0,0 +1,117 @@
+---
+id: task_351_enterprise_industry_policy_analysis_medium_medium045
+name: enterprise_industry_policy_analysis-medium-medium045
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium045.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is the number of central ministry/agency policies issued by the Ministry of Housing and Urban-Rural Development in the ministerial policies for the industry of Zhao Ye Ze Jin Real Estate Holdings Co., Ltd. greater than the total number of policies for the Metal Smelting and Rolling Processing industry in Gansu Province?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_352_enterprise_industry_policy_analysis_medium_medium046.md b/tasks/task_352_enterprise_industry_policy_analysis_medium_medium046.md
new file mode 100644
index 0000000000000000000000000000000000000000..1038898c87fbce970a175f18fb83f8e2bb25b241
--- /dev/null
+++ b/tasks/task_352_enterprise_industry_policy_analysis_medium_medium046.md
@@ -0,0 +1,117 @@
+---
+id: task_352_enterprise_industry_policy_analysis_medium_medium046
+name: enterprise_industry_policy_analysis-medium-medium046
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium046.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Zhao Ye Ze Jin Di Chan Holdings Co., Ltd.industry's Ministry of Housing and Urban-Rural DevelopmentNumber of policies (indicator) and Gansu Province Development and Reform CommissionNumber of policies (indicator)what is the gap?
+
+Output guidelines:
+The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer"No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_353_enterprise_industry_policy_analysis_medium_medium047.md b/tasks/task_353_enterprise_industry_policy_analysis_medium_medium047.md
new file mode 100644
index 0000000000000000000000000000000000000000..6ec82ac183fadc598fe795925de293af48158cad
--- /dev/null
+++ b/tasks/task_353_enterprise_industry_policy_analysis_medium_medium047.md
@@ -0,0 +1,117 @@
+---
+id: task_353_enterprise_industry_policy_analysis_medium_medium047
+name: enterprise_industry_policy_analysis-medium-medium047
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium047.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of local policies issued by the Chengdu Municipality Bureau of Economy and Information Technology in the local policies for the industry of Hua Xin Yuan Shi New Materials Co., Ltd. and the number of local policies issued by the Gansu Province General Office of the People's Government in the local policies for the Pharmaceutical Manufacturing industry in Gansu Province, which is greater?
+
+Output guidelines:
+The answer must be a company name or the word "Equal"; output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_354_enterprise_industry_policy_analysis_medium_medium048.md b/tasks/task_354_enterprise_industry_policy_analysis_medium_medium048.md
new file mode 100644
index 0000000000000000000000000000000000000000..c2da90af046ebfb82c7b60f4104cd108092cb742
--- /dev/null
+++ b/tasks/task_354_enterprise_industry_policy_analysis_medium_medium048.md
@@ -0,0 +1,117 @@
+---
+id: task_354_enterprise_industry_policy_analysis_medium_medium048
+name: enterprise_industry_policy_analysis-medium-medium048
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium048.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of policies issued by the Shandong Province Department of Industry and Information Technology in the local policies for the industry of Hua Xin Yuan Shi Xin Cai Liao Co., Ltd. and the number of policies for the Pharmaceutical Manufacturing industry in Gansu Province, which is higher?
+
+Output guidelines:
+The answer must be a company name or the word "Equal"; output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_355_enterprise_industry_policy_analysis_medium_medium049.md b/tasks/task_355_enterprise_industry_policy_analysis_medium_medium049.md
new file mode 100644
index 0000000000000000000000000000000000000000..e9f85d3c09b23d492b44235c70c92180529e3e8d
--- /dev/null
+++ b/tasks/task_355_enterprise_industry_policy_analysis_medium_medium049.md
@@ -0,0 +1,117 @@
+---
+id: task_355_enterprise_industry_policy_analysis_medium_medium049
+name: enterprise_industry_policy_analysis-medium-medium049
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium049.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of local policies for the industry of Wan Hui Jin Sheng Real Estate Development Co., Ltd. and the number of local policies issued by the Jiangxi Province People's Government for the Commercial Electrical Machinery and Equipment Manufacturing industry in Jiangxi Province, which is greater?
+
+Output guidelines:
+The answer must be a company name or the word "industry"; output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Wan Hui Jin Sheng Real Estate Development Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_356_enterprise_industry_policy_analysis_medium_medium050.md b/tasks/task_356_enterprise_industry_policy_analysis_medium_medium050.md
new file mode 100644
index 0000000000000000000000000000000000000000..b1e17862ff60b4635e73e06c9048dcf49d91f3e9
--- /dev/null
+++ b/tasks/task_356_enterprise_industry_policy_analysis_medium_medium050.md
@@ -0,0 +1,117 @@
+---
+id: task_356_enterprise_industry_policy_analysis_medium_medium050
+name: enterprise_industry_policy_analysis-medium-medium050
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium050.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Wan Hui Jin Sheng Fang Di Chan Kai Fa Co., Ltd.industry's Shandong ProvinceNumber of policies (indicator) and Jiangxi Province Number of policies (indicator)how much？
+
+Output guidelines:
+The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer"No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_357_enterprise_industry_policy_analysis_medium_medium051.md b/tasks/task_357_enterprise_industry_policy_analysis_medium_medium051.md
new file mode 100644
index 0000000000000000000000000000000000000000..abdbe631916fab83c0bc5199f806fb2de4e71193
--- /dev/null
+++ b/tasks/task_357_enterprise_industry_policy_analysis_medium_medium051.md
@@ -0,0 +1,117 @@
+---
+id: task_357_enterprise_industry_policy_analysis_medium_medium051
+name: enterprise_industry_policy_analysis-medium-medium051
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium051.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Wu Li Hui Da Chain Co., Ltd.industry's ministerial policies_Number of policies (indicator) and China Shenzhen CitypersonsNumber of policies (indicator)compared with difference how much？
+
+Output guidelines:
+The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer"No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_358_enterprise_industry_policy_analysis_medium_medium052.md b/tasks/task_358_enterprise_industry_policy_analysis_medium_medium052.md
new file mode 100644
index 0000000000000000000000000000000000000000..0f26620fd8e84aef2a5ed95bae7ecbfdf7969dbd
--- /dev/null
+++ b/tasks/task_358_enterprise_industry_policy_analysis_medium_medium052.md
@@ -0,0 +1,117 @@
+---
+id: task_358_enterprise_industry_policy_analysis_medium_medium052
+name: enterprise_industry_policy_analysis-medium-medium052
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium052.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of local policies issued by the Sichuan Province People's Government in the local policies for the industry of Wu Li Hui Da Chain Co., Ltd. and the number of central ministry/agency policies in the China conglomerates category, which is greater?
+
+Output guidelines:
+The answer must be a company name or the word "industry"; output only the name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Wu Li Hui Da Chain Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_359_enterprise_industry_policy_analysis_medium_medium053.md b/tasks/task_359_enterprise_industry_policy_analysis_medium_medium053.md
new file mode 100644
index 0000000000000000000000000000000000000000..a8a404d91b0996797035d332670c4d52315b4427
--- /dev/null
+++ b/tasks/task_359_enterprise_industry_policy_analysis_medium_medium053.md
@@ -0,0 +1,117 @@
+---
+id: task_359_enterprise_industry_policy_analysis_medium_medium053
+name: enterprise_industry_policy_analysis-medium-medium053
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium053.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of policies issued by the State Administration of Foreign Exchange in the ministerial policies for the industry of Hua Ying Tai Sheng Cai Fu Management Co., Ltd. and the number of local policies issued by the Chongqing Municipality General Office of the People's Government in the local policies for the Financial Services industry in China, which is greater?
+
+Output guidelines:
+The answer must be "Equal", a company name, or the word "industry"; output only one word or company name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Chongqing Municipality General Office of the People's Government"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_360_enterprise_industry_policy_analysis_medium_medium054.md b/tasks/task_360_enterprise_industry_policy_analysis_medium_medium054.md
new file mode 100644
index 0000000000000000000000000000000000000000..3770a784e9a7fc355cee7ce5dc78d6370ecee0c1
--- /dev/null
+++ b/tasks/task_360_enterprise_industry_policy_analysis_medium_medium054.md
@@ -0,0 +1,117 @@
+---
+id: task_360_enterprise_industry_policy_analysis_medium_medium054
+name: enterprise_industry_policy_analysis-medium-medium054
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium054.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Hua Ying Tai Sheng Wealth Management Co., Ltd.industry's Shanghai Municipalitypersons Number of policies (indicator) and China Number of policies issued by Ministry of Educationcompared with which unitsgreater？
+
+Output guidelines:
+The answer must a company name or "industry""Equal", Output onlyname, without any explanation or description.Wu, answer"No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_361_enterprise_industry_policy_analysis_medium_medium055.md b/tasks/task_361_enterprise_industry_policy_analysis_medium_medium055.md
new file mode 100644
index 0000000000000000000000000000000000000000..4b60a0ddebc07c756a1c02104470b8db9c1be99a
--- /dev/null
+++ b/tasks/task_361_enterprise_industry_policy_analysis_medium_medium055.md
@@ -0,0 +1,117 @@
+---
+id: task_361_enterprise_industry_policy_analysis_medium_medium055
+name: enterprise_industry_policy_analysis-medium-medium055
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium055.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of policies issued by the Ministry of Public Security in the ministerial policies for the industry of Tong Tong Ze Hong Securities Co., Ltd. and the number of policies issued by the State Administration of Foreign Exchange in the ministerial policies for the Leasing and Business Services industry in China, which is greater?
+
+Output guidelines:
+The answer must be "Equal", a company name, or the word "industry"; output only one word or company name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_362_enterprise_industry_policy_analysis_medium_medium056.md b/tasks/task_362_enterprise_industry_policy_analysis_medium_medium056.md
new file mode 100644
index 0000000000000000000000000000000000000000..e8f16adf0ae7f559a61aefe8094a8de455f1fbc1
--- /dev/null
+++ b/tasks/task_362_enterprise_industry_policy_analysis_medium_medium056.md
@@ -0,0 +1,117 @@
+---
+id: task_362_enterprise_industry_policy_analysis_medium_medium056
+name: enterprise_industry_policy_analysis-medium-medium056
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium056.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Tong Tong Ze Hong Zheng Quan Co., Ltd.industry's local policiesGuangzhou CitypersonsNumber of policies (indicator) and China ministerial policiesNational Development and Reform CommissionNumber of policies (indicator)compared with difference how much？
+
+Output guidelines:
+The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer"No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_363_enterprise_industry_policy_analysis_medium_medium057.md b/tasks/task_363_enterprise_industry_policy_analysis_medium_medium057.md
new file mode 100644
index 0000000000000000000000000000000000000000..7d492e8b637f913de6266feb0f0919f00aa8c9d5
--- /dev/null
+++ b/tasks/task_363_enterprise_industry_policy_analysis_medium_medium057.md
@@ -0,0 +1,117 @@
+---
+id: task_363_enterprise_industry_policy_analysis_medium_medium057
+name: enterprise_industry_policy_analysis-medium-medium057
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium057.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Hua Ying Tai Sheng Wealth Management Co., Ltd.industry's Number of policies (indicator) and China Sichuan ProvinceNumber of policies (indicator)what is the gap?
+
+Output guidelines:
+The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer"No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`2.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_364_enterprise_industry_policy_analysis_medium_medium058.md b/tasks/task_364_enterprise_industry_policy_analysis_medium_medium058.md
new file mode 100644
index 0000000000000000000000000000000000000000..4d4773bbbd3b5ea9f2e686b52af09addde93779c
--- /dev/null
+++ b/tasks/task_364_enterprise_industry_policy_analysis_medium_medium058.md
@@ -0,0 +1,117 @@
+---
+id: task_364_enterprise_industry_policy_analysis_medium_medium058
+name: enterprise_industry_policy_analysis-medium-medium058
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium058.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Shi Yang Jin Jin Electrical Appliances Co., Ltd.province Guangdong ProvinceNumber of policies (indicator) and China Hainan ProvinceNumber of policies (indicator)compared with difference how much？
+
+Output guidelines:
+The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer"No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_365_enterprise_industry_policy_analysis_medium_medium059.md b/tasks/task_365_enterprise_industry_policy_analysis_medium_medium059.md
new file mode 100644
index 0000000000000000000000000000000000000000..eb0105262cb0036d68103425914495864ea58a73
--- /dev/null
+++ b/tasks/task_365_enterprise_industry_policy_analysis_medium_medium059.md
@@ -0,0 +1,117 @@
+---
+id: task_365_enterprise_industry_policy_analysis_medium_medium059
+name: enterprise_industry_policy_analysis-medium-medium059
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium059.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Between the number of policies issued by the Guangdong Province Development and Reform Commission in the local policies for the province where Shi Yang Jin Jin Electrical Appliances Co., Ltd. is located and the number of local policies issued by the Department of Digitalization and Future Industries in the Food and Beverage industry in China, which is greater?
+
+Output guidelines:
+The answer must be a company name, the word "industry", or "Equal"; output only one word or one company name, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Equal"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_366_enterprise_industry_policy_analysis_medium_medium060.md b/tasks/task_366_enterprise_industry_policy_analysis_medium_medium060.md
new file mode 100644
index 0000000000000000000000000000000000000000..32dcb956744f9a4c7337b640c05913e90bef2e34
--- /dev/null
+++ b/tasks/task_366_enterprise_industry_policy_analysis_medium_medium060.md
@@ -0,0 +1,117 @@
+---
+id: task_366_enterprise_industry_policy_analysis_medium_medium060
+name: enterprise_industry_policy_analysis-medium-medium060
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium060.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd.province Number of policies (indicator) and China  in Guangdong ProvinceNumber of policies (indicator)compared with which unitsgreater？
+
+Output guidelines:
+The answer must a company name or "industry", Output onlyname, without any explanation or description.Wu, answer"No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Jian Fan Ning Ze Yang Lao Fu Wu Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_367_enterprise_industry_policy_analysis_medium_medium061.md b/tasks/task_367_enterprise_industry_policy_analysis_medium_medium061.md
new file mode 100644
index 0000000000000000000000000000000000000000..92bd1ea8aac0ec6c2256a7747351f174f9ab2842
--- /dev/null
+++ b/tasks/task_367_enterprise_industry_policy_analysis_medium_medium061.md
@@ -0,0 +1,117 @@
+---
+id: task_367_enterprise_industry_policy_analysis_medium_medium061
+name: enterprise_industry_policy_analysis-medium-medium061
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium061.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Hua Cheng Sheng Yuan Integrated Development Co., Ltd.province Guangzhou CityHuang Pu DistrictNumber of policies (indicator) and China Number of policies (indicator)compared with which unitsgreater？
+
+Output guidelines:
+The answer must a company name or "industry", Output onlyname, without any explanation or description.Wu, answer"No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Hua Cheng Sheng Yuan Integrated Development Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_368_enterprise_industry_policy_analysis_medium_medium062.md b/tasks/task_368_enterprise_industry_policy_analysis_medium_medium062.md
new file mode 100644
index 0000000000000000000000000000000000000000..04016ea77a61fa896acdcddc54ebb61ebddc3fdb
--- /dev/null
+++ b/tasks/task_368_enterprise_industry_policy_analysis_medium_medium062.md
@@ -0,0 +1,117 @@
+---
+id: task_368_enterprise_industry_policy_analysis_medium_medium062
+name: enterprise_industry_policy_analysis-medium-medium062
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium062.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the difference between the number of local policies issued by the Guangdong Provincial Committee of the CPC in the local policies for the industry of Hua Cheng Sheng Yuan Integrated Development Co., Ltd. and the number of local policies issued by the Fujian Province Department of Industry and Information Technology in the local policies for the Semiconductor Industry in China?
+
+Output guidelines:
+The answer must be a number rounded to one decimal place. Output only the number, without units, commas, or explanations. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_369_enterprise_industry_policy_analysis_medium_medium063.md b/tasks/task_369_enterprise_industry_policy_analysis_medium_medium063.md
new file mode 100644
index 0000000000000000000000000000000000000000..39bf1ee8506a0c01242318c8a0a6e9accf274c2d
--- /dev/null
+++ b/tasks/task_369_enterprise_industry_policy_analysis_medium_medium063.md
@@ -0,0 +1,117 @@
+---
+id: task_369_enterprise_industry_policy_analysis_medium_medium063
+name: enterprise_industry_policy_analysis-medium-medium063
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium063.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd.province Number of policies (indicator) and China TransportationNumber of policies (indicator)compared with  how much？
+
+Output guidelines:
+The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer"No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_370_enterprise_industry_policy_analysis_medium_medium064.md b/tasks/task_370_enterprise_industry_policy_analysis_medium_medium064.md
new file mode 100644
index 0000000000000000000000000000000000000000..c8c8278b5b99aac516b28e3528ae089696061c3a
--- /dev/null
+++ b/tasks/task_370_enterprise_industry_policy_analysis_medium_medium064.md
@@ -0,0 +1,117 @@
+---
+id: task_370_enterprise_industry_policy_analysis_medium_medium064
+name: enterprise_industry_policy_analysis-medium-medium064
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium064.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Is the number of specific local policies in the province where Bi Yuan Zhi Ze Cheng Shi Development Co., Ltd. is located the same as the number of policies issued by the Hainan Province Department of Industry and Information Technology for China's rubber and plastic products industry local policies?
+
+Output guidelines:
+The answer must be "Yes" or "No"; output only one word, without any explanation or description. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Yes"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_371_enterprise_industry_policy_analysis_medium_medium065.md b/tasks/task_371_enterprise_industry_policy_analysis_medium_medium065.md
new file mode 100644
index 0000000000000000000000000000000000000000..86f3a434c00481f374b9ebad90d864a137ec57f2
--- /dev/null
+++ b/tasks/task_371_enterprise_industry_policy_analysis_medium_medium065.md
@@ -0,0 +1,117 @@
+---
+id: task_371_enterprise_industry_policy_analysis_medium_medium065
+name: enterprise_industry_policy_analysis-medium-medium065
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium065.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Zhong Ke Ke Shu Software Co., Ltd.province  Number of policies (indicator) and China Sichuan Provincepersons Number of policies (indicator)what is the gap?
+
+Output guidelines:
+The answer must units, rounded to one decimal place.Output only, without units, commas, or explanations.Wu, answer"No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_372_enterprise_industry_policy_analysis_medium_medium066.md b/tasks/task_372_enterprise_industry_policy_analysis_medium_medium066.md
new file mode 100644
index 0000000000000000000000000000000000000000..eb4248fae1b90a2661282adc56ee0fe1813e59d7
--- /dev/null
+++ b/tasks/task_372_enterprise_industry_policy_analysis_medium_medium066.md
@@ -0,0 +1,117 @@
+---
+id: task_372_enterprise_industry_policy_analysis_medium_medium066
+name: enterprise_industry_policy_analysis-medium-medium066
+category: enterprise_industry_policy_analysis
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/enterprise_industry_policy_analysis/medium066.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Zhong Ke Ke Shu Software Co., Ltd.province Guangzhou CitypersonsNumber of policies (indicator) and China Hunan ProvincepersonsNumber of policies (indicator)compared with which unitsgreater？
+
+Output guidelines:
+The answer must a company name or "industry", Output onlyname, without any explanation or description.Wu, answer"No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Zhong Ke Ke Shu Software Co., Ltd."`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_373_hypothesis_verification_hard_hard001.md b/tasks/task_373_hypothesis_verification_hard_hard001.md
new file mode 100644
index 0000000000000000000000000000000000000000..e1d258ef5d32ab20a49c244e802092db153cb243
--- /dev/null
+++ b/tasks/task_373_hypothesis_verification_hard_hard001.md
@@ -0,0 +1,119 @@
+---
+id: task_373_hypothesis_verification_hard_hard001
+name: hypothesis_verification-hard-hard001
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard001.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+作为研发驱动型产业，医药制造业的企业创新投入与收入表现之间的关联一直备受研究者关注，而地方政策环境可能影响这种关联的强弱。请以2022年数据为基础，将医药制造业上市企业按所在省份是否出台了地方生物医药产业发展促进政策(政策名称含生物医药，且含发展或促进)分为两组，分别计算两组企业的研发投入占比与营业收入同比增减幅之间的斯皮尔曼等级相关系数，并报告两个相关系数的差值（有政策省份系数减去无政策省份系数）。
+
+Output guidelines:
+依次回答出台政策省份的相关系数和未出台政策省份的相关系数及两者差值。相关系数保留4位小数，差值保留2位小数。如["0.2356", "-0.1048", "0.34"]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[0.1142, -0.2132, 0.33]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_374_hypothesis_verification_hard_hard002.md b/tasks/task_374_hypothesis_verification_hard_hard002.md
new file mode 100644
index 0000000000000000000000000000000000000000..98a30997983f618b175029be2f0c7d369ee36a67
--- /dev/null
+++ b/tasks/task_374_hypothesis_verification_hard_hard002.md
@@ -0,0 +1,119 @@
+---
+id: task_374_hypothesis_verification_hard_hard002
+name: hypothesis_verification-hard-hard002
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard002.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+有观点认为，在推动消费电子及电气业数字化转型与智能制造升级方面，国家层面的产业政策与地方政府政策在表述深度和覆盖方向上存在系统性差异。请检验这一判断：针对所有与消费电子及电气业相关的政策文件，分别统计国家级政策（含国务院及各部委发布的政策）与地方级政策中，明确提出数字化转型或智能制造相关目标或措施的政策数量及其占各自总数的比例，并给出国家级覆盖率减去地方级覆盖率的差值（以百分点计）。
+
+Output guidelines:
+依次回答国家级政策覆盖率、地方级政策覆盖率、差值（国家级占比减去地方级占比）。覆盖率和差值均以百分点表示，保留2位小数。如[80.00, 71.43, 8.57]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[75.0, 66.67, 8.33]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_375_hypothesis_verification_hard_hard003.md b/tasks/task_375_hypothesis_verification_hard_hard003.md
new file mode 100644
index 0000000000000000000000000000000000000000..0a57081d7610c39e3374cc66d54c8283257fa251
--- /dev/null
+++ b/tasks/task_375_hypothesis_verification_hard_hard003.md
@@ -0,0 +1,119 @@
+---
+id: task_375_hypothesis_verification_hard_hard003
+name: hypothesis_verification-hard-hard003
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard003.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+在2022年通用设备制造业上市企业中，计算有地方制造业创新与科技发展促进政策支撑的省份（有政策省份）与无政策省份的企业政府补贴金额与年度中国发明专利申请数之间的斯皮尔曼等级相关系数，并给出两组系数之差（有政策省份系数减无政策省份系数）。要求依次回答：有政策省份的相关系数、无政策省份的相关系数、两组系数之差（均保留4位小数）。
+
+Output guidelines:
+依次回答有政策省份的相关系数、无政策省份的相关系数和两者差值（有政策-无政策）。相关系数和差值均保留4位小数。如["0.6812", "0.3590", "0.3222"]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[0.7531, 0.4245, 0.3285]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_376_hypothesis_verification_hard_hard004.md b/tasks/task_376_hypothesis_verification_hard_hard004.md
new file mode 100644
index 0000000000000000000000000000000000000000..e2121ce330f94f063f199557725f0e88cb73f60e
--- /dev/null
+++ b/tasks/task_376_hypothesis_verification_hard_hard004.md
@@ -0,0 +1,119 @@
+---
+id: task_376_hypothesis_verification_hard_hard004
+name: hypothesis_verification-hard-hard004
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard004.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+在2022年纺织鞋服业上市企业中对比政策支持对企业盈亏的影响，计算有地方制造业转型升级政策支撑的省份（有政策省份）与无政策省份的企业营业利润亏损占比，并给出两组占比之差（有政策省份占比减无政策省份占比，以百分点计）。要求依次回答：有政策省份亏损占比、无政策省份亏损占比、两组占比的差值（均保留2位小数）。
+
+Output guidelines:
+依次回答有政策省份亏损占比、无政策省份亏损占比和差值（有政策组占比减去无政策组占比）。占比以百分数表示，保留2位小数。如[38.46, 27.50, 10.96]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[43.75, 31.96, 11.79]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_377_hypothesis_verification_hard_hard005.md b/tasks/task_377_hypothesis_verification_hard_hard005.md
new file mode 100644
index 0000000000000000000000000000000000000000..e603598fe3d19f8ae34b041b732c65a5d83b218b
--- /dev/null
+++ b/tasks/task_377_hypothesis_verification_hard_hard005.md
@@ -0,0 +1,119 @@
+---
+id: task_377_hypothesis_verification_hard_hard005
+name: hypothesis_verification-hard-hard005
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard005.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+在化学原料和化学制品制造业中，碳达峰与节能减排政策的推进可能给部分企业带来额外合规成本，由此引发一种反常现象：企业获得的政府补贴较高（以全体有效企业政府补贴金额的中位数作为划定高补贴的分界点），但利润同比却在下滑。本题以2022年度该行业的上市企业为分析对象。请分别计算出台了地方碳达峰或节能减排促进政策的省份、以及未出台此类政策的省份中，反常企业占各组有效企业总数的比例，以及两组比例之差（有政策省份减无政策省份，以百分点计）。
+
+Output guidelines:
+依次回答有政策省份的反常企业占比、无政策省份的反常企业占比和两组占比的差值。占比和差值均以百分数表示，保留2位小数。如[32.14, 21.05, 11.09]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["29.29", "24.43", "4.87"]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_378_hypothesis_verification_hard_hard006.md b/tasks/task_378_hypothesis_verification_hard_hard006.md
new file mode 100644
index 0000000000000000000000000000000000000000..74520a4d60652459a9278067a6e3fde6632d24dc
--- /dev/null
+++ b/tasks/task_378_hypothesis_verification_hard_hard006.md
@@ -0,0 +1,119 @@
+---
+id: task_378_hypothesis_verification_hard_hard006
+name: hypothesis_verification-hard-hard006
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard006.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+在2022年铁路、船舶、航空航天和其他运输设备制造业上市企业中，计算有地方先进制造与装备产业促进政策支撑的省份（有政策省份）与无政策省份的企业总资产与累计中国发明专利授权数之间的斯皮尔曼等级相关系数，并给出两组系数之差（有政策省份系数减无政策省份系数，保留2位小数）。要求依次回答：有政策省份的相关系数、无政策省份的相关系数、两组系数的差值。
+
+Output guidelines:
+依次回答有政策省份的相关系数、无政策省份的相关系数和两者差值（有政策-无政策）。相关系数保留2位小数。如["0.63", "0.48", "0.15"]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["0.56", "0.72", "-0.16"]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_379_hypothesis_verification_hard_hard007.md b/tasks/task_379_hypothesis_verification_hard_hard007.md
new file mode 100644
index 0000000000000000000000000000000000000000..13cad430aa6daa37fa92e5fb77a3edb2f53f93b2
--- /dev/null
+++ b/tasks/task_379_hypothesis_verification_hard_hard007.md
@@ -0,0 +1,119 @@
+---
+id: task_379_hypothesis_verification_hard_hard007
+name: hypothesis_verification-hard-hard007
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard007.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+在2022年非金属矿物制品业上市企业中研究政策对不同规模企业营业利润率的影响，计算已出台专项推动建材行业碳达峰或节能减排政策的省份（有政策省份）中大型企业与小型企业的平均营业利润率差距（有政策省份规模差距），以及未出台此类政策的省份（无政策省份）中同一差距（无政策省份规模差距），并计算两者的差值（有政策省份规模差距减无政策省份规模差距）。要求依次回答：有政策省份规模差距、无政策省份规模差距、两类省份差距之差（均以百分点表示，保留2位小数）。
+
+Output guidelines:
+依次回答有政策省份规模差距、无政策省份规模差距、两类省份差距之差。均以百分点表示，保留2位小数。如[5.46, 1.23, 4.23]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[7.72, 0.69, 7.03]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_380_hypothesis_verification_hard_hard008.md b/tasks/task_380_hypothesis_verification_hard_hard008.md
new file mode 100644
index 0000000000000000000000000000000000000000..2b9d9afdead3a63abd58650d297b9430ac4ca9d1
--- /dev/null
+++ b/tasks/task_380_hypothesis_verification_hard_hard008.md
@@ -0,0 +1,119 @@
+---
+id: task_380_hypothesis_verification_hard_hard008
+name: hypothesis_verification-hard-hard008
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard008.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，集成电路产业的地方政策竞争进入白热化阶段，各省在专项激励力度上差异显著。本题统计口径说明如下：①统计对象为营业利润与营业收入数据均完整且营业收入非零的内地企业，港澳台地区企业不纳入；②营业利润率 = 营业利润 ÷ 营业收入 × 100%；③认定为专项集成电路产业促进政策，须是专门针对集成电路或半导体产业的地方政策，且明确包含流片补贴、企业落户奖励、研发设计人才支持、产业规模发展目标等专项措施中的至少一项，仅泛提数字经济或科技创新的通用政策不符合要求。在此基础上，请计算2022年半导体业中出台了上述专项政策的省份与未出台省份的企业平均营业利润率，并给出差值（有政策省份均值减去无政策省份均值，以百分点计）。
+
+Output guidelines:
+依次回答有政策省份平均营业利润率、无政策省份平均营业利润率、两者差值（有政策-无政策）。数值均保留2位小数，以百分点表示。如[10.55, 13.87, -3.32]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[7.02, 16.33, -9.31]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_381_hypothesis_verification_hard_hard009.md b/tasks/task_381_hypothesis_verification_hard_hard009.md
new file mode 100644
index 0000000000000000000000000000000000000000..4244c1da381a364859bcc6e488d3cd4b545d13ec
--- /dev/null
+++ b/tasks/task_381_hypothesis_verification_hard_hard009.md
@@ -0,0 +1,119 @@
+---
+id: task_381_hypothesis_verification_hard_hard009
+name: hypothesis_verification-hard-hard009
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard009.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+在2022年汽车制造业上市企业中验证政策对企业营业利润有更好的促进作用的假设，计算出台了新能源汽车产业专项促进政策省份（有政策省份）的企业营业利润同比增减幅中位数，未出台此类政策省份（无政策省份）的企业营业利润同比增减幅中位数，以及两者的差值（有政策省份中位数减无政策省份中位数，以百分点计）。要求依次回答：有政策省份企业中位数、无政策省份企业中位数、两组中位数之差（均保留2位小数，单位为百分点）。
+
+Output guidelines:
+依次回答有政策省份企业中位数、无政策省份企业中位数和差值。数值均保留2位小数，单位为百分点（%）。差值=有政策省份中位数-无政策省份中位数。如[4.16, -3.80, 7.96]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[2.73, -5.52, 8.25]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_382_hypothesis_verification_hard_hard010.md b/tasks/task_382_hypothesis_verification_hard_hard010.md
new file mode 100644
index 0000000000000000000000000000000000000000..9528c2bcd9cc82ddfc8312254dddf2c119715936
--- /dev/null
+++ b/tasks/task_382_hypothesis_verification_hard_hard010.md
@@ -0,0 +1,119 @@
+---
+id: task_382_hypothesis_verification_hard_hard010
+name: hypothesis_verification-hard-hard010
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard010.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+在2022年金属冶炼和压延加工业上市企业中验证政策对民因企业有更好的资产收益率假设，计算有专项推动有色金属冶炼产业高质量发展政策省份（有政策省份）的所有制效率差距（国有均值减民营均值）、无政策省份的所有制效率差距，以及两类省份差距之差（有政策省份差距减无政策省份差距，以百分点计）。要求依次回答：有政策省份的所有制效率差距、无政策省份的所有制效率差距、两类省份差距之差（均以百分点表示，保留2位小数）。
+
+Output guidelines:
+依次回答有政策省份的所有制效率差距、无政策省份的所有制效率差距、两类省份差距之差，均以百分点表示，保留2位小数。如[-7.52, -2.10, -5.42]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[-9.89, -3.28, -6.61]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_383_hypothesis_verification_hard_hard011.md b/tasks/task_383_hypothesis_verification_hard_hard011.md
new file mode 100644
index 0000000000000000000000000000000000000000..c0a997dc463411499c713d0dfd46e63b7b94d655
--- /dev/null
+++ b/tasks/task_383_hypothesis_verification_hard_hard011.md
@@ -0,0 +1,119 @@
+---
+id: task_383_hypothesis_verification_hard_hard011
+name: hypothesis_verification-hard-hard011
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard011.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+在2022年专用设备制造业上市企业中验证政策对头头部企业影响更大的假设，计算有地方重大技术装备专项促进政策或先进制造业专项法规省份（有政策省份）的CR20%、无政策省份的CR20%，以及两者的差值（有政策省份CR20%减无政策省份CR20%，以百分点计）。要求依次回答：有政策省份CR20%、无政策省份CR20%、差值（均保留2位小数，单位为百分点）。
+
+Output guidelines:
+依次回答有政策省份CR20%、无政策省份CR20%、差值。数值保留2位小数，以百分点表示。如[82.35, 70.14, 12.21]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[79.88, 73.71, 6.17]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_384_hypothesis_verification_hard_hard012.md b/tasks/task_384_hypothesis_verification_hard_hard012.md
new file mode 100644
index 0000000000000000000000000000000000000000..4140c7cc20a73ae53aefc97f16b0909f80f77b4d
--- /dev/null
+++ b/tasks/task_384_hypothesis_verification_hard_hard012.md
@@ -0,0 +1,119 @@
+---
+id: task_384_hypothesis_verification_hard_hard012
+name: hypothesis_verification-hard-hard012
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard012.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+一个行业内营业利润率的分布离散程度，在一定条件下可反映企业间的竞争分化态势或政策的结构性影响效应。以四分位距（IQR = Q3减去Q1，均基于企业个体营业利润率的分布计算）作为离散程度的测量工具，对2022年橡胶和塑料制品业（排除港澳台）数据进行分析，营业利润率=营业利润金额/营业收入金额×100%。若以出台了面向制造业企业、按发展阶段设定梯度化现金奖励或培育支持机制且将橡胶和塑料制品业列为受益行业的专项产业培育激励政策的省份为一组，其余省份为另一组，两组企业营业利润率的IQR分别是多少个百分点？两组IQR的差值（有政策省份减无政策省份）为多少个百分点？
+
+Output guidelines:
+依次回答有政策省份的IQR、无政策省份的IQR、两者差值。数值以百分点表示，保留2位小数。如[5.82, 9.45, -3.63]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[4.37, 11.59, -7.23]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_385_hypothesis_verification_hard_hard013.md b/tasks/task_385_hypothesis_verification_hard_hard013.md
new file mode 100644
index 0000000000000000000000000000000000000000..d8d32430788509b54c6137fe4353662804a94d49
--- /dev/null
+++ b/tasks/task_385_hypothesis_verification_hard_hard013.md
@@ -0,0 +1,119 @@
+---
+id: task_385_hypothesis_verification_hard_hard013
+name: hypothesis_verification-hard-hard013
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard013.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+国家与地方政策的协同效应，在产业经济学中通常以政策叠加框架加以讨论——双重政策覆盖的企业是否在劳动效率上具有系统性优势，是检验政策层级互补性的核心命题之一。人均营业收入作为劳动效率的代理变量，以2022年食品饮料业为例，将企业所在省份按政策覆盖状态分为三组：第一组，省份同时被国家消费品工业促进政策明确列为活动实施省份且已出台地方食品相关产业政策；第二组，省份仅有地方食品相关产业政策、未被前述国家政策明确覆盖；第三组，上述两类政策均无。有效企业须营业收入金额非空且雇员总数为正的内地企业（排除港澳台）。三组企业的平均人均营业收入分别为多少万元？第一组与第二组的均值差为多少万元？
+
+Output guidelines:
+依次回答双重覆盖组、仅地方覆盖组、无政策覆盖组的平均人均营业收入（万元/人），以及双重覆盖组与仅地方覆盖组的差值（万元/人）。所有数值保留2位小数。如[185.30, 152.75, 126.40, 32.55]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[201.94, 168.2, 140.96, 33.74]`
+
+Scoring rules:
+- The gold answer is a list with N=4 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_3` each as 0 or 1.
+- Return `total = (sum(part_i)) / 4` exactly.
+- If the model output is missing or cannot be parsed into 4 comparable parts, score all parts 0.
+
diff --git a/tasks/task_386_hypothesis_verification_hard_hard014.md b/tasks/task_386_hypothesis_verification_hard_hard014.md
new file mode 100644
index 0000000000000000000000000000000000000000..6a5486f751527e338a2de342fa70c0dac1c3d0bf
--- /dev/null
+++ b/tasks/task_386_hypothesis_verification_hard_hard014.md
@@ -0,0 +1,119 @@
+---
+id: task_386_hypothesis_verification_hard_hard014
+name: hypothesis_verification-hard-hard014
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard014.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+针对验证政府补贴能高效地转化为了企业的创新产出对于不同企业存在差异的假设。以通信传输设备业为例，这一转化效率在有无专项研发创新激励政策的省份之间是否存在差异。认定的专项扶持企业研发创新的地方促进政策须同时满足：①政策涵盖通信传输设备业；②政策明确包含金额确定的企业研发奖励、科技创新补贴或专项创新资金等直接企业激励措施。有效企业指政府补贴金额大于0且年度中国发明专利申请数有完整记录的内地企业。请以每百万元政府补贴所对应的年度发明专利申请数作为衡量补贴转化效率的指标，分别计算并给出2022年有政策省份与无政策省份中有效企业的该指标均值，并给出差值（有政策省份均值减去无政策省份均值）。
+
+Output guidelines:
+依次回答有政策省份的补贴转化效率、无政策省份的补贴转化效率、两者差值。均保留2位小数，单位为件/百万元。如[3.05, 2.47, 0.58]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[2.38, 3.18, -0.81]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_387_hypothesis_verification_hard_hard015.md b/tasks/task_387_hypothesis_verification_hard_hard015.md
new file mode 100644
index 0000000000000000000000000000000000000000..3be211fa2fceed9685f8bd998b708bb2db1a9623
--- /dev/null
+++ b/tasks/task_387_hypothesis_verification_hard_hard015.md
@@ -0,0 +1,119 @@
+---
+id: task_387_hypothesis_verification_hard_hard015
+name: hypothesis_verification-hard-hard015
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/hard015.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+政策文本的表述精度，即政策目标是以可量化的具体数值呈现，还是以方向性、原则性的定性语言为主，可能反映政府的政策执行意志，进而影响辖区内企业的研发行为。在仪器仪表制造业中，对已出台涉及本行业地方政策的省份进行内容分析：凡政策正文中含有具体产业发展数值目标（如产业规模达X亿元、增长X%、新建X家/X座等可核查数值）的省份，归为量化目标组；政策正文仅涵盖定性方向、原则性要求或门槛条件而不包含上述产业发展数值目标的省份，归为定性目标组。基于2022年数据，请分别计算两组省份内仪器仪表制造业有效企业（研发投入占比数据非空且数值在0%到100%之间）的研发投入占比均值，并给出量化目标组均值减去定性目标组均值的差值（单位：百分点）。
+
+Output guidelines:
+依次回答含量化目标省份的研发投入占比均值、仅含定性目标省份的研发投入占比均值、两者差值（量化组均值减去定性组均值）。均值和差值均保留2位小数，以百分点表示。如[12.53, 8.21, 4.32]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[14.07, 9.67, 4.4]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_388_hypothesis_verification_medium_medium001.md b/tasks/task_388_hypothesis_verification_medium_medium001.md
new file mode 100644
index 0000000000000000000000000000000000000000..15af7c1da3a063ba2cd7a325fb31307bbae70768
--- /dev/null
+++ b/tasks/task_388_hypothesis_verification_medium_medium001.md
@@ -0,0 +1,117 @@
+---
+id: task_388_hypothesis_verification_medium_medium001
+name: hypothesis_verification-medium-medium001
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/medium001.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Using 2022 data, to verify the hypothesis that policy has a positive effect on corporate R&D: among listed semiconductor firms, divide them by whether their registered province has ever issued a local industrial policy whose name or covered-industry field contains "semiconductor" or "integrated circuit" (either condition suffices). What is the difference in mean R&D investment ratio between the policy-covered group and the non-covered group (difference = policy-province group mean minus non-policy-province group mean), in percentage points?
+
+Output guidelines:
+Answer format: a numeric value (two decimal places, in percentage points). A positive value means the policy-province group is higher; a negative value means the non-policy-province group is higher. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-0.73`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_389_hypothesis_verification_medium_medium002.md b/tasks/task_389_hypothesis_verification_medium_medium002.md
new file mode 100644
index 0000000000000000000000000000000000000000..038ea984bddc6c80aeab8ced9f27c0ca731a60db
--- /dev/null
+++ b/tasks/task_389_hypothesis_verification_medium_medium002.md
@@ -0,0 +1,117 @@
+---
+id: task_389_hypothesis_verification_medium_medium002
+name: hypothesis_verification-medium-medium002
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/medium002.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in the pharmaceutical manufacturing industry, what is the difference in Pearson correlation coefficient between private enterprises and state-owned enterprises(State-owned enterprises include central SOEs, local SOEs, SOEs under research institutes, and other SOEs.) with respect to R&D investment amount and cumulative number of granted Chinese invention patents?
+
+Output guidelines:
+Answer format: the difference value (four decimal places). Difference = private enterprise correlation − state-owned enterprise correlation. A positive value means private enterprises have stronger correlation. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-0.1452`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_390_hypothesis_verification_medium_medium003.md b/tasks/task_390_hypothesis_verification_medium_medium003.md
new file mode 100644
index 0000000000000000000000000000000000000000..ce924f077ec28506e746dc3e27d28c37926a2090
--- /dev/null
+++ b/tasks/task_390_hypothesis_verification_medium_medium003.md
@@ -0,0 +1,117 @@
+---
+id: task_390_hypothesis_verification_medium_medium003
+name: hypothesis_verification-medium-medium003
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/medium003.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+To verify the hypothesis that policy density has a positive effect on average corporate profitability, using 2022 automotive manufacturing as an example, calculate the Spearman rank correlation coefficient between each province's policy density indicator (number of policies whose industry field contains "automotive") and average profitability (total operating profit amount / total operating revenue amount × 100%). Provinces without automotive manufacturing operating data or with zero operating revenue are excluded from the calculation; valid provinces with no policy records are assigned a policy count of 0.
+
+Output guidelines:
+Answer format: numeric value (rounded to 4 decimal places). If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-0.0023`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_391_hypothesis_verification_medium_medium004.md b/tasks/task_391_hypothesis_verification_medium_medium004.md
new file mode 100644
index 0000000000000000000000000000000000000000..c755f788db362dc64800e2027b82e34a6767b728
--- /dev/null
+++ b/tasks/task_391_hypothesis_verification_medium_medium004.md
@@ -0,0 +1,117 @@
+---
+id: task_391_hypothesis_verification_medium_medium004
+name: hypothesis_verification-medium-medium004
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/medium004.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, to verify the hypothesis on the effect of debt ratio on R&D investment ratio in Raw Chemical Materials and Chemical Products Manufacturing, calculate the difference in the mean R&D investment ratio (in percentage points) between the high-debt group (asset-liability ratio above the national industry median) and the low-debt group (asset-liability ratio below the national industry median).
+
+Output guidelines:
+Answer format: numeric value (rounded to 2 decimal places, unit: percentage points). A positive value indicates the high-debt group is higher. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-0.61`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_392_hypothesis_verification_medium_medium005.md b/tasks/task_392_hypothesis_verification_medium_medium005.md
new file mode 100644
index 0000000000000000000000000000000000000000..d576694e03ba49520e9ba7d491d8ce57353f7fab
--- /dev/null
+++ b/tasks/task_392_hypothesis_verification_medium_medium005.md
@@ -0,0 +1,117 @@
+---
+id: task_392_hypothesis_verification_medium_medium005
+name: hypothesis_verification-medium-medium005
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/medium005.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, to verify the hypothesis that there is a clear relationship between enterprise total assets and invention patent count in the Consumer Electronics and Electrical Industry, we focus only on provinces with high R&D density (provinces where the mean R&D investment ratio in provincial industry aggregate data exceeds the corresponding national industry aggregate mean). Among all listed enterprises in this industry within these provinces, what is the Pearson correlation coefficient between enterprise total asset scale and annual China invention patent applications?
+
+Output guidelines:
+Answer format: numeric value (rounded to 4 decimal places). If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.8294`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_393_hypothesis_verification_medium_medium006.md b/tasks/task_393_hypothesis_verification_medium_medium006.md
new file mode 100644
index 0000000000000000000000000000000000000000..31a624344eb5f60fadfde1ddd47f246101ce729a
--- /dev/null
+++ b/tasks/task_393_hypothesis_verification_medium_medium006.md
@@ -0,0 +1,117 @@
+---
+id: task_393_hypothesis_verification_medium_medium006
+name: hypothesis_verification-medium-medium006
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/medium006.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, there was a view that government subsidies and revenue are clearly correlated among large-asset-scale enterprises in the food and beverage industry. Therefore, researchers sampled the top one-third of large enterprises ranked (rounded down) by total assets as the research subject. What is the Pearson correlation coefficient between government subsidy amount and year-over-year revenue growth rate?
+
+Output guidelines:
+Answer format: numerical value (retain 4 decimal places). If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`-0.1237`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_394_hypothesis_verification_medium_medium007.md b/tasks/task_394_hypothesis_verification_medium_medium007.md
new file mode 100644
index 0000000000000000000000000000000000000000..72c88379f8f4bedc22d36099ec7ebc400311b667
--- /dev/null
+++ b/tasks/task_394_hypothesis_verification_medium_medium007.md
@@ -0,0 +1,117 @@
+---
+id: task_394_hypothesis_verification_medium_medium007
+name: hypothesis_verification-medium-medium007
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/medium007.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In the 2022 data, to verify the hypothesis that private enterprises in the communications transmission equipment industry benefit more from policies than state-owned enterprises, we focus on communications transmission equipment enterprises in provinces that have issued local policies involving the "communications"-related industry. Enterprises are grouped by ownership type (private enterprises vs. state-owned enterprises, where state-owned enterprises include only central state-owned enterprises and local state-owned enterprises). Within each group, the per capita revenue (total revenue / total number of employees) is calculated using the weighted consolidation method. Finally, return the specific difference between per capita revenue of the private enterprise group and that of the state-owned enterprise group.
+
+Output guidelines:
+Answer format: numerical value (retain 2 decimal places, unit: yuan/person). A positive value indicates private enterprises are higher. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`245418.07`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_395_hypothesis_verification_medium_medium008.md b/tasks/task_395_hypothesis_verification_medium_medium008.md
new file mode 100644
index 0000000000000000000000000000000000000000..6b70a34db0d3b7605e41d1e42284e50d1061b017
--- /dev/null
+++ b/tasks/task_395_hypothesis_verification_medium_medium008.md
@@ -0,0 +1,117 @@
+---
+id: task_395_hypothesis_verification_medium_medium008
+name: hypothesis_verification-medium-medium008
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/medium008.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Using the specialized equipment manufacturing industry in 2022 as the research subject, some researchers believe that enterprise size may confound the relationship between listing history length and cumulative patent accumulation. To test this hypothesis: First, among all valid enterprises (with non-null cumulative China patent applications), calculate the Pearson correlation coefficient r1 between listing years (derived by subtracting listing year from 2022) and cumulative China patent applications; second, restrict the sample to the large enterprise subset whose total assets exceed the industry median total assets in the national industry aggregate data, then calculate the Pearson correlation coefficient r2 for the same pair of variables; finally, report the specific value of the difference r2 − r1.
+
+Output guidelines:
+Answer format: the difference between the two correlation coefficients (retain 4 decimal places). Difference = large enterprise correlation coefficient − all enterprises correlation coefficient. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0.0093`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_396_hypothesis_verification_medium_medium009.md b/tasks/task_396_hypothesis_verification_medium_medium009.md
new file mode 100644
index 0000000000000000000000000000000000000000..b7c064af8fbda2aec15e6c8a16c0b897d47ac877
--- /dev/null
+++ b/tasks/task_396_hypothesis_verification_medium_medium009.md
@@ -0,0 +1,117 @@
+---
+id: task_396_hypothesis_verification_medium_medium009
+name: hypothesis_verification-medium-medium009
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/medium009.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In the pharmaceutical manufacturing industry in 2022, verify whether the 'policy-innovation paradox' exists (i.e., the phenomenon where the strength of local pharmaceutical innovation support policies is negatively associated with innovation output). Among provinces that have issued pharmaceutical support policies, use the median number of policy entries as the threshold for support strength, and use the national average invention patent grants per province for pharmaceutical manufacturing as the innovation output benchmark. Count the number of provinces where policy support strength exceeds the median but total invention patent grants are below the national average.
+
+Output guidelines:
+Answer format: numerical value (integer). If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`3`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_397_hypothesis_verification_medium_medium010.md b/tasks/task_397_hypothesis_verification_medium_medium010.md
new file mode 100644
index 0000000000000000000000000000000000000000..17231df3dac8879029607b6443e02c372ea28783
--- /dev/null
+++ b/tasks/task_397_hypothesis_verification_medium_medium010.md
@@ -0,0 +1,117 @@
+---
+id: task_397_hypothesis_verification_medium_medium010
+name: hypothesis_verification-medium-medium010
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/medium010.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, based on the 'high input low output' hypothesis in Romer's endogenous growth theory, verify whether state-owned enterprises in the semiconductor industry exhibit the dual anomaly of 'large asset scale but low operating profit margin' and 'high R&D investment but low patent conversion efficiency'. The hypothesis states: when state-owned enterprises rank high in the industry in both asset scale (total assets) and R&D investment (amount), their operating profit margin (operating profit / revenue) and patent conversion efficiency (cumulative China invention patent grants / R&D investment × 100 million) should rank low in the industry. Please count the number of enterprises that simultaneously satisfy the following conditions (state-owned enterprises include central state-owned enterprises, local state-owned enterprises, state-owned enterprises (research institutes), and state-owned enterprises (other)): ① total assets > median total assets of industry-wide state-owned enterprises; ② operating profit margin < median operating profit margin of industry-wide state-owned enterprises; ③ R&D investment amount > median R&D investment amount of industry-wide state-owned enterprises; ④ patent conversion efficiency < median patent conversion efficiency of industry-wide state-owned enterprises.
+
+Output guidelines:
+Answer format: numerical value (integer, unit: enterprises). If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`3`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_398_hypothesis_verification_medium_medium011.md b/tasks/task_398_hypothesis_verification_medium_medium011.md
new file mode 100644
index 0000000000000000000000000000000000000000..5ae9688cb9c9a64fd0006b7f0a0ade4ec997e627
--- /dev/null
+++ b/tasks/task_398_hypothesis_verification_medium_medium011.md
@@ -0,0 +1,117 @@
+---
+id: task_398_hypothesis_verification_medium_medium011
+name: hypothesis_verification-medium-medium011
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/medium011.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In the 2022 provincial-level data for the automotive manufacturing industry, some provinces exhibit a dual structural contradiction: first, the number of automotive manufacturing enterprises in the province exceeds the average number of enterprises across all provinces with automotive manufacturing, but the province's total automotive manufacturing revenue is below the average revenue of all provinces; second, the province's average R&D investment ratio is higher than the mean of this indicator across provinces, but the average profit margin (measured as total operating profit / total revenue × 100%) is lower than the mean of this profit margin indicator across all provinces. Please count the number of provinces in 2022 that simultaneously meet both of the above contradiction conditions.
+
+Output guidelines:
+Answer format: numerical value (integer). If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_399_hypothesis_verification_medium_medium012.md b/tasks/task_399_hypothesis_verification_medium_medium012.md
new file mode 100644
index 0000000000000000000000000000000000000000..d97c40db1b42a482a374a21a4cfff5c9ba22b028
--- /dev/null
+++ b/tasks/task_399_hypothesis_verification_medium_medium012.md
@@ -0,0 +1,117 @@
+---
+id: task_399_hypothesis_verification_medium_medium012
+name: hypothesis_verification-medium-medium012
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/medium012.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, based on the 'structural efficiency paradox' hypothesis in industrial economics (i.e., during enterprise technology transformation, the anomalous phenomenon may occur where revenue growth coexists with workforce reduction, and reduced R&D investment coexists with increased innovation output), verify whether the dual paradox exists in the chemical fiber manufacturing industry. The hypothesis states: when enterprises are in the technology upgrading stage, revenue grows but automation replaces labor leading to fewer employees; meanwhile, R&D investment ratio is below the industry median but patent grants are above the industry median. Please count the number of valid enterprises that simultaneously exhibit both paradox characteristics.
+
+Output guidelines:
+Answer format: numerical value (integer, unit: enterprises). If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`2`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_400_hypothesis_verification_medium_medium013.md b/tasks/task_400_hypothesis_verification_medium_medium013.md
new file mode 100644
index 0000000000000000000000000000000000000000..e7f1fb54fc2d1094135d5bb4913176d1317e8985
--- /dev/null
+++ b/tasks/task_400_hypothesis_verification_medium_medium013.md
@@ -0,0 +1,117 @@
+---
+id: task_400_hypothesis_verification_medium_medium013
+name: hypothesis_verification-medium-medium013
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/medium013.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in the information transmission, software and information technology services industry, researchers aim to characterize a type of dual anomaly enterprises with high valuation and high R&D investment, yet weak profitability and insufficient patent influence. The specific criteria are: valid enterprises must have complete data for all four indicators—market cap, revenue (non-zero), R&D investment ratio, and cumulative total patent citations; net profit margin is calculated as net profit amount divided by revenue amount times 100%; among all valid enterprises, using each indicator's median as the threshold, filter enterprises where market cap exceeds the median and net profit margin is below the median, while R&D investment ratio exceeds the median and cumulative patent citations are below the median. What is the proportion of such enterprises among all valid enterprises?
+
+Output guidelines:
+Answer format: proportional value (retain 2 decimal places, unit: %). If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`3.49`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_401_hypothesis_verification_medium_medium014.md b/tasks/task_401_hypothesis_verification_medium_medium014.md
new file mode 100644
index 0000000000000000000000000000000000000000..ded8895ddfba77c8ecab27b199fc601fa2ff7b2b
--- /dev/null
+++ b/tasks/task_401_hypothesis_verification_medium_medium014.md
@@ -0,0 +1,117 @@
+---
+id: task_401_hypothesis_verification_medium_medium014
+name: hypothesis_verification-medium-medium014
+category: hypothesis_verification
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/hypothesis_verification/medium014.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, based on the 'leverage-growth coupling effect' hypothesis in the context of industrial upgrading (i.e., there is a non-linear relationship between corporate financial leverage and operating growth, potentially showing a reverse coupling pattern of high leverage with high growth or low leverage with negative growth), in provinces covered by local policies for the rubber and plastic products industry, verify whether listed enterprises in this industry exhibit structural leverage-growth association. Specifically: among valid enterprises, using the median asset-liability ratio as the financial leverage threshold and the sign of revenue growth rate as the growth direction indicator, count the number of coupling enterprises satisfying 'high leverage (asset-liability ratio > median) and high growth (growth rate > 0)' (A) and coupling enterprises satisfying 'low leverage (asset-liability ratio < median) and negative growth' (B), and calculate the difference A − B (integer). What is the specific value of this difference?
+
+Output guidelines:
+Answer format: numerical value (integer). A positive value indicates that 'high leverage high growth' coupling is more prominent; a negative value indicates that 'low leverage negative growth' coupling is more prominent. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`4`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_402_industry_planning_hard_hard001.md b/tasks/task_402_industry_planning_hard_hard001.md
new file mode 100644
index 0000000000000000000000000000000000000000..599375ba889f6403e7403cfa3eef6423d7e8ebfc
--- /dev/null
+++ b/tasks/task_402_industry_planning_hard_hard001.md
@@ -0,0 +1,119 @@
+---
+id: task_402_industry_planning_hard_hard001
+name: industry_planning-hard-hard001
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/hard001.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+基于2022年的数据，现对中国大陆半导体行业各省（相关企业数量>=5）进行政策分化情景推演：凡已落地集成电路产业发展促进类专项政策的省份，其辖内半导体企业可维持现有研发扩张节奏（以省内各企业研发投入同比增减幅的中位数为准）；而尚未出台上述专项政策的省份，受政策缺位影响，预计其研发增速将压缩至原有水平的一半。在此分化情景下，以3年复合增长方式推算至2025年，请问届时研发投入规模居于各省首位的是哪个省份？该省预测研发投入总额约为多少亿元？
+
+Output guidelines:
+依次回答省份名称和预计研发投入总额。金额以亿元为单位，保留2位小数。如["浙江省", 312.75]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["上海市", 508.26]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_403_industry_planning_hard_hard002.md b/tasks/task_403_industry_planning_hard_hard002.md
new file mode 100644
index 0000000000000000000000000000000000000000..76caeab5c0fcb26f68c317a94888455b3bab8dac
--- /dev/null
+++ b/tasks/task_403_industry_planning_hard_hard002.md
@@ -0,0 +1,119 @@
+---
+id: task_403_industry_planning_hard_hard002
+name: industry_planning-hard-hard002
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/hard002.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+以2022年化学原料和化学制品制造业的实际数据为起点，构建如下政策差异化情景：若某省（仅纳入中国大陆化学原料和化学制品制造业存续企业数量不低于8家的省份，港澳台不在统计范围内）已发布新材料领域的产业相关发展政策且涉及具体目标和量化指标，则该省化工企业得以按照各自当前研发增速（取省内企业研发投入同比增减幅的中位数）持续推进研发扩张；反之，凡无此类专项政策的省份，其研发增速将以现有水平的一半计算。在3年复合增长模型下展望至2025年，试问：哪个省份的研发投入省际排名跃升幅度最为显著？该省届时预估的研发投入总规模是多少亿元？
+
+Output guidelines:
+依次回答省份名称和预计研发投入总额。研发投入总额以亿元为单位，保留2位小数。如["湖北省", 72.31]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["湖南省", 25.13]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_404_industry_planning_hard_hard003.md b/tasks/task_404_industry_planning_hard_hard003.md
new file mode 100644
index 0000000000000000000000000000000000000000..a6c598a9424af65ebfb17ceda5211e1dfe6281d3
--- /dev/null
+++ b/tasks/task_404_industry_planning_hard_hard003.md
@@ -0,0 +1,119 @@
+---
+id: task_404_industry_planning_hard_hard003
+name: industry_planning-hard-hard003
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/hard003.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+针对中国大陆医药制造业，以2022年各省数据为基准，现设定如下政策效应假设：已颁布生物医药产业发展促进相关政策的省份（仅计入中国大陆医药制造业企业数量不少于8家的省级行政区，不含港澳台地区），其辖区内医药企业在未来三年内营业收入年增速可在原有基础上额外叠加5个百分点（原增速以该省全部医药企业营业收入同比增减幅的中位数衡量）；而那些尚未出台此类促进政策的省份，因缺乏政策催化，营业收入增速将较现有水平收缩20%。在上述差异化情景下，对各省营业收入以3年复合增长方式推算至2025年，请问：哪个省份在这轮重新洗牌后实现了最大幅度的营收排名晋升？对应的2025年预计营业收入总量为多少亿元？
+
+Output guidelines:
+依次回答省份名称和预计营业收入总额。营业收入总额以亿元为单位，保留2位小数。如["湖北省", 425.18]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["河南省", 350.45]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_405_industry_planning_hard_hard004.md b/tasks/task_405_industry_planning_hard_hard004.md
new file mode 100644
index 0000000000000000000000000000000000000000..a8153d13652117d99551a1f2801bbd85749cc421
--- /dev/null
+++ b/tasks/task_405_industry_planning_hard_hard004.md
@@ -0,0 +1,119 @@
+---
+id: task_405_industry_planning_hard_hard004
+name: industry_planning-hard-hard004
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/hard004.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+以2022年数据为起点，对中国大陆汽车制造业各省（统计范围限定为汽车制造业在册企业不少于5家的中国大陆省份，不含港澳台）进行如下情景模拟：已出台地方性新能源汽车及智能汽车产业发展专项政策的省份，其汽车制造业企业营业收入年增速将在当前中位增速基础上再叠加3个百分点（当前增速以各省企业营业收入同比增减幅中位数为准）；未出台此类专项政策的省份则呈现增长动力不足的局面，营业收入（增速取各省企业营业收入同比增减幅的中位数））增速将萎缩至原有水平的70%。按3年复合增长推算至2025年，请找出：在拥有政策支持的省份中，哪个省份将凭借政策加持超越其2022年时排名本高于自身的某个无政策省份，从而实现排名反超？该省届时的预计营业收入总量是多少亿元？
+
+Output guidelines:
+依次回答省份名称和预计营业收入总额。营业收入总额以亿元为单位，保留2位小数。如["广东省", 5230.41]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["北京市", 4850.79]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_406_industry_planning_hard_hard005.md b/tasks/task_406_industry_planning_hard_hard005.md
new file mode 100644
index 0000000000000000000000000000000000000000..63acd3f2d5476b0664fd88b8b585c91f9cb803c5
--- /dev/null
+++ b/tasks/task_406_industry_planning_hard_hard005.md
@@ -0,0 +1,119 @@
+---
+id: task_406_industry_planning_hard_hard005
+name: industry_planning-hard-hard005
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/hard005.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+云南省于2022年正式发布了绿色铝产业发展三年行动计划，其中载明了2024年绿色铝全产业链产值的量化目标。若以该省金属冶炼和压延加工业所有上市企业（仅覆盖中国大陆范围内的金属冶炼和压延加工业企业，不含港澳台数据））2022年的实际营业收入为基数，并假设各企业按自身现有增速（取全部上市企业营业收入同比增减幅的中位数作为统一测算基准）持续保持增长，经2年复合增长推算至2024年，请问：届时这些上市企业的营收汇总值与政策文件明确设定的产业链产值目标之间还存在多大缺口（以亿元计）？若要在2年内完全弥合上述缺口，在现有增速基础上还需年均额外拉升多少个百分点？
+
+Output guidelines:
+依次回答缺口金额和额外增速。缺口金额以亿元为单位，保留2位小数；额外增速以百分点为单位，保留2位小数。如[280.50, 4.12]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[356.36, 5.85]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_407_industry_planning_hard_hard006.md b/tasks/task_407_industry_planning_hard_hard006.md
new file mode 100644
index 0000000000000000000000000000000000000000..5704127a7cfee85bba5824fb342a11a9f8232449
--- /dev/null
+++ b/tasks/task_407_industry_planning_hard_hard006.md
@@ -0,0 +1,119 @@
+---
+id: task_407_industry_planning_hard_hard006
+name: industry_planning-hard-hard006
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/hard006.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+聚焦2022年中国大陆非金属矿物制品业，考察以下政策差异化情景对各省（非金属矿物制品业企业数不少于5家的中国大陆省级行政区，不纳入港澳台数据）研发格局的重塑效果：凡已发布专门涉及建材行业绿色转型内容的省级碳达峰或节能减排专项实施方案的省份，其辖内非金属矿物制品业企业可保持现有研发投入增速不变（增速以该省各企业研发投入同比增减幅的中位数为准）；而尚未落地此类省级专项方案的省份，其研发扩张动能将打折，增速降至当前水平的50%。以上述差异化增速进行3年复合增长测算，预测各省2025年研发投入规模并进行重新排序，请问：从2022年到2025年，哪个省份实现了最大幅度的研发投入排名跃升？该省2025年研发投入总额约为多少亿元？
+
+Output guidelines:
+依次回答省份名称和预计研发投入总额。总额以亿元为单位，保留2位小数。如["湖北省", 20.04]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["湖南省", 31.15]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_408_industry_planning_hard_hard007.md b/tasks/task_408_industry_planning_hard_hard007.md
new file mode 100644
index 0000000000000000000000000000000000000000..29ec874199100484d6cb1ad404bac6457636e9b5
--- /dev/null
+++ b/tasks/task_408_industry_planning_hard_hard007.md
@@ -0,0 +1,119 @@
+---
+id: task_408_industry_planning_hard_hard007
+name: industry_planning-hard-hard007
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/hard007.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+就2022年中国大陆软件和信息技术服务业的人才规模格局而言，若设定如下政策导向假设（仅含软件和信息技术服务业存续企业数量达到5家及以上门槛的中国大陆省份，港澳台数据不纳入计算）：凡已正式出台软件产业高质量发展专项政策的省份，其企业员工总量可沿现有轨道持续扩张，年增速以该省各企业雇员同比增减幅的中位数+13%为准；而对于尚未出台此类专项政策的省份，因政策引领欠缺导致人才吸附力不足，员工扩张速度将缩减至现有增速的一半。在这一情景设定下以3年复合增长推算至2025年，哪个省份在员工总量省际排名中实现了最大幅度的正向位次变动？请一并报告该省届时预计的从业人员总规模。
+
+Output guidelines:
+依次回答省份名称和预计员工总人数。员工总人数保留2位小数，单位为万人。如["四川省", 2.58]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["安徽省", 3.49]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_409_industry_planning_hard_hard008.md b/tasks/task_409_industry_planning_hard_hard008.md
new file mode 100644
index 0000000000000000000000000000000000000000..675b57b4ed5178602bfb1c16a3ed0ff0aae526eb
--- /dev/null
+++ b/tasks/task_409_industry_planning_hard_hard008.md
@@ -0,0 +1,119 @@
+---
+id: task_409_industry_planning_hard_hard008
+name: industry_planning-hard-hard008
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/hard008.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+以2022年为基期，针对中国大陆通用设备制造业设计双情景对比测算框架（通用设备制造业在册企业数量不少于5家的中国大陆省份，港澳台不纳入）。情景一（政策分化情景）：已出台制造业高质量发展省级专项文件的省份，其通用设备制造业企业总资产按各自当前增速+6%（以省内各企业营业收入同比增减幅的中位数作为总资产增速的替代指标）持续扩张，无政策省份则以减半增速计算；情景二（全量减半基准情景）：所有省份不论有无政策，一律按当前增速的一半推算总资产增长。以3年复合增长分别推算两种情景下各省2025年总资产，并以两情景之差作为"政策带来的额外总资产增量"，请问：在出台了上述专项政策且符合最低企业数量门槛的省份中，哪个省份从该政策中撬动的额外总资产增量（额外增量=政策分化情景下2025年总资产 - 全量减半情景下2025年总资产）最为可观？该增量具体为多少亿元？
+
+Output guidelines:
+依次回答省份名称和额外总资产增量。额外增量保留2位小数，单位为亿元。如["湖南省", 520.38]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["广东省", 140.38]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_410_industry_planning_hard_hard009.md b/tasks/task_410_industry_planning_hard_hard009.md
new file mode 100644
index 0000000000000000000000000000000000000000..3f2e2ab3c7e31ab131e1e75e38a356e48c7ef172
--- /dev/null
+++ b/tasks/task_410_industry_planning_hard_hard009.md
@@ -0,0 +1,119 @@
+---
+id: task_410_industry_planning_hard_hard009
+name: industry_planning-hard-hard009
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/hard009.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+以2022年铁路、船舶、航空航天和其他运输设备制造业的省级（限定为铁路、船舶、航空航天和其他运输设备制造业辖内企业数量达到3家以上的中国大陆省级行政区，港澳台数据不纳入）数据为基准，模拟如下政策分化对行业格局的冲击：已出台船舶与海洋工程装备产业专项发展政策的省份，其企业营业收入增速在未来三年将在现有中位水平上额外叠加5个百分点；而没有落地此类专项政策的省份，受制于政策真空，营业收入增速将萎缩至现有水平的70%（增速以各省企业营业收入同比增减幅中位数为准）。按各省调整后的增速进行3年复合增长推算，对比2022年与2025年的省际营收排名变动，请从无政策省份中找出：哪个省份享受政策红利而获得最大幅度的排名上升？它一共上升了几位？该省2025年的预测营业收入总量为多少亿元？
+
+Output guidelines:
+依次回答省份名称、排名下降名数（整数）、该省2025年预计营业收入总额。营业收入总额保留2位小数，单位为亿元。如["广东省", 2, 512.34]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["山东省", 2, 697.09]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_411_industry_planning_hard_hard010.md b/tasks/task_411_industry_planning_hard_hard010.md
new file mode 100644
index 0000000000000000000000000000000000000000..72c1f7c47d16122f91db91c47577574860b94ec5
--- /dev/null
+++ b/tasks/task_411_industry_planning_hard_hard010.md
@@ -0,0 +1,119 @@
+---
+id: task_411_industry_planning_hard_hard010
+name: industry_planning-hard-hard010
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/hard010.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+在铁路、船舶、航空航天和其他运输设备制造业（即广义轨道交通装备制造业）领域，以2022年为基期，构建以下轨道交通政策激励传导模型（轨道交通装备制造相关企业存续数量不低于3家的中国大陆省级行政区，港澳台不计入）：若某省已明确将轨道交通装备产业列入重点打造产业集群目录并配套专项支持措施，则该省企业营业收入的年增速可在现有中位水平上再叠加3个百分点；而未落地此类产业集群专项支持政策的省份，其营业收入增速将在原有基础上收缩30%（增速取省内各企业营业收入同比增减幅中位数）。以上述情景增速进行3年复合增长预测，并计算各有政策省份从2022年到2025年的营业收入绝对增量（定义为：2025年预测营业收入 − 2022年实际营业收入），请问：哪个获得政策支持的省份营业收入的绝对增量最为可观？该省这一增量数值为多少亿元？
+
+Output guidelines:
+依次回答省份名称和营业收入绝对增量。绝对增量保留2位小数。如["上海市", 185.42]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["山东省", 413.38]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_412_industry_planning_hard_hard011.md b/tasks/task_412_industry_planning_hard_hard011.md
new file mode 100644
index 0000000000000000000000000000000000000000..feaafb3ac2766e2b58d6152ae270ce13a8d8a6fd
--- /dev/null
+++ b/tasks/task_412_industry_planning_hard_hard011.md
@@ -0,0 +1,119 @@
+---
+id: task_412_industry_planning_hard_hard011
+name: industry_planning-hard-hard011
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/hard011.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+在铁路、船舶、航空航天和其他运输设备制造业中，以2022年各省（港澳台地区数据不计入本题统计范围）实际数据为起点，设定如下情景假设：已颁布航空航天产业发展专项支持政策的省份，其行业内企业能够维持现有营业收入增速（以各省企业营业收入同比增减幅的中位数衡量）持续扩张；而未出台此类专项支持政策的省份，由于缺乏政策引导，营业收入增速将被动压缩至当前水平衰减40%。在以上差异化条件下按3年复合增长推算至2025年，请重点关注2022年营业收入总量尚未跨越100亿元门槛的有政策省份——在这一子集中，哪个省份将在政策扶持下率先实现百亿营收的历史性突破？该省届时的预测营业收入总额为多少亿元？
+
+Output guidelines:
+依次回答省份名称和2025年预计营业收入总额。营业收入总额保留2位小数，单位为亿元。如["江西省", 89.37]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["四川省", 116.56]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_413_industry_planning_hard_hard012.md b/tasks/task_413_industry_planning_hard_hard012.md
new file mode 100644
index 0000000000000000000000000000000000000000..bc7ec61bb6c5abc34baee8c3c0b8eea10ac46462
--- /dev/null
+++ b/tasks/task_413_industry_planning_hard_hard012.md
@@ -0,0 +1,119 @@
+---
+id: task_413_industry_planning_hard_hard012
+name: industry_planning-hard-hard012
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/hard012.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，在中国大陆医疗仪器设备及器械制造业中（仅统计仪器仪表制造业企业不少于3家的中国大陆省份，不含港澳台），假设出台了医疗器械产业高端化（适用于医药制造和仪表仪器）发展专项政策的省份，其企业净利润增速在未来3年额外提升5个百分点，而未出台此类政策的省份净利润增速衰减20%，到2025年净利润总额排名上升幅度最大的省份是哪个（增速使用各省份企业净利润同比增减幅的中位数；增长方式为3年复合增长）？其预计净利润总额是多少亿元？
+
+Output guidelines:
+依次回答省份名称和预计净利润总额。净利润总额保留2位小数。如["安徽省", 6.92]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["北京市", 4.34]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_414_industry_planning_hard_hard013.md b/tasks/task_414_industry_planning_hard_hard013.md
new file mode 100644
index 0000000000000000000000000000000000000000..a2cba1a38b0236acebb073e358d15cb0c8e96d46
--- /dev/null
+++ b/tasks/task_414_industry_planning_hard_hard013.md
@@ -0,0 +1,119 @@
+---
+id: task_414_industry_planning_hard_hard013
+name: industry_planning-hard-hard013
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/hard013.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+假设针对金属冶炼和压延加工业出台了新材料产业发展专项扶持政策的省份（仅统计2022年净利润数据完整的中国大陆省份，不含港澳台），其金属冶炼和压延加工业上市企业净利润按2022年的增速（各省企业净利润同比增减幅中位数）持续增长，而未出台此类政策的省份净利润增速减半，在3年复合增长模型下，到2025年有政策省份净利润总和与无政策省份净利润总和的比值是多少？相比2022年该比值变化了多少（提升或下降）？
+
+Output guidelines:
+依次回答2025年的比值和比值变化量。比值保留2位小数，变化量保留2位小数并注明提升或下降。如[1.85, "提升0.93"]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[3.61, "提升1.79"]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_415_industry_planning_hard_hard014.md b/tasks/task_415_industry_planning_hard_hard014.md
new file mode 100644
index 0000000000000000000000000000000000000000..639f706115a102816b403d7e12e6342d17665f17
--- /dev/null
+++ b/tasks/task_415_industry_planning_hard_hard014.md
@@ -0,0 +1,119 @@
+---
+id: task_415_industry_planning_hard_hard014
+name: industry_planning-hard-hard014
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/hard014.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，在中国大陆消费电子及电气业中，假设出台了电子信息产业集群培育专项政策的省份（仅统计消费电子及电气业企业不少于5家的中国大陆省份，不含港澳台），其企业净利润增速在未来3年额外提升8个百分点，而未出台此类专项政策的省份净利润增速衰减至当前水平的50%（净利润增速取各省份企业净利润同比增速的中位数；增长方式为3年复合增长），到2025年，在2022年净利润总额排名处于后半段（排名靠后一半）的有政策省份中，排名提升幅度最大的是哪个省份？其预计净利润总额是多少亿元？
+
+Output guidelines:
+依次回答省份名称和预计净利润总额。净利润总额保留2位小数。如["江西省", 45.82]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["四川省", 63.45]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_416_industry_planning_medium_medium001.md b/tasks/task_416_industry_planning_medium_medium001.md
new file mode 100644
index 0000000000000000000000000000000000000000..c69e2033f715bf29c05c03979e7d6fb84c3c55c0
--- /dev/null
+++ b/tasks/task_416_industry_planning_medium_medium001.md
@@ -0,0 +1,119 @@
+---
+id: task_416_industry_planning_medium_medium001
+name: industry_planning-medium-medium001
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/medium001.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Based on 2022 data, the following policy effect scenario is simulated: For provinces that have promulgated industrial policies containing the keywords "semiconductor" or "integrated circuit", policy empowerment accelerates the R&D investment expansion pace of their semiconductor enterprises to 2 times the current growth rate over the next 3 years; for provinces that have not yet issued such policies, R&D growth rate remains unchanged. Using the median year-on-year change in enterprise R&D investment as the baseline growth rate for each province, and projecting with 3-year compound growth, which province will have the highest total semiconductor industry R&D investment nationwide by 2025? What is the corresponding estimated amount?
+
+Output guidelines:
+Answer format: [province name, value (2 decimal places, unit: yuan)]. If relevant data cannot be found, please answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["Shanghai", 97732260069.03]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_417_industry_planning_medium_medium002.md b/tasks/task_417_industry_planning_medium_medium002.md
new file mode 100644
index 0000000000000000000000000000000000000000..490e5a5e4030619123e5cf439797f026876f39d4
--- /dev/null
+++ b/tasks/task_417_industry_planning_medium_medium002.md
@@ -0,0 +1,117 @@
+---
+id: task_417_industry_planning_medium_medium002
+name: industry_planning-medium-medium002
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/medium002.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, assume that in the pharmaceutical manufacturing industry, the annual operating revenue growth rate of private enterprises is 5 percentage points higher than state-owned enterprises (including central and local state-owned enterprises) in the same province, while state-owned enterprises maintain their current growth rate unchanged (growth rate measured by the median year-on-year change in operating revenue of enterprises in the same province). By 2025, how many provinces will have private enterprise total revenue exceeding state-owned enterprise total revenue for the first time?
+
+Output guidelines:
+Answer format: integer (unit: count). If relevant data cannot be found, please answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`1`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_418_industry_planning_medium_medium003.md b/tasks/task_418_industry_planning_medium_medium003.md
new file mode 100644
index 0000000000000000000000000000000000000000..49906bcb4509c657ee3787c6e1ef3771e767b531
--- /dev/null
+++ b/tasks/task_418_industry_planning_medium_medium003.md
@@ -0,0 +1,117 @@
+---
+id: task_418_industry_planning_medium_medium003
+name: industry_planning-medium-medium003
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/medium003.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Using 2022 as the base period, a differentiated policy incentive is proposed for the automobile manufacturing industry: For provinces that have implemented automobile industry policies containing the keywords "new energy" or "electric", the annual operating profit growth rate of automobile manufacturing enterprises within their jurisdiction will add 10 percentage points on top of the current median growth rate, creating a policy acceleration effect; other provinces are unaffected, and operating profit growth rate continues along the current trajectory. Under this differentiated scenario, using the median year-on-year change in operating profit as the baseline growth rate for each province, and projecting to 2025 via 3-year compound growth, which province has the most prominent increase in total operating profit compared to actual 2022 levels? What is the specific increase (increase = (2025 estimated value - 2022 actual value) / 2022 actual value × 100%)?
+
+Output guidelines:
+Answer format: value (2 decimal places, unit: %). If relevant data cannot be found, please answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`859.71`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_419_industry_planning_medium_medium004.md b/tasks/task_419_industry_planning_medium_medium004.md
new file mode 100644
index 0000000000000000000000000000000000000000..6f4eec7d62b2f97086cac9f9e2035b8b9cc7c63f
--- /dev/null
+++ b/tasks/task_419_industry_planning_medium_medium004.md
@@ -0,0 +1,117 @@
+---
+id: task_419_industry_planning_medium_medium004
+name: industry_planning-medium-medium004
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/medium004.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in the communication transmission equipment industry, assuming that enterprises with R&D investment ratio below the national industry median will be eliminated from the market within 3 years, and only enterprises with R&D investment ratio not lower than the national median will be retained, what is the proportion for the province with the highest ratio of remaining enterprises' operating revenue after elimination to operating revenue before elimination?
+
+Output guidelines:
+Answer format: Value (2 decimal places, unit: %). If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`100.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_420_industry_planning_medium_medium005.md b/tasks/task_420_industry_planning_medium_medium005.md
new file mode 100644
index 0000000000000000000000000000000000000000..9b6a0d8bd641ae2c4e1758672297c81e09d0b948
--- /dev/null
+++ b/tasks/task_420_industry_planning_medium_medium005.md
@@ -0,0 +1,117 @@
+---
+id: task_420_industry_planning_medium_medium005
+name: industry_planning-medium-medium005
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/medium005.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, regarding the strategic choice for Guangdong Province's consumer electronics and electrical industry, a policy consulting agency proposed two competing development paths: the first is the "high-end transformation route", evaluated by the average R&D investment ratio of private enterprises, invention patent density (= total annual Chinese invention patent grants ÷ total number of enterprises), and the number of relevant industrial policies in that province; the second is the "export-oriented route", evaluated by per capita revenue (= total operating revenue ÷ total number of employees), average asset turnover rate (= mean operating revenue ÷ mean total assets), and total number of enterprises. Both routes use inter-provincial peer comparison ranking scores (score = (N - ranking) / (N - 1) × 100), with equal weight across dimensions to calculate the route total score. What is the difference between Guangdong Province's total score on the "high-end transformation route" and the "export-oriented route" (former minus latter)?
+
+Output guidelines:
+Answer format: Value (2 decimal places). A positive number indicates the high-end transformation route has a higher score, a negative number indicates the export-oriented route has a higher score. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`4.55`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_421_industry_planning_medium_medium006.md b/tasks/task_421_industry_planning_medium_medium006.md
new file mode 100644
index 0000000000000000000000000000000000000000..85ad457426be9c3987459d5080618d322d6f1eff
--- /dev/null
+++ b/tasks/task_421_industry_planning_medium_medium006.md
@@ -0,0 +1,117 @@
+---
+id: task_421_industry_planning_medium_medium006
+name: industry_planning-medium-medium006
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/medium006.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, regarding the industrial development direction of Hebei Province's metal smelting and rolling processing industry, researchers intend to compare two alternative transformation paths through a multi-dimensional scoring method. Path one "green and low-carbon route" covers three evaluation indicators: average enterprise R&D investment amount (reflecting technology upgrade willingness), invention patent density (= total cumulative Chinese invention patent grants ÷ total number of enterprises, weight 0.3), and count of provincial green-related policies (i.e., policy records whose name contains "green", "low-carbon", or "energy-saving", weight 0.4), with average R&D investment amount weight 0.3; Path two "traditional capacity expansion route" also includes three indicators: total assets (weight 0.4), total operating revenue (weight 0.3), and total number of enterprises (weight 0.3). Each province's score for each indicator is calculated by inter-provincial peer ranking (score = (N - ranking) / (N - 1) × 100). Please calculate the score difference between Hebei Province on the above two routes (green and low-carbon route score minus traditional capacity expansion route score).
+
+Output guidelines:
+Answer format: Value (2 decimal places). A positive number indicates the green and low-carbon route has a higher score, a negative number indicates the traditional capacity expansion route has a higher score. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`42.97`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_422_industry_planning_medium_medium007.md b/tasks/task_422_industry_planning_medium_medium007.md
new file mode 100644
index 0000000000000000000000000000000000000000..50d99fa5be6b266cca0a51c7c730fa95ca6e3a65
--- /dev/null
+++ b/tasks/task_422_industry_planning_medium_medium007.md
@@ -0,0 +1,117 @@
+---
+id: task_422_industry_planning_medium_medium007
+name: industry_planning-medium-medium007
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/medium007.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In the 2022 data for the chemical raw materials and chemical products manufacturing industry, first identify the enterprise with the largest total assets in the entire industry and determine its registration province; then take that province as the research object, and use the four-indicator equal-weight scoring method (each indicator score = (N - inter-provincial ranking) / (N - 1) × 100) to compare two industrial strategy routes: the "R&D-driven route" comprehensively evaluates four indicators: total private enterprise R&D investment, total state-owned enterprise R&D investment (central state-owned + local state-owned + other state-owned enterprises + state-owned enterprises (research institutes)), regional R&D intensity (mean R&D investment ratio), and count of relevant R&D policies (policy name contains "R&D", "innovation", or "technology/science"); the "scale expansion route" comprehensively evaluates four indicators: total assets, total operating revenue, total number of enterprises, and total government subsidies. Which route has the higher comprehensive score for that province?
+
+Output guidelines:
+Answer format: "R&D-driven route" or "Scale expansion route". If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Scale expansion route"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_423_industry_planning_medium_medium008.md b/tasks/task_423_industry_planning_medium_medium008.md
new file mode 100644
index 0000000000000000000000000000000000000000..818ffbc9f94531a8b03f8f4da2355ffe947f070d
--- /dev/null
+++ b/tasks/task_423_industry_planning_medium_medium008.md
@@ -0,0 +1,117 @@
+---
+id: task_423_industry_planning_medium_medium008
+name: industry_planning-medium-medium008
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/medium008.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in the food and beverage industry, for the province where the enterprise with the most cumulative Chinese invention patent grants is located, if that province chooses the "brand upgrade route" (evaluating market-cap-to-revenue ratio, profit margin, per capita market cap, with weights of 35%, 35%, 30% respectively) versus the "industrial chain extension route" (evaluating total number of enterprises, revenue scale, upstream-downstream enterprise diversity, with weights of 40%, 30%, 30% respectively), which route has the higher score?
+
+Output guidelines:
+Answer format: "Brand upgrade route" or "Industrial chain extension route". If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Industrial chain extension route"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_424_industry_planning_medium_medium009.md b/tasks/task_424_industry_planning_medium_medium009.md
new file mode 100644
index 0000000000000000000000000000000000000000..0af5e62b6edbd54082161181371546360b134ead
--- /dev/null
+++ b/tasks/task_424_industry_planning_medium_medium009.md
@@ -0,0 +1,117 @@
+---
+id: task_424_industry_planning_medium_medium009
+name: industry_planning-medium-medium009
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/medium009.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Among the enterprises in the textile, footwear and apparel industry in 2022, find the enterprise with the largest R&D personnel scale; its province is the analysis object. For that province's textile, footwear and apparel industry, use the following two route scoring systems to determine which development route has the advantage—the "automation upgrade route" scores by three indicators: R&D investment intensity (= total R&D investment amount / total operating revenue, weight 0.4), capitalized R&D investment ratio (= total capitalized R&D investment / total R&D investment amount, weight 0.3), and count of equipment manufacturing policies (policy name contains "equipment" or "intelligent manufacturing", weight 0.3); the "brand overseas expansion route" scores by cumulative PCT patent applications (weight 0.4), per capita revenue (weight 0.3), and count of export-related policies (policy name contains "export", "foreign trade", or "international", weight 0.3); both routes use inter-provincial ranking scores for each indicator (score = (N - ranking) / (N - 1) × 100). Which route has the higher score for that province?
+
+Output guidelines:
+Answer format: "Automation upgrade route" or "Brand overseas expansion route". If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Automation upgrade route"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_425_industry_planning_medium_medium010.md b/tasks/task_425_industry_planning_medium_medium010.md
new file mode 100644
index 0000000000000000000000000000000000000000..1b6ffb4ffaa96b9264f3cde30ad385d848bcc6be
--- /dev/null
+++ b/tasks/task_425_industry_planning_medium_medium010.md
@@ -0,0 +1,117 @@
+---
+id: task_425_industry_planning_medium_medium010
+name: industry_planning-medium-medium010
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/medium010.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Sort the semiconductor industry in 2022 by operating revenue from high to low, identify the top 10 enterprises by revenue scale, and count the provinces to which these leading enterprises belong; the province with the highest frequency is the research target. Next, conduct a national horizontal comparison of that province based on four industrial competitiveness dimensions—dimension one is enterprise agglomeration (total number of semiconductor industry enterprises per province), dimension two is innovation activity (= total annual Chinese invention patent grants ÷ total number of enterprises), dimension three is policy support (= count of relevant policies whose name contains "semiconductor" or "integrated circuit"), dimension four is industry scale (= total operating revenue). Among the national provincial rankings for each dimension, in how many dimensions did that province rank in the top 3?
+
+Output guidelines:
+Answer format: Integer. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`3`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_426_industry_planning_medium_medium011.md b/tasks/task_426_industry_planning_medium_medium011.md
new file mode 100644
index 0000000000000000000000000000000000000000..853429280599455c748e23ef8dcc911cfd31c63a
--- /dev/null
+++ b/tasks/task_426_industry_planning_medium_medium011.md
@@ -0,0 +1,117 @@
+---
+id: task_426_industry_planning_medium_medium011
+name: industry_planning-medium-medium011
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/medium011.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In the 2022 pharmaceutical manufacturing industry data, find the listed enterprise with the highest total Chinese patent grants in that year; after identifying that enterprise's province, construct a four-dimensional evaluation framework around "innovation ecosystem": R&D intensity (= total R&D investment amount ÷ total operating revenue), patent conversion efficiency (= total cumulative Chinese invention patent grants ÷ total cumulative Chinese invention patent applications), industry scale (= total operating revenue), and policy support (= count of relevant policies whose name contains "pharmaceutical" or "biotechnology"). For each dimension, identify the top 5 provinces nationwide and calculate their mean. Is that province's overall innovation ecosystem level above or below the average of the national top 5 across these four dimensions?
+
+Output guidelines:
+Answer format: "Above average" or "Below average". If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Below average"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_427_industry_planning_medium_medium012.md b/tasks/task_427_industry_planning_medium_medium012.md
new file mode 100644
index 0000000000000000000000000000000000000000..0612a9329b32707489e39c8d52cfad9fc16f6433
--- /dev/null
+++ b/tasks/task_427_industry_planning_medium_medium012.md
@@ -0,0 +1,117 @@
+---
+id: task_427_industry_planning_medium_medium012
+name: industry_planning-medium-medium012
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/medium012.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in the railway, ship, aerospace and other transport equipment manufacturing industry, take the province with the highest concentration of the top 5 enterprises by total assets as the research object. What is that province's comprehensive ranking (ranking based on the arithmetic mean of the three dimension rankings) across three high-end manufacturing dimensions: technology intensity (technology intensity = total R&D personnel / total employees), capital intensity (capital intensity = total assets / total employees), and policy concentration (policy concentration = count of relevant policies)?
+
+Output guidelines:
+Answer format: Integer (unit: rank). If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`4`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_428_industry_planning_medium_medium013.md b/tasks/task_428_industry_planning_medium_medium013.md
new file mode 100644
index 0000000000000000000000000000000000000000..6dad7673a7951417660afb3d00b13f5ac803aef3
--- /dev/null
+++ b/tasks/task_428_industry_planning_medium_medium013.md
@@ -0,0 +1,117 @@
+---
+id: task_428_industry_planning_medium_medium013
+name: industry_planning-medium-medium013
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/medium013.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In the 2022 automotive manufacturing industry, take the top 10 provinces by profitability (measured by net profit) as the candidate set, and construct a three-dimensional comprehensive scoring model to identify the province with the best industrial development quality: Dimension one "industrial chain completeness" (weight 0.3) comprehensively assesses the number of enterprise ownership types (ownership diversity) and the interquartile range of total assets of enterprises within the province (scale diversity); dimension two "technological capability" (weight 0.4) comprehensively assesses mean R&D investment ratio (R&D intensity) and total annual Chinese invention patent grants divided by total number of enterprises (patent density); dimension three "market performance" (weight 0.3) comprehensively assesses total operating revenue (revenue scale) and total operating profit divided by total operating revenue (profit margin). Each dimension's score is represented by the average ranking of its sub-indicators among the candidate provinces (comprehensive score = weighted average of each dimension's mean ranking). Which province has the highest comprehensive score (i.e., the best overall performance)?
+
+Output guidelines:
+Answer format: Province name. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Guangdong Province"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_429_industry_planning_medium_medium014.md b/tasks/task_429_industry_planning_medium_medium014.md
new file mode 100644
index 0000000000000000000000000000000000000000..2c854917e302490c8509a1b1a85c098c79b5a9a2
--- /dev/null
+++ b/tasks/task_429_industry_planning_medium_medium014.md
@@ -0,0 +1,117 @@
+---
+id: task_429_industry_planning_medium_medium014
+name: industry_planning-medium-medium014
+category: industry_planning
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/industry_planning/medium014.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In the 2022 general equipment manufacturing industry provincial data, find the province with the highest total government reward and subsidy amount; after identifying that province, conduct a comprehensive rating of its industrial competitiveness across two strategic dimensions: the "industrial upgrade capability" dimension combines the inter-provincial mean ranking of three sub-indicators—mean year-on-year change in R&D investment (R&D investment growth rate), year-on-year growth rate of annual Chinese patent applications (patent application growth rate), and mean R&D personnel ratio (high-end talent ratio); the "industrial foundation" dimension combines the inter-provincial mean ranking of three sub-indicators—total number of enterprises (enterprise scale), total operating revenue (revenue scale), and the ratio of total operating revenue to total government reward and subsidy (subsidy efficiency). The overall comprehensive performance ranking is determined by the average of the two dimension rankings. Can that province's comprehensive industrial performance rank among the national top 5?
+
+Output guidelines:
+Answer format: "Yes" or "No". If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_430_international_comparison_hard_hard001.md b/tasks/task_430_international_comparison_hard_hard001.md
new file mode 100644
index 0000000000000000000000000000000000000000..0b87de50c94224efeb303fcbba25df9d600c8281
--- /dev/null
+++ b/tasks/task_430_international_comparison_hard_hard001.md
@@ -0,0 +1,119 @@
+---
+id: task_430_international_comparison_hard_hard001
+name: international_comparison-hard-hard001
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/hard001.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Against the backdrop of China's strong push for semiconductor industry self-sufficiency and many regions issuing special support policies, what are the R&D expenses and operating revenue disclosed in United Microelectronics Corporation (UMC)'s 2022 annual report (in New Taiwan Dollars)? What is the R&D investment ratio calculated therefrom? In the domestic semiconductor industry, the top 10% by operating revenue are classified as leading enterprises. By how many percentage points does UMC's R&D investment ratio differ from the median R&D investment ratio of these leading enterprises?
+
+Output guidelines:
+Please answer in order: (1) UMC 2022 R&D expenses (hundred million NTD, 2 decimal places); (2) UMC 2022 operating revenue (hundred million NTD, 2 decimal places); (3) UMC R&D investment ratio (%, 2 decimal places); (4) Difference between UMC's R&D investment ratio and the median R&D investment ratio of domestic semiconductor industry leading enterprises (top 10% by revenue) (percentage points, 2 decimal places, negative if lower；return as an array). If relevant data cannot be found, please answer "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[129.53, 2787.05, 4.65, -1.16]`
+
+Scoring rules:
+- The gold answer is a list with N=4 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_3` each as 0 or 1.
+- Return `total = (sum(part_i)) / 4` exactly.
+- If the model output is missing or cannot be parsed into 4 comparable parts, score all parts 0.
+
diff --git a/tasks/task_431_international_comparison_hard_hard002.md b/tasks/task_431_international_comparison_hard_hard002.md
new file mode 100644
index 0000000000000000000000000000000000000000..357040e15a2261d8ca410a064d635d573d3752b4
--- /dev/null
+++ b/tasks/task_431_international_comparison_hard_hard002.md
@@ -0,0 +1,119 @@
+---
+id: task_431_international_comparison_hard_hard002
+name: international_comparison-hard-hard002
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/hard002.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, against the backdrop of the expiration of new energy vehicle purchase subsidy policies and the transition of industry support from direct subsidies to indirect incentives such as purchase tax exemption, what is the ratio of government subsidies and related income to operating revenue in Li Auto's annual report? Compared with the median government subsidy-to-revenue ratio of private enterprises and state-owned enterprises (including central and local state-owned enterprises) in the domestic automotive manufacturing industry, by how many percentage points is it higher for each?
+
+Output guidelines:
+Answer in order: Li Auto government subsidy-to-revenue ratio (%), percentage points above private enterprise median, percentage points above state-owned enterprise median. Retain 2 decimal places and return as an array. If relevant data cannot be found, please answer "No relevant data found"
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[1.38, 0.66, 0.8]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_432_international_comparison_hard_hard003.md b/tasks/task_432_international_comparison_hard_hard003.md
new file mode 100644
index 0000000000000000000000000000000000000000..20101b3ae877c9898f01fcb78dfcd7d208f1611d
--- /dev/null
+++ b/tasks/task_432_international_comparison_hard_hard003.md
@@ -0,0 +1,119 @@
+---
+id: task_432_international_comparison_hard_hard003
+name: international_comparison-hard-hard003
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/hard003.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，在碳达峰政策驱动光伏装机需求激增、多晶硅阶段性供不应求导致价格大幅上涨的背景下，大全新能源（Daqo New Energy）年报中的净利润率（净利润÷营业收入×100%）是多少？分别与国内化学原料和化学制品制造业中民营企业和国有企业（含中央及地方国有企业）的净利润率中位数相差多少个百分点？
+
+Output guidelines:
+依次回答：大全新能源2022年净利润率（%）、与民营企业中位数之差（百分点）、与国有企业中位数之差（百分点）。数值保留2位小数，并返回数组。如果无法找到相关数据，请回答"未查询到相关数据"
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[53.81, 45.48, 44.76]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_433_international_comparison_hard_hard004.md b/tasks/task_433_international_comparison_hard_hard004.md
new file mode 100644
index 0000000000000000000000000000000000000000..ac9ff3f4036bc9020fcc7721844c58f7a4767604
--- /dev/null
+++ b/tasks/task_433_international_comparison_hard_hard004.md
@@ -0,0 +1,119 @@
+---
+id: task_433_international_comparison_hard_hard004
+name: international_comparison-hard-hard004
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/hard004.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, against the backdrop of intensively issued policies promoting biopharmaceutical industry cluster development and encouraging innovative drug R&D across regions, what is Zai Lab's price-to-sales ratio (market cap ÷ annual operating revenue) in multiples? How many times is its price-to-sales ratio relative to the median price-to-sales ratio of domestic pharmaceutical manufacturing industry leading enterprises (top 10% by revenue)?
+
+Output guidelines:
+Answer in order: (1) Zai Lab 2022 price-to-sales ratio (multiples); (2) Zai Lab's price-to-sales ratio as a multiple of the domestic pharmaceutical manufacturing industry top 10% by revenue leading enterprises' median price-to-sales ratio. Retain 2 decimal places for all values and return as an array. If relevant data cannot be found, please answer "No relevant data found"
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[15.17, 7.73]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_434_international_comparison_hard_hard005.md b/tasks/task_434_international_comparison_hard_hard005.md
new file mode 100644
index 0000000000000000000000000000000000000000..d683c67f99bd94059b40f76db5e1c2fc7d734c96
--- /dev/null
+++ b/tasks/task_434_international_comparison_hard_hard005.md
@@ -0,0 +1,119 @@
+---
+id: task_434_international_comparison_hard_hard005
+name: international_comparison-hard-hard005
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/hard005.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, against the backdrop of intensively issued policies in Shanghai, Hefei, Hangzhou and elsewhere promoting high-quality development of the integrated circuit industry, what is Silicon Motion Technology's per capita net profit in the annual report converted to RMB in ten thousand yuan? Compared with the median per capita net profit of private enterprises and state-owned enterprises in the domestic semiconductor industry respectively, by how many times is it higher?
+
+Output guidelines:
+Answer in order: Silicon Motion 2022 per capita net profit (ten thousand yuan, converted at 2022 average exchange rate 1 USD ≈ 6.73 RMB), ratio to domestic semiconductor industry private enterprise median per capita net profit (times), ratio to state-owned enterprise median per capita net profit (times). Retain 2 decimal places and return as an array. If relevant data cannot be found, please answer "No relevant data found"
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[70.66, 8.23, 4.79]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_435_international_comparison_hard_hard006.md b/tasks/task_435_international_comparison_hard_hard006.md
new file mode 100644
index 0000000000000000000000000000000000000000..eae17b8a8eae4d3bdd9fe633a4cb7b6c8bb91686
--- /dev/null
+++ b/tasks/task_435_international_comparison_hard_hard006.md
@@ -0,0 +1,119 @@
+---
+id: task_435_international_comparison_hard_hard006
+name: international_comparison-hard-hard006
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/hard006.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+A global technology thematic portfolio uses a "technological moat + policy diffusion" framework for its semiconductor sub-portfolio. For candidate companies, first calculate: 1. Technological moat gap = company advanced process revenue ratio - median R&D investment ratio of A-share semiconductor industry top 10% by revenue in 2022, where advanced process revenue ratio = (5nm + 7nm revenue) ÷ wafer revenue; 2. Policy diffusion ratio = count of China semiconductor industry policies ÷ number of provincial-level administrative regions covered by non-national policies; 3. Theme conviction score = 0.6 × technological moat gap + 0.4 × policy diffusion ratio. If technological moat gap > 40 and policy diffusion ratio > 2.5, list as core overweight with active weight = min(3.00%, theme conviction score ÷ 10); otherwise do not include in core overweight list. Using TSMC's 2022 annual report and local database, calculate and determine the most appropriate conclusion.
+
+Output guidelines:
+Answer in order: technological moat gap (percentage points), policy diffusion ratio, theme conviction score, active weight, most appropriate conclusion (conclusion must specify position action and active weight). Retain 2 decimal places and return as an array. If relevant data cannot be found, please answer "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[46.6, 2.93, 29.13, "Core overweight, active weight 2.91%"]`
+
+Scoring rules:
+- The gold answer is a list with N=4 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_3` each as 0 or 1.
+- Return `total = (sum(part_i)) / 4` exactly.
+- If the model output is missing or cannot be parsed into 4 comparable parts, score all parts 0.
+
diff --git a/tasks/task_436_international_comparison_hard_hard007.md b/tasks/task_436_international_comparison_hard_hard007.md
new file mode 100644
index 0000000000000000000000000000000000000000..b148cfc7da7121ef1cdf247de450e8074cc22c3c
--- /dev/null
+++ b/tasks/task_436_international_comparison_hard_hard007.md
@@ -0,0 +1,119 @@
+---
+id: task_436_international_comparison_hard_hard007
+name: international_comparison-hard-hard007
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/hard007.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+A global electric vehicle growth portfolio uses an "innovation offsets earnings deficit" framework for loss-making but high-R&D complete vehicle companies. For each candidate, first calculate: 1. Innovation excess = company R&D-to-revenue ratio − median R&D-to-revenue ratio among A-share automotive manufacturing firms in the top 10% by operating revenue in 2022; 2. Profit gap = median net profit margin among those top-10%-by-revenue A-share automotive manufacturing firms − company net profit margin; 3. Policy leverage = number of China automotive manufacturing industry policies ÷ number of provincial-level administrative regions covered by non-national policies. If innovation excess > 8, profit gap < 10, and policy leverage > 1.5, tactical overweight is allowed with active weight = min(2.00%, innovation excess ÷ 5 − profit gap ÷ 10 + policy leverage ÷ 10); otherwise only the watch list applies. Using Li Auto (LI) 2022 annual report and the local database, compute and state the most appropriate conclusion.
+
+Output guidelines:
+Answer in order: innovation excess (percentage points), profit gap (percentage points), policy leverage, most appropriate conclusion. Retain 2 decimal places and return as an array; the conclusion must specify position action and active weight. If relevant data cannot be found, please answer "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[10.19, 7.73, 3.45, "Tactical overweight, active weight 1.61%"]`
+
+Scoring rules:
+- The gold answer is a list with N=4 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_3` each as 0 or 1.
+- Return `total = (sum(part_i)) / 4` exactly.
+- If the model output is missing or cannot be parsed into 4 comparable parts, score all parts 0.
+
diff --git a/tasks/task_437_international_comparison_hard_hard008.md b/tasks/task_437_international_comparison_hard_hard008.md
new file mode 100644
index 0000000000000000000000000000000000000000..4b8534fb162e58ea560e830e81eae24364f868b1
--- /dev/null
+++ b/tasks/task_437_international_comparison_hard_hard008.md
@@ -0,0 +1,119 @@
+---
+id: task_437_international_comparison_hard_hard008
+name: international_comparison-hard-hard008
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/hard008.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+A global consumer defensive portfolio, when screening platform retailers, views fulfillment expenses as having quasi-fixed cost characteristics. For candidate companies, apply the following stress test: ① Fulfillment expense ratio = Fulfillment ÷ Net revenues; ② Stressed net profit margin = company net profit margin - 0.2 × fulfillment expense ratio; ③ Defense gap = stressed net profit margin - median net profit margin of A-share wholesale and retail industry top 10% by revenue in 2022. If stressed net profit margin < 0, do not include in defensive core position; if stressed net profit margin is between 0 and industry median, benchmark hold only; if above industry median, overweight is allowed. Using JD.com (JD) 2022 annual report and local database, calculate and determine the most appropriate conclusion.
+
+Output guidelines:
+Answer in order: fulfillment expense ratio (%), stressed net profit margin (%), defense gap (percentage points), most appropriate conclusion. Retain 2 decimal places and return as an array; conclusion must specify position action. If relevant data cannot be found, please answer "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[6.02, -0.28, -1.16, "Do not include in defensive core position"]`
+
+Scoring rules:
+- The gold answer is a list with N=4 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_3` each as 0 or 1.
+- Return `total = (sum(part_i)) / 4` exactly.
+- If the model output is missing or cannot be parsed into 4 comparable parts, score all parts 0.
+
diff --git a/tasks/task_438_international_comparison_hard_hard009.md b/tasks/task_438_international_comparison_hard_hard009.md
new file mode 100644
index 0000000000000000000000000000000000000000..f105330fdfc1bfee33420bf096ef7123721c8dce
--- /dev/null
+++ b/tasks/task_438_international_comparison_hard_hard009.md
@@ -0,0 +1,119 @@
+---
+id: task_438_international_comparison_hard_hard009
+name: international_comparison-hard-hard009
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/hard009.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+An active equity manager uses a "high-profit reinvestment" framework for the internet retail growth sub-portfolio. For candidate companies, first calculate: ① High-quality growth score = difference between company net profit margin and median net profit margin of A-share wholesale and retail industry top 10% by revenue in 2022 + 0.5 × transaction services revenue ratio + 0.5 × R&D investment ratio; ② Policy diffusion ratio = count of wholesale and retail industry policies ÷ number of provincial-level administrative regions covered by non-national policies. If high-quality growth score > 35 and policy diffusion ratio > 2.0, list as strategic overweight with active weight = min(4.00%, high-quality growth score ÷ 10); otherwise ordinary position only. Using Pinduoduo (PDD) 2022 annual report and local database, calculate and determine the most appropriate conclusion.
+
+Output guidelines:
+Answer in order: high-quality growth score, policy diffusion ratio, active weight (%), most appropriate conclusion. Retain 2 decimal places and return as an array; conclusion must specify position action. If relevant data cannot be found, please answer "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[37.73, 2.0, 0.0, "Ordinary position"]`
+
+Scoring rules:
+- The gold answer is a list with N=4 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_3` each as 0 or 1.
+- Return `total = (sum(part_i)) / 4` exactly.
+- If the model output is missing or cannot be parsed into 4 comparable parts, score all parts 0.
+
diff --git a/tasks/task_439_international_comparison_hard_hard010.md b/tasks/task_439_international_comparison_hard_hard010.md
new file mode 100644
index 0000000000000000000000000000000000000000..677159654272d515164fbcd47d1bcc06094baf7f
--- /dev/null
+++ b/tasks/task_439_international_comparison_hard_hard010.md
@@ -0,0 +1,119 @@
+---
+id: task_439_international_comparison_hard_hard010
+name: international_comparison-hard-hard010
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/hard010.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+A private wealth global consumption themed account wishes to incorporate the feature of 'overseas cash flow hedging domestic cycle' for China optional consumption. For candidate companies, apply the following rules: ① Overseas Hedging Quality Score = the difference between the company's net profit margin and the median net profit margin of A-share wholesale and retail industry in 2022 + 0.5 × international market revenue share; ② If the score ≥ 15, list as core holding, with active weight = min(3.00%, Overseas Hedging Quality Score ÷ 8); otherwise, only satellite holding is permitted. Based on MINISO (MNSO) FY2022 annual report and local database, calculate and determine the most appropriate conclusion.
+
+Output guidelines:
+Answer in order: International market revenue share (%), the difference between net profit margin and the median net profit margin of A-share wholesale and retail industry (percentage points), Overseas Hedging Quality Score, and the most appropriate conclusion. Retain 2 decimal places for numerical values and return as an array. The conclusion must include position action and active weight. Example: ["26.20", "4.79", "17.89", "Core holding, active weight 2.24%"]. If relevant data cannot be found, respond with "Relevant data not found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[26.2, 4.38, 17.48, "Core holding, active weight 2.18%"]`
+
+Scoring rules:
+- The gold answer is a list with N=4 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_3` each as 0 or 1.
+- Return `total = (sum(part_i)) / 4` exactly.
+- If the model output is missing or cannot be parsed into 4 comparable parts, score all parts 0.
+
diff --git a/tasks/task_440_international_comparison_hard_hard011.md b/tasks/task_440_international_comparison_hard_hard011.md
new file mode 100644
index 0000000000000000000000000000000000000000..c02b9dc7c71d60f7a7ffc8231934bf34b5bf6c14
--- /dev/null
+++ b/tasks/task_440_international_comparison_hard_hard011.md
@@ -0,0 +1,119 @@
+---
+id: task_440_international_comparison_hard_hard011
+name: international_comparison-hard-hard011
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/hard011.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+When an industrial internet fund evaluation distribution platform transforms toward high value-added services, it adopts a two-step method: ? Intangible input excess = company's R&D investment ratio - median R&D investment ratio of the A-share wholesale and retail industry in 2022; ? Transformation score = net service revenue ratio + intangible input excess. Only when net service revenue ratio >= 5 and transformation score >= 5 can a company enter the platform-based overweight list; otherwise, it enters the watch list. Based on ZKH's 2022 annual report and local database, calculate and determine the most appropriate conclusion.
+
+Output guidelines:
+Answer in order: Net service revenue ratio (%), intangible input excess (percentage points), transformation score, and the most appropriate conclusion. Retain 2 decimal places for numerical values and return as an array. The conclusion must clearly indicate whether to enter the platform-based overweight list or the watch list. If relevant data cannot be found, respond with "Relevant data not found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["2.16", "2.49", "4.65", "Watch list"]`
+
+Scoring rules:
+- The gold answer is a list with N=4 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_3` each as 0 or 1.
+- Return `total = (sum(part_i)) / 4` exactly.
+- If the model output is missing or cannot be parsed into 4 comparable parts, score all parts 0.
+
diff --git a/tasks/task_441_international_comparison_hard_hard012.md b/tasks/task_441_international_comparison_hard_hard012.md
new file mode 100644
index 0000000000000000000000000000000000000000..3089a5b030fefaaa54e669ec250248197df2c403
--- /dev/null
+++ b/tasks/task_441_international_comparison_hard_hard012.md
@@ -0,0 +1,119 @@
+---
+id: task_441_international_comparison_hard_hard012
+name: international_comparison-hard-hard012
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/hard012.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+A real estate transformation special account evaluates companies that replace development cycles with existing property services. For candidate companies, apply the following rules: ① Transformation buffer = home renovation and furnishing revenue ratio + R&D investment ratio; ② Profit gap = median net profit margin of top 10% A-share real estate companies by revenue in 2022 - company net profit margin; ③ Net transformation score = transformation buffer - profit gap. If net transformation score ≤ 0, exclude; if 0 < net transformation score < 5, only tactical small overweight is permitted, with active weight = min(1.50%, net transformation score ÷ 10); if net transformation score ≥ 5, standard overweight is permitted. Based on KE Holdings (BEKE) 2022 annual report and local database, calculate and determine the most appropriate conclusion.
+
+Output guidelines:
+Answer in order: Transformation buffer, profit gap (percentage points), net transformation score, and the most appropriate conclusion. Retain 2 decimal places for numerical values and return as an array. The conclusion must include position action and active weight. If relevant data cannot be found, respond with "Relevant data not found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[12.51, 4.22, 8.29, "Standard overweight; the rules do not specify a concrete active weight percentage"]`
+
+Scoring rules:
+- The gold answer is a list with N=4 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_3` each as 0 or 1.
+- Return `total = (sum(part_i)) / 4` exactly.
+- If the model output is missing or cannot be parsed into 4 comparable parts, score all parts 0.
+
diff --git a/tasks/task_442_international_comparison_hard_hard013.md b/tasks/task_442_international_comparison_hard_hard013.md
new file mode 100644
index 0000000000000000000000000000000000000000..969fc7d50a8bcbd37da604fe311fedc6df76d889
--- /dev/null
+++ b/tasks/task_442_international_comparison_hard_hard013.md
@@ -0,0 +1,119 @@
+---
+id: task_442_international_comparison_hard_hard013
+name: international_comparison-hard-hard013
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/hard013.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+A quasi-infrastructure growth portfolio allows allocation to data center operators during accounting loss periods, but requires a significant operating profit buffer. For candidate companies, apply the following rules: ① Profit conversion penalty = |net profit margin| ÷ adjusted EBITDA margin × 100; ② Policy diffusion ratio = number of data center, Eastern Data Western Computing, or computing power related policies ÷ number of provincial-level administrative regions covered by non-national policies; ③ Infrastructure capacity score = adjusted EBITDA margin - profit conversion penalty + 5 × policy diffusion ratio. If profit conversion penalty < 35 and infrastructure capacity score > 25, the company may be listed for satellite overweight 1.00%; otherwise, benchmark allocation only. Based on GDS Holdings (GDS) 2022 annual report and local database, calculate and determine the most appropriate conclusion.
+
+Output guidelines:
+Answer in order: Profit conversion penalty, policy diffusion ratio, infrastructure capacity score, and the most appropriate conclusion. Retain 2 decimal places for numerical values and return as an array. The conclusion must clearly indicate position action. If relevant data cannot be found, respond with "Relevant data not found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[29.77, 1.25, 22.08, "Benchmark allocation only"]`
+
+Scoring rules:
+- The gold answer is a list with N=4 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_3` each as 0 or 1.
+- Return `total = (sum(part_i)) / 4` exactly.
+- If the model output is missing or cannot be parsed into 4 comparable parts, score all parts 0.
+
diff --git a/tasks/task_443_international_comparison_hard_hard014.md b/tasks/task_443_international_comparison_hard_hard014.md
new file mode 100644
index 0000000000000000000000000000000000000000..fce4abce28105232031c04ce00cef2486b6c12f8
--- /dev/null
+++ b/tasks/task_443_international_comparison_hard_hard014.md
@@ -0,0 +1,119 @@
+---
+id: task_443_international_comparison_hard_hard014
+name: international_comparison-hard-hard014
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/hard014.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+A distressed-reversal fund applies a "screen first, value second" rule to platform retail stocks. For each candidate company, first compute: ① Survival score = technology service revenue ratio + R&D intensity - 0.1×|company net margin - median net margin of A-share wholesale and retail industry in 2022|; ② If company net margin is below -100%, trigger one-vote veto and remove directly; otherwise, only enter the watch pool when survival score ≥ 20. Based on Mogujie (MOGU) annual report for the fiscal year ended March 31, 2022 and local database, calculate and determine the most appropriate conclusion.
+
+Output guidelines:
+Answer in order: technology service revenue ratio (%), R&D intensity (%), survival score, and most appropriate conclusion. Values rounded to 2 decimal places; conclusion must clearly state whether the stock is removed; return as an array. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[13.65, 24.49, 18.95, "Remove"]`
+
+Scoring rules:
+- The gold answer is a list with N=4 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_3` each as 0 or 1.
+- Return `total = (sum(part_i)) / 4` exactly.
+- If the model output is missing or cannot be parsed into 4 comparable parts, score all parts 0.
+
diff --git a/tasks/task_444_international_comparison_medium_medium001.md b/tasks/task_444_international_comparison_medium_medium001.md
new file mode 100644
index 0000000000000000000000000000000000000000..118a0a5f2676d33fc1675ffe3cb79614c5bcf1de
--- /dev/null
+++ b/tasks/task_444_international_comparison_medium_medium001.md
@@ -0,0 +1,119 @@
+---
+id: task_444_international_comparison_medium_medium001
+name: international_comparison-medium-medium001
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium001.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What was NetEase's R&D intensity (R&D expenses as a percentage of revenue) in 2022? Compared with the median R&D intensity of listed companies in China's information transmission, software and IT services industry, how many percentage points higher or lower is it?
+
+Output guidelines:
+Answer both sub-questions: 1) NetEase 2022 R&D intensity (2 decimal places, %); 2) Difference from industry median (2 decimal places, percentage points; positive value means above industry median; return as an array). If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[15.59, 3.77]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_445_international_comparison_medium_medium002.md b/tasks/task_445_international_comparison_medium_medium002.md
new file mode 100644
index 0000000000000000000000000000000000000000..c65568e41d6da733d69d08e2e5835d1be84b27cd
--- /dev/null
+++ b/tasks/task_445_international_comparison_medium_medium002.md
@@ -0,0 +1,119 @@
+---
+id: task_445_international_comparison_medium_medium002
+name: international_comparison-medium-medium002
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium002.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the net profit margin (net profit / operating revenue × 100%) of Trip.com Group according to its 2022 annual report? What is the difference in percentage points compared with the median net profit margin of listed companies in China's transport, storage and postal services industry?
+
+Output guidelines:
+Answer in order: Trip.com net profit margin (%), and the difference from domestic industry median (percentage points; negative value means Trip.com is below industry median). Values rounded to 2 decimal places, returned as an array, e.g. [-3.50, -1.20]. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[6.82, -1.46]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_446_international_comparison_medium_medium003.md b/tasks/task_446_international_comparison_medium_medium003.md
new file mode 100644
index 0000000000000000000000000000000000000000..d9d11fb3a2a9dab96d50ff7001f41be61a0e8a59
--- /dev/null
+++ b/tasks/task_446_international_comparison_medium_medium003.md
@@ -0,0 +1,119 @@
+---
+id: task_446_international_comparison_medium_medium003
+name: international_comparison-medium-medium003
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium003.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the revenue per employee (operating revenue / total employees, unit: ten thousand CNY) of Bilibili according to its 2022 annual report? Compared with the median revenue per employee of listed companies in China's "information transmission, software and IT services" industry, how many times the industry median is Bilibili's revenue per employee?
+
+Output guidelines:
+Answer in order: Bilibili revenue per employee (ten thousand CNY), and the multiple of industry median. Values rounded to 2 decimal places, e.g. [197.43, 2.16]. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[197.43, 2.16]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_447_international_comparison_medium_medium004.md b/tasks/task_447_international_comparison_medium_medium004.md
new file mode 100644
index 0000000000000000000000000000000000000000..1483b98b62d18dd1b90268f8bcd8562379944428
--- /dev/null
+++ b/tasks/task_447_international_comparison_medium_medium004.md
@@ -0,0 +1,119 @@
+---
+id: task_447_international_comparison_medium_medium004
+name: international_comparison-medium-medium004
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium004.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is the net profit margin (net profit ÷ operating revenue × 100%) of Vipshop according to its 2022 annual report? How many percentage points higher is it compared with the median net profit margin of listed companies in China's wholesale and retail industry?
+
+Output guidelines:
+Answer in order: Vipshop net profit margin (%, 2 decimal places) and the difference from domestic industry median (percentage points, 2 decimal places; positive value means Vipshop is higher), e.g. [6.12, 4.57]. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[6.12, 4.57]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_448_international_comparison_medium_medium005.md b/tasks/task_448_international_comparison_medium_medium005.md
new file mode 100644
index 0000000000000000000000000000000000000000..32d728f21896ab85dfb29c1c2d2cf20abc2ca139
--- /dev/null
+++ b/tasks/task_448_international_comparison_medium_medium005.md
@@ -0,0 +1,119 @@
+---
+id: task_448_international_comparison_medium_medium005
+name: international_comparison-medium-medium005
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium005.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+ZTO Express is a profit leader in China's express delivery industry. Based on ZTO Express 2022 annual report, answer in order: (1) ZTO Express 2022 net profit per employee (net profit ÷ total employees, unit: ten thousand CNY); (2) Median net profit per employee of listed companies in China's transport, storage and postal services industry (ten thousand CNY); (3) How many times the industry median is ZTO Express net profit per employee?
+
+Output guidelines:
+Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [26.76, 13.26, 2.02]. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[26.76, 13.26, 2.02]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_449_international_comparison_medium_medium006.md b/tasks/task_449_international_comparison_medium_medium006.md
new file mode 100644
index 0000000000000000000000000000000000000000..074c9ab4656dab80fc5743917eef0960c3889efc
--- /dev/null
+++ b/tasks/task_449_international_comparison_medium_medium006.md
@@ -0,0 +1,119 @@
+---
+id: task_449_international_comparison_medium_medium006
+name: international_comparison-medium-medium006
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium006.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+For MINISO (MNSO) FY2022 (ended June 30, 2022), answer in order: (1) What is the total asset turnover ratio? (2) What is the median total asset turnover ratio of listed companies in China's wholesale and retail industry in 2022? (3) What is the difference between the two (positive value means industry median is higher)?
+
+Output guidelines:
+Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [0.89, 1.02, 0.13]. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[0.89, 1.02, 0.13]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_450_international_comparison_medium_medium007.md b/tasks/task_450_international_comparison_medium_medium007.md
new file mode 100644
index 0000000000000000000000000000000000000000..9e58386f79b59085e581978703ee01c0d7d4013c
--- /dev/null
+++ b/tasks/task_450_international_comparison_medium_medium007.md
@@ -0,0 +1,119 @@
+---
+id: task_450_international_comparison_medium_medium007
+name: international_comparison-medium-medium007
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium007.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+H World Group (HTHT) is one of China's leading hotel groups. Based on H World Group 2022 annual report, answer in order: (1) Revenue per employee (total operating revenue ÷ total employees, unit: ten thousand CNY); (2) Median revenue per employee of listed companies in China's accommodation and catering industry (ten thousand CNY); (3) How many times the industry median is H World Group's revenue per employee?
+
+Output guidelines:
+Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [56.96, 41.29, 1.38]. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[56.96, 41.29, 1.38]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_451_international_comparison_medium_medium008.md b/tasks/task_451_international_comparison_medium_medium008.md
new file mode 100644
index 0000000000000000000000000000000000000000..f01bcd28a36c9a52397dfbcabb3e2161a47be582
--- /dev/null
+++ b/tasks/task_451_international_comparison_medium_medium008.md
@@ -0,0 +1,119 @@
+---
+id: task_451_international_comparison_medium_medium008
+name: international_comparison-medium-medium008
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium008.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Yum China (YUMC) is one of China's largest restaurant chains, with brands including KFC and Pizza Hut. Based on Yum China 2022 annual report, answer in order: (1) Net profit margin (net profit ÷ operating revenue, %); (2) Median net profit margin of listed companies in China's accommodation and catering industry (%); (3) How many percentage points higher is Yum China's net profit margin than that median?
+
+Output guidelines:
+Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [4.62, -16.17, 20.79]. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[4.62, -16.17, 20.79]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_452_international_comparison_medium_medium009.md b/tasks/task_452_international_comparison_medium_medium009.md
new file mode 100644
index 0000000000000000000000000000000000000000..69516b5bf89f0b9300207d259e928d55f2e76c11
--- /dev/null
+++ b/tasks/task_452_international_comparison_medium_medium009.md
@@ -0,0 +1,119 @@
+---
+id: task_452_international_comparison_medium_medium009
+name: international_comparison-medium-medium009
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium009.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What is ASE Technology Holding's net profit margin in 2022? Among listed companies in China's semiconductor industry, the top 10% by operating revenue (count rounded up) are defined as industry leaders. What is ASE Technology Holding's net profit margin? How many percentage points does it differ from the median net profit margin of industry leaders?
+
+Output guidelines:
+Answer in order: 1) ASE Technology Holding's 2022 net profit margin (2 decimal places, %); 2) Difference between ASE's net profit margin and the median net profit margin of domestic semiconductor industry leaders (top 10% by revenue) in percentage points (2 decimal places). Return as a list. If relevant data cannot be found, answer "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[9.17, -0.43]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_453_international_comparison_medium_medium010.md b/tasks/task_453_international_comparison_medium_medium010.md
new file mode 100644
index 0000000000000000000000000000000000000000..fe649571adaafeceed1742f4068dbbf91d792df7
--- /dev/null
+++ b/tasks/task_453_international_comparison_medium_medium010.md
@@ -0,0 +1,119 @@
+---
+id: task_453_international_comparison_medium_medium010
+name: international_comparison-medium-medium010
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium010.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+What was Futu Holdings' (FUTU) net profit per employee in 2022, in ten thousand CNY? What is the multiple of Futu's net profit per employee compared with the median net profit per employee of listed companies in China's capital market services industry (securities and diversified financials)?
+
+Output guidelines:
+Answer both sub-questions: 1) Futu Holdings 2022 net profit per employee (2 decimal places, ten thousand CNY; convert HKD to CNY using 2022 average rate 1 HKD ≈ 0.86 RMB); 2) Multiple of Futu net profit per employee vs domestic capital market services industry median (2 decimal places). If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[90.42, 5.62]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_454_international_comparison_medium_medium011.md b/tasks/task_454_international_comparison_medium_medium011.md
new file mode 100644
index 0000000000000000000000000000000000000000..cdbdb95e5046d36ffa916e9b61a511ae98a18d4d
--- /dev/null
+++ b/tasks/task_454_international_comparison_medium_medium011.md
@@ -0,0 +1,119 @@
+---
+id: task_454_international_comparison_medium_medium011
+name: international_comparison-medium-medium011
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium011.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Atour Lifestyle Group (ATAT) is a representative mid-to-high-end chain hotel brand in China. Based on Atour 2022 annual report, answer in order: (1) Asset-liability ratio (total liabilities ÷ total assets, %); (2) Median asset-liability ratio of listed companies in China's accommodation and catering industry (%); (3) How many percentage points higher is Atour's asset-liability ratio than that median?
+
+Output guidelines:
+Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [75.07, 63.83, 11.24]. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[75.07, 63.83, 11.24]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_455_international_comparison_medium_medium012.md b/tasks/task_455_international_comparison_medium_medium012.md
new file mode 100644
index 0000000000000000000000000000000000000000..ce97298b1fb72bcfdf47950a1685b4c9a9f36e43
--- /dev/null
+++ b/tasks/task_455_international_comparison_medium_medium012.md
@@ -0,0 +1,119 @@
+---
+id: task_455_international_comparison_medium_medium012
+name: international_comparison-medium-medium012
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium012.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Full Truck Alliance (NYSE: YMM) is a leading digital freight platform in China. Based on Full Truck Alliance 2022 annual report, answer in order: (1) R&D intensity (R&D expenses ÷ operating revenue, %); (2) Median R&D intensity of listed companies in China's transport, storage and postal services industry (%); (3) How many times the industry median is Full Truck Alliance's R&D intensity?
+
+Output guidelines:
+Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [13.58, 0.56, 24.25]. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[13.58, 0.56, 24.25]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_456_international_comparison_medium_medium013.md b/tasks/task_456_international_comparison_medium_medium013.md
new file mode 100644
index 0000000000000000000000000000000000000000..8936972a9a16136e3e8a751e0c9691dfe26449e9
--- /dev/null
+++ b/tasks/task_456_international_comparison_medium_medium013.md
@@ -0,0 +1,119 @@
+---
+id: task_456_international_comparison_medium_medium013
+name: international_comparison-medium-medium013
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium013.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Himax Technologies (HIMX) is a leading display driver IC design company globally. Based on Himax 2022 annual report, answer in order: (1) R&D intensity (R&D expenses as percentage of operating revenue, %); (2) Median R&D intensity of listed semiconductor companies in China (%); (3) How many times the domestic semiconductor industry median is Himax's R&D intensity?
+
+Output guidelines:
+Answer the three sub-questions in order; values rounded to 2 decimal places, e.g. [14.61, 7.17, 2.04]. Reasoning may be shown in the process, but final answer should include only these three items. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[14.61, 7.17, 2.04]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_457_international_comparison_medium_medium014.md b/tasks/task_457_international_comparison_medium_medium014.md
new file mode 100644
index 0000000000000000000000000000000000000000..3241609cb6a8516c9c5c765e92603404e2b52da6
--- /dev/null
+++ b/tasks/task_457_international_comparison_medium_medium014.md
@@ -0,0 +1,119 @@
+---
+id: task_457_international_comparison_medium_medium014
+name: international_comparison-medium-medium014
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium014.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Weibo (WB / Weibo Corporation) is a leading social media platform in China. Based on Weibo 2022 annual report, compute its net profit margin (net profit attributable to Weibo shareholders ÷ operating revenue, in percentage), and conduct peer analysis against listed companies in China's information transmission, software and IT services industry. Rank by operating revenue; top 10% are defined as industry leaders. Answer in order: Weibo's net profit margin, and the difference (in percentage points) between Weibo's net profit margin and the median net profit margin of industry leaders.
+
+Output guidelines:
+Answer in order: Weibo 2022 net profit margin (%), and the difference between Weibo and industry leaders (top 10% by revenue) median net profit margin (percentage points). Values rounded to 2 decimal places, e.g. [4.66, 1.21]. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[4.66, 1.21]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_458_international_comparison_medium_medium015.md b/tasks/task_458_international_comparison_medium_medium015.md
new file mode 100644
index 0000000000000000000000000000000000000000..0fb5eb4eac4020022f227631db6c04f0e0c22761
--- /dev/null
+++ b/tasks/task_458_international_comparison_medium_medium015.md
@@ -0,0 +1,119 @@
+---
+id: task_458_international_comparison_medium_medium015
+name: international_comparison-medium-medium015
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium015.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Delta Electronics (2308.TW) is a leading global provider of power and thermal management solutions. Based on Delta Electronics' 2022 annual report, compute its consolidated net profit margin (net profit / operating revenue × 100%) and benchmark against listed companies in China's Electrical Machinery and Equipment Manufacturing industry: the top 10% by operating revenue are defined as industry leaders. What is Delta Electronics' net profit margin? What is the gap (in percentage points) between Delta's net profit margin and the median net profit margin of industry leaders?
+
+Output guidelines:
+Answer in order: Delta Electronics 2022 consolidated net profit margin (%), and the gap (percentage points) between Delta Electronics and the median net profit margin of domestic Electrical Machinery and Equipment Manufacturing industry leaders (top 10% by revenue). Round all values to 2 decimal places, e.g. [9.62, 3.30]. If relevant data cannot be found, answer "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[9.62, 3.3]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_459_international_comparison_medium_medium016.md b/tasks/task_459_international_comparison_medium_medium016.md
new file mode 100644
index 0000000000000000000000000000000000000000..3768482806846ef128bb6c597bfcf870c00b1717
--- /dev/null
+++ b/tasks/task_459_international_comparison_medium_medium016.md
@@ -0,0 +1,119 @@
+---
+id: task_459_international_comparison_medium_medium016
+name: international_comparison-medium-medium016
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium016.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, amid strong policy support for domestic semiconductor self-reliance in China, what was TSMC's R&D intensity (R&D expenses as percentage of revenue) in its annual report? What is the difference in percentage points compared with the median R&D intensity of listed semiconductor companies in China?
+
+Output guidelines:
+Answer in order: TSMC R&D intensity (%), and the difference from domestic median (percentage points; positive value means TSMC is higher). Values rounded to 2 decimal places. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[7.21, 0.04]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_460_international_comparison_medium_medium017.md b/tasks/task_460_international_comparison_medium_medium017.md
new file mode 100644
index 0000000000000000000000000000000000000000..241f8b588e44f7badd52632ae8ea1360018ca521
--- /dev/null
+++ b/tasks/task_460_international_comparison_medium_medium017.md
@@ -0,0 +1,119 @@
+---
+id: task_460_international_comparison_medium_medium017
+name: international_comparison-medium-medium017
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium017.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, under national policies promoting innovative drug R&D, what was BeiGene's R&D expense as a percentage of revenue in its annual report? How many percentage points does that ratio differ from the median R&D-to-revenue ratio among domestic pharmaceutical manufacturing firms that are (1) private enterprises and (2) state-owned enterprises (including centrally administered SOEs, locally administered SOEs, and institute-type SOEs), after excluding zeros and invalid data? What is the gap between the median R&D ratios of private vs. state-owned domestic pharmaceutical firms?
+
+Output guidelines:
+Answer in order: BeiGene's R&D-to-revenue ratio (%); gap vs. private enterprises' median (percentage points); gap vs. state-owned enterprises' median (percentage points); gap between private and state-owned medians (percentage points). Use 2 decimal places; return as a list. If relevant data cannot be found, answer "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[115.86, 108.4, 111.23, 2.83]`
+
+Scoring rules:
+- The gold answer is a list with N=4 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_3` each as 0 or 1.
+- Return `total = (sum(part_i)) / 4` exactly.
+- If the model output is missing or cannot be parsed into 4 comparable parts, score all parts 0.
+
diff --git a/tasks/task_461_international_comparison_medium_medium018.md b/tasks/task_461_international_comparison_medium_medium018.md
new file mode 100644
index 0000000000000000000000000000000000000000..d9c2a812f9d4011111d0c22c2b79e5ae584053bb
--- /dev/null
+++ b/tasks/task_461_international_comparison_medium_medium018.md
@@ -0,0 +1,119 @@
+---
+id: task_461_international_comparison_medium_medium018
+name: international_comparison-medium-medium018
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium018.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, amid China's new energy vehicle industry policy support, what was XPeng's net profit margin (net profit ÷ operating revenue × 100%) in its annual report? What is the difference in percentage points compared with the median net profit margin of domestic automotive manufacturing industry leaders (top 10% by revenue)?
+
+Output guidelines:
+Answer in order: XPeng net profit margin (%) and the difference from industry leaders median (percentage points; negative value means XPeng is lower). Values rounded to 2 decimal places, e.g. [-34.03, -37.27]. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[-34.03, -37.27]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_462_international_comparison_medium_medium019.md b/tasks/task_462_international_comparison_medium_medium019.md
new file mode 100644
index 0000000000000000000000000000000000000000..5927254ffb02778deb731b2caa164d39108f919a
--- /dev/null
+++ b/tasks/task_462_international_comparison_medium_medium019.md
@@ -0,0 +1,119 @@
+---
+id: task_462_international_comparison_medium_medium019
+name: international_comparison-medium-medium019
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium019.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, amid policy support for digital economy and e-commerce development in China, what was PDD Holdings' net profit margin (net profit ÷ operating revenue × 100%) in its annual report? How many percentage points higher is it compared with the median net profit margin of listed companies in China's wholesale and retail industry?
+
+Output guidelines:
+Answer in order: PDD Holdings net profit margin (%), and the number of percentage points above domestic median. Values rounded to 2 decimal places, e.g. [24.16, 22.60]. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[24.16, 22.6]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_463_international_comparison_medium_medium020.md b/tasks/task_463_international_comparison_medium_medium020.md
new file mode 100644
index 0000000000000000000000000000000000000000..cd7c1d9f12918029f0ac6c13d10fbbdfb023ed9c
--- /dev/null
+++ b/tasks/task_463_international_comparison_medium_medium020.md
@@ -0,0 +1,119 @@
+---
+id: task_463_international_comparison_medium_medium020
+name: international_comparison-medium-medium020
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium020.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, amid national carbon peaking and clean energy development policies, what was JinkoSolar's asset-liability ratio (total liabilities ÷ total assets × 100%) in its annual report? Compared with the median asset-liability ratios among listed companies in China's Electricity, Heat, Gas and Water Supply industry—separately for state-owned enterprises (including centrally administered SOEs, locally administered SOEs, and other SOEs) and for private enterprises—how many percentage points higher is JinkoSolar's ratio in each case?
+
+Output guidelines:
+Answer in order: JinkoSolar's asset-liability ratio (%); percentage points above the state-owned median; percentage points above the private enterprise median. Use 2 decimal places; return as an array. If relevant data cannot be found, answer "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[75.15, 13.88, 22.57]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_464_international_comparison_medium_medium021.md b/tasks/task_464_international_comparison_medium_medium021.md
new file mode 100644
index 0000000000000000000000000000000000000000..b04ce68edd9316ab667682d744a60bbd89e2ae3f
--- /dev/null
+++ b/tasks/task_464_international_comparison_medium_medium021.md
@@ -0,0 +1,119 @@
+---
+id: task_464_international_comparison_medium_medium021
+name: international_comparison-medium-medium021
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium021.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, amid policy support for digital economy development and healthy platform economy regulation in China, what was Alibaba's revenue per employee (operating revenue ÷ total employees) in ten thousand CNY? How many times the median revenue per employee of domestic IT industry leaders (top 10% (round down) by operating revenue in information transmission, software and IT services) is it?
+
+Output guidelines:
+Answer in order: Alibaba revenue per employee (ten thousand CNY), and the multiple of industry leaders median revenue per employee. Values rounded to 2 decimal places, returned as an array. If relevant data cannot be found, reply "No relevant data found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[369.31, 1.5]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_465_international_comparison_medium_medium022.md b/tasks/task_465_international_comparison_medium_medium022.md
new file mode 100644
index 0000000000000000000000000000000000000000..8daf3391ff222c508be90d1bbb6ff8e49d8ab75e
--- /dev/null
+++ b/tasks/task_465_international_comparison_medium_medium022.md
@@ -0,0 +1,119 @@
+---
+id: task_465_international_comparison_medium_medium022
+name: international_comparison-medium-medium022
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium022.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, against the backdrop of intensive policy rollouts for the new energy vehicle industry, what was the proportion of R&D personnel to total employees in NIO's annual report? By how many percentage points did it differ from the median R&D personnel ratio among private enterprises and state-owned enterprises (including central and local state-owned enterprises) in China's automobile manufacturing industry?
+
+Output guidelines:
+Answer in order: NIO's R&D personnel ratio (%), the number of percentage points above the private enterprise median, and the number of percentage points above the state-owned enterprise median. Retain 2 decimal places for values. E.g. [37.46, 24.19, 22.02]. If the relevant data cannot be found, answer "Relevant data not found"
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[37.46, 24.19, 22.02]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_466_international_comparison_medium_medium023.md b/tasks/task_466_international_comparison_medium_medium023.md
new file mode 100644
index 0000000000000000000000000000000000000000..7589a51e2592d53e854ad72fdfb88320dab23388
--- /dev/null
+++ b/tasks/task_466_international_comparison_medium_medium023.md
@@ -0,0 +1,119 @@
+---
+id: task_466_international_comparison_medium_medium023
+name: international_comparison-medium-medium023
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium023.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Against the backdrop of the state promoting high-quality development of the digital economy and platform economy, what were JD.com's total revenue and total number of employees disclosed in its 2022 annual report? What is the per capita revenue in 10,000 yuan calculated therefrom? Among industries with no fewer than 30 domestic enterprises (excluding financial, real estate and diversified industries), how many times is JD.com's per capita revenue compared to the industry with the highest median per capita revenue?
+
+Output guidelines:
+Please answer in order: (1) JD.com's 2022 total revenue (100 million yuan, 2 decimal places); (2) JD.com's 2022 total employees (persons); (3) JD.com's per capita revenue (10,000 yuan, 2 decimal places); (4) The ratio of JD.com's per capita revenue to the industry with highest median per capita revenue (2 decimal places). Return as an array. If the relevant data cannot be found, answer "Relevant data not found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[10462.36, 450679, 232.15, 0.69]`
+
+Scoring rules:
+- The gold answer is a list with N=4 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_3` each as 0 or 1.
+- Return `total = (sum(part_i)) / 4` exactly.
+- If the model output is missing or cannot be parsed into 4 comparable parts, score all parts 0.
+
diff --git a/tasks/task_467_international_comparison_medium_medium024.md b/tasks/task_467_international_comparison_medium_medium024.md
new file mode 100644
index 0000000000000000000000000000000000000000..3c25174d7b84205dbf3fca28369d137a7cd125bd
--- /dev/null
+++ b/tasks/task_467_international_comparison_medium_medium024.md
@@ -0,0 +1,119 @@
+---
+id: task_467_international_comparison_medium_medium024
+name: international_comparison-medium-medium024
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium024.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Against the backdrop of national policies promoting high-level AI application and supporting digital economy enterprises in strengthening hard-tech innovation, what was Baidu's (BIDU, listed on NASDAQ) R&D intensity (R&D expenses as a percentage of revenue) in 2022? Compared with A-share listed companies in China's information transmission, software, and information technology services industry, when divided into leading enterprises (revenue ≥ 90th percentile) and non-leading enterprises, by how many percentage points did Baidu's R&D intensity exceed the median R&D ratio of each group?
+
+Output guidelines:
+Answer in order: Baidu's R&D intensity (%), the number of percentage points above the A-share leading group (revenue ≥ sample 90th percentile) median R&D ratio, and the number of percentage points above the A-share non-leading group. Retain 2 decimal places, return as an array. If the relevant data cannot be found, answer "Relevant data not found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[18.85, 12.59, 5.79]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_468_international_comparison_medium_medium025.md b/tasks/task_468_international_comparison_medium_medium025.md
new file mode 100644
index 0000000000000000000000000000000000000000..ffd846bdd3af53c3f07afe09039383aa5b00dc1c
--- /dev/null
+++ b/tasks/task_468_international_comparison_medium_medium025.md
@@ -0,0 +1,119 @@
+---
+id: task_468_international_comparison_medium_medium025
+name: international_comparison-medium-medium025
+category: international_comparison
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/international_comparison/medium025.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Against the backdrop of intensive policy rollouts for the new energy vehicle industry, what was ZEEKR's asset turnover ratio (revenue divided by total assets) in 2022? How did it compare to the median asset turnover ratio of China's automobile manufacturing industry as a whole?
+
+Output guidelines:
+Answer in order: ZEEKR's asset turnover ratio (2 decimal places), and the difference from the industry median (2 decimal places; positive means ZEEKR is higher). Return as an array. If the relevant data cannot be found, answer "Relevant data not found".
+
+You may use files under `./database/` and web search.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`[1.64, 1.05]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_469_risk_assessment_hard_hard001.md b/tasks/task_469_risk_assessment_hard_hard001.md
new file mode 100644
index 0000000000000000000000000000000000000000..f0770c59032290823a60eabefe3cbdb4610ec0e1
--- /dev/null
+++ b/tasks/task_469_risk_assessment_hard_hard001.md
@@ -0,0 +1,117 @@
+---
+id: task_469_risk_assessment_hard_hard001
+name: risk_assessment-hard-hard001
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/hard001.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in the automobile manufacturing industry, focusing on provinces with a total of 8 or more relevant enterprises, what proportion of the qualifying provinces simultaneously meet both the 'high policy dependence' risk (government subsidies as a proportion of operating profit >30%) and the 'high market concentration' risk (operating revenue CR4>60%)?
+
+Output guidelines:
+Answer format: percentage value (2 decimal places). E.g. 25.67. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`20.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_470_risk_assessment_hard_hard002.md b/tasks/task_470_risk_assessment_hard_hard002.md
new file mode 100644
index 0000000000000000000000000000000000000000..10f0079bb07c337f963a331c41ca6d3ef1902776
--- /dev/null
+++ b/tasks/task_470_risk_assessment_hard_hard002.md
@@ -0,0 +1,117 @@
+---
+id: task_470_risk_assessment_hard_hard002
+name: risk_assessment-hard-hard002
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/hard002.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in the communication transmission equipment industry, assume that enterprises with R&D investment intensity below the national median for that industry are eliminated from the market within the next three years, and only enterprises with R&D investment intensity not below the national median remain. What is the ratio of remaining enterprises' operating revenue after elimination to operating revenue before elimination for the province with the highest such ratio?
+
+Output guidelines:
+Answer format: numeric value (2 decimal places, unit %). If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`100.0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_471_risk_assessment_hard_hard003.md b/tasks/task_471_risk_assessment_hard_hard003.md
new file mode 100644
index 0000000000000000000000000000000000000000..632c70850550dbb1e84adc9658da8db6fdbcd031
--- /dev/null
+++ b/tasks/task_471_risk_assessment_hard_hard003.md
@@ -0,0 +1,119 @@
+---
+id: task_471_risk_assessment_hard_hard003
+name: risk_assessment-hard-hard003
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/hard003.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，国家层面发布了数字经济发展相关规划，同时部分省份也配套出台了支持通信传输设备业的政策，形成中央-地方协同支持格局。在同时受益于上述两级政策支持、且通信传输设备业上市企业数量不低于8家的省份中（计算政府补贴依赖度时，仅纳入政府补贴金额、营业利润金额、营业收入金额三项同时有完整记录的企业，依赖度=省内企业政府补贴总额÷省内企业营业利润总额），哪个省份对政府补贴的财务依赖最为突出？进一步模拟：一旦该省所有通信传输设备企业同时遭遇50%补贴缩减，且利润等额受损，全省通信传输设备业的营业利润将萎缩多大比例？
+
+Output guidelines:
+依次回答省份名称、政府补贴依赖度和营业利润下降比例。依赖度和下降比例均以百分数表示，保留2位小数。如["广东省", 25.33, 12.67]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["北京市", 19.52, 9.76]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_472_risk_assessment_hard_hard004.md b/tasks/task_472_risk_assessment_hard_hard004.md
new file mode 100644
index 0000000000000000000000000000000000000000..3df3b921720fc0a7f126c07df57fcaa4dbf987dc
--- /dev/null
+++ b/tasks/task_472_risk_assessment_hard_hard004.md
@@ -0,0 +1,117 @@
+---
+id: task_472_risk_assessment_hard_hard004
+name: risk_assessment-hard-hard004
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/hard004.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, suppose an industrial policy researcher is screening vulnerable provinces for the chemical raw materials and chemical products manufacturing industry: she defines "government subsidy amount exceeding 5% of that province's industry total operating revenue" as excessive subsidy dependence, and "the single largest enterprise's operating revenue as a share of that province's industry total operating revenue above 40%" as market structure imbalance. Only among provinces with total enterprises not less than 12 (operating profit margin = total operating profit / total operating revenue × 100%), among provinces that simultaneously trigger both alerts, what is the operating profit margin of the province with the weakest profitability?
+
+Output guidelines:
+Answer format: percentage value (2 decimal places). E.g. "8.56". If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No relevant data found"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_473_risk_assessment_hard_hard005.md b/tasks/task_473_risk_assessment_hard_hard005.md
new file mode 100644
index 0000000000000000000000000000000000000000..a1e872595533bcd22b7751f08f1bd819873f871e
--- /dev/null
+++ b/tasks/task_473_risk_assessment_hard_hard005.md
@@ -0,0 +1,117 @@
+---
+id: task_473_risk_assessment_hard_hard005
+name: risk_assessment-hard-hard005
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/hard005.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+At the end of 2022, semiconductor industry analysts modeled 2023 scenarios: revenue across enterprises is projected to decline by 20%, while fixed operating costs (defined as operating revenue minus operating profit) rise by 15% from their baseline. Under these assumptions, compute the change in operating profit margin for each enterprise in Jiangsu Province, Guangdong Province, and Shanghai (original operating profit margin = operating profit / operating revenue × 100%; new operating profit margin = (new operating revenue − new operating cost) / new operating revenue × 100%; decline magnitude = original operating profit margin − new operating profit margin), then take the average decline per region—limited to enterprises with complete operating revenue and operating profit data and positive operating revenue. Among the three regions, which province/municipality has the largest average decline in enterprise operating profit margin?
+
+Output guidelines:
+Answer format: province or city name. E.g. "Zhejiang Province" or "Beijing". If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Guangdong"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_474_risk_assessment_hard_hard006.md b/tasks/task_474_risk_assessment_hard_hard006.md
new file mode 100644
index 0000000000000000000000000000000000000000..500339c06608b19cd62b04f397e6e4de3257122f
--- /dev/null
+++ b/tasks/task_474_risk_assessment_hard_hard006.md
@@ -0,0 +1,117 @@
+---
+id: task_474_risk_assessment_hard_hard006
+name: risk_assessment-hard-hard006
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/hard006.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in the comprehensive assessment of the semiconductor industry ecosystem, the scope is limited to provinces with total enterprises not less than 6. The assessment framework has three dimensions: ? "Industry Scale Foundation" (weight 30%): half the sum of each province's share of national semiconductor enterprise count and each province's share of national semiconductor operating revenue as the raw score for this dimension; ? "Policy Ecosystem Density" (weight 30%): each province's count of policies involving the Chinese terms for semiconductors, chips, or integrated circuits divided by the national total of such policies; ? "Technological Accumulation Depth" (weight 40%): each province's share of cumulative Chinese invention patent grants in the national total for the industry. Each dimension is min-max normalized (mapped to 0-100), then weighted to yield the comprehensive development potential index. Which province has the highest comprehensive score?
+
+Output guidelines:
+Answer format: province name. E.g. "Guangdong Province". If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"Shanghai"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_475_risk_assessment_hard_hard007.md b/tasks/task_475_risk_assessment_hard_hard007.md
new file mode 100644
index 0000000000000000000000000000000000000000..5b1348478cc0ba8c81c015fda82d4870f095426e
--- /dev/null
+++ b/tasks/task_475_risk_assessment_hard_hard007.md
@@ -0,0 +1,117 @@
+---
+id: task_475_risk_assessment_hard_hard007
+name: risk_assessment-hard-hard007
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/hard007.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, Guangdong Province's automobile manufacturing industry faces pressure from subsidy phase-out. A three-year phase-out path is set: compared with the actual subsidy amount in 2022, cut 20% in 2023, 40% in 2024, and 60% in 2025. For each Guangdong automobile manufacturing enterprise that has a recorded government reward/subsidy amount, its net profit loss in each year is exactly equal to that year's subsidy reduction; summing the losses for 2023–2025 gives the enterprise's cumulative net profit loss. Under this definition, how many hundred million yuan is the three-year total loss for the enterprise with the heaviest cumulative damage?
+
+Output guidelines:
+Answer format: a numeric value (2 decimal places). For example, "5.67" means a cumulative loss of 5.67 hundred million yuan. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"20.53"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_476_risk_assessment_hard_hard008.md b/tasks/task_476_risk_assessment_hard_hard008.md
new file mode 100644
index 0000000000000000000000000000000000000000..fe24929a006c12683b150d73ed8a40bb7a70baff
--- /dev/null
+++ b/tasks/task_476_risk_assessment_hard_hard008.md
@@ -0,0 +1,117 @@
+---
+id: task_476_risk_assessment_hard_hard008
+name: risk_assessment-hard-hard008
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/hard008.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, suppose the chemical raw materials and chemical products manufacturing industry faces dual pressure in 2023: environmental protection investment must increase by an amount equivalent to 8% of operating revenue, and raw material price increases lead to a 10% increase in existing costs. If existing costs account for 75% of operating revenue, among enterprises in this industry in Jiangsu Province, how many enterprises will have negative new operating profit?
+
+Output guidelines:
+Answer format: integer. E.g. 15. If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`0`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_477_risk_assessment_hard_hard009.md b/tasks/task_477_risk_assessment_hard_hard009.md
new file mode 100644
index 0000000000000000000000000000000000000000..5f24720d8b269cf019482a83a3eabb6c3fa5147c
--- /dev/null
+++ b/tasks/task_477_risk_assessment_hard_hard009.md
@@ -0,0 +1,117 @@
+---
+id: task_477_risk_assessment_hard_hard009
+name: risk_assessment-hard-hard009
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/hard009.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, build a comprehensive industry-level risk resilience scoring system covering all manufacturing industries with total enterprises not less than 25 (excluding Finance, Real Estate, and Conglomerate industries). The scoring framework has three dimensions with different weights: Asset Return Efficiency (weight 40%), measured as the industry's average operating profit per enterprise divided by average total assets per enterprise; R&D Investment Intensity (weight 30%), measured as the industry's average R&D investment per enterprise divided by average operating revenue per enterprise; and Financial Desensitization to Policy Subsidies (weight 30%), measured as the industry's total operating profit divided by (total government subsidies + 1)—adding 1 to the denominator avoids division by zero when subsidies are zero. Each of the three raw indicators is min-max normalized across industries, scaled to a 0–100 score, then weighted by the above weights to yield the final score. Under this system, what is the comprehensive score of the top-ranked industry?
+
+Output guidelines:
+Answer format: numeric value (2 decimal places). If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`65.11`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_478_risk_assessment_hard_hard010.md b/tasks/task_478_risk_assessment_hard_hard010.md
new file mode 100644
index 0000000000000000000000000000000000000000..34e93febf30a994207bac50488679f26509f61ca
--- /dev/null
+++ b/tasks/task_478_risk_assessment_hard_hard010.md
@@ -0,0 +1,117 @@
+---
+id: task_478_risk_assessment_hard_hard010
+name: risk_assessment-hard-hard010
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/hard010.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, to measure the resilience reserve of each province's pharmaceutical manufacturing industry under extreme external shocks, a research team established the following three-dimensional evaluation framework (including only provinces with total enterprises not less than 10): Dimension 1 "Policy Shield Thickness" (weight 35%): for each province, the number of local policies involving the Chinese terms for pharmaceuticals or bio-industry divided by the national total of such policies yields the raw relative policy density; Dimension 2 "Independent Profitability Buffer" (weight 35%): each province's total pharmaceutical manufacturing operating profit minus total government subsidies, divided by total assets, reflecting actual asset profitability excluding subsidies; Dimension 3 "Technological Innovation Reserve Thickness" (weight 30%): patent grants per unit of R&D investment by province, measuring R&D output efficiency. Each dimension is min-max normalized across provinces (mapped to 0-100), then weighted by the above weights to produce the comprehensive score. Among the top three ranked provinces, what is the average comprehensive score?
+
+Output guidelines:
+Answer format: numeric value (2 decimal places). If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`69.4`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_479_risk_assessment_hard_hard011.md b/tasks/task_479_risk_assessment_hard_hard011.md
new file mode 100644
index 0000000000000000000000000000000000000000..766b3ed96aad9e2373ed3db9d28596f68ce4e963
--- /dev/null
+++ b/tasks/task_479_risk_assessment_hard_hard011.md
@@ -0,0 +1,119 @@
+---
+id: task_479_risk_assessment_hard_hard011
+name: risk_assessment-hard-hard011
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/hard011.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，国家从消费品工业数字化转型、质量标准提升等多个维度出台了政策支持，部分省份也因势利导推出了地方消费电子产业相关政策方案。现聚焦受上述双重政策覆盖的省份（同时须满足：省内消费电子及电气业上市企业不少于5家，且全省总营业利润大于零）。在这批省份中，哪个省份的企业负债结构对利率最为敏感——即以全省总负债相对于总营业利润的倍率（总负债/总营业利润）来衡量，哪个省份的这一比值最高？在此基础上，若利率水平抬升2个百分点，以各企业总负债×2%估算其新增利息负担，全省新增利息成本总额将相当于该省消费电子及电气业总营业利润的百分之多少？
+
+Output guidelines:
+依次回答省份名称和额外利息成本占总营业利润的比例。比例以百分数表示，保留2位小数。如["广东省", 23.36]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["江西省", 60.36]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_480_risk_assessment_hard_hard012.md b/tasks/task_480_risk_assessment_hard_hard012.md
new file mode 100644
index 0000000000000000000000000000000000000000..1f00ff02a3a9559629729173dfce88a484fbbdd2
--- /dev/null
+++ b/tasks/task_480_risk_assessment_hard_hard012.md
@@ -0,0 +1,117 @@
+---
+id: task_480_risk_assessment_hard_hard012
+name: risk_assessment-hard-hard012
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/hard012.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, in the chemical raw materials and chemical products manufacturing industry, comprehensively assess each province's ability to withstand external shocks. What is the comprehensive score of the lowest-scoring province? (Assessment indicators: Profit Diversification 30%, Financial Safety Cushion 35%, Policy Buffer 35%. Profit Diversification = 1 − (largest enterprise operating profit / total operating profit) normalized score; Financial Safety Cushion = (total operating profit / total liabilities) normalized score; Policy Buffer = (total operating profit / (total government subsidies + 1)) normalized score; each dimension min-max normalized to 0–100; only provinces with total enterprises>=15 are included)
+
+Output guidelines:
+Answer format: numeric value (2 decimal places). If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`15.78`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_481_risk_assessment_hard_hard013.md b/tasks/task_481_risk_assessment_hard_hard013.md
new file mode 100644
index 0000000000000000000000000000000000000000..b32ea42e1036cd1036ed0987573b1d8191900c2a
--- /dev/null
+++ b/tasks/task_481_risk_assessment_hard_hard013.md
@@ -0,0 +1,117 @@
+---
+id: task_481_risk_assessment_hard_hard013
+name: risk_assessment-hard-hard013
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/hard013.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022 regional data for the general purpose equipment manufacturing industry, for the province with the highest total government reward and subsidy funds: comprehensively rate its industrial competitiveness along two strategic dimensions. "Industrial Upgrading Capacity" combines the province's average rank across three sub-indicators (R&D investment growth rate, year-on-year change in Chinese patent applications, R&D personnel share); "Industrial Base" combines the province's average rank across three sub-indicators (total enterprises, total operating revenue, subsidy efficiency = total operating revenue / total government subsidies). The overall comprehensive performance rank is the average of the two dimension ranks. Answer only: Can this province's comprehensive industrial performance rank among the top 5 nationally?
+
+Output guidelines:
+Answer format: "Yes" or "No". If relevant data cannot be found, please answer "No relevant data found"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`"No"`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_482_risk_assessment_medium_medium001.md b/tasks/task_482_risk_assessment_medium_medium001.md
new file mode 100644
index 0000000000000000000000000000000000000000..66cf1e23e4beaeea5187e44cc7d2becd154902c7
--- /dev/null
+++ b/tasks/task_482_risk_assessment_medium_medium001.md
@@ -0,0 +1,119 @@
+---
+id: task_482_risk_assessment_medium_medium001
+name: risk_assessment-medium-medium001
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/medium001.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+Based on 2022 data, consider the following policy effect scenario: For provinces that have promulgated local industrial policies whose titles contain the keywords "semiconductor" or "integrated circuit", policy support will accelerate their semiconductor industry enterprises' R&D expansion pace to 2× the current growth rate over the next 3 years; for provinces that have not yet issued such policies, R&D growth remains unchanged. Using the median year-on-year change in enterprise R&D investment by province as the baseline growth rate, projected with 3-year compound growth, which province will rank first nationwide in total semiconductor industry R&D investment by 2025? What is the projected amount?
+
+Output guidelines:
+Answer format: [Province name, value (2 decimal places, unit: yuan)]. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["Shanghai", 97732260069.03]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_483_risk_assessment_medium_medium002.md b/tasks/task_483_risk_assessment_medium_medium002.md
new file mode 100644
index 0000000000000000000000000000000000000000..0a936fad3462b77fa007eeaa039066fb445021dd
--- /dev/null
+++ b/tasks/task_483_risk_assessment_medium_medium002.md
@@ -0,0 +1,117 @@
+---
+id: task_483_risk_assessment_medium_medium002
+name: risk_assessment-medium-medium002
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/medium002.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, focusing on pharmaceutical manufacturing: among provinces included in the statistics (requiring total related enterprises in the province ≥ 10), if a province has R&D investment concentration CR3 greater than 60% and cumulative granted Chinese invention patent concentration CR3 also greater than 60%, classify that province as a high-risk "R&D–patent dual head concentration" province. How many provinces satisfy both dual-concentration conditions?
+
+Output guidelines:
+Answer format: an integer, e.g. "3". If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`3`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_484_risk_assessment_medium_medium003.md b/tasks/task_484_risk_assessment_medium_medium003.md
new file mode 100644
index 0000000000000000000000000000000000000000..2dc799b71f76046cb3a255e372ff4f5a42944b9d
--- /dev/null
+++ b/tasks/task_484_risk_assessment_medium_medium003.md
@@ -0,0 +1,117 @@
+---
+id: task_484_risk_assessment_medium_medium003
+name: risk_assessment-medium-medium003
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/medium003.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+In 2022, assume all national R&D tax incentives are cancelled. Among manufacturing industries (only industries with total enterprises ≥ 20 and complete R&D investment and operating revenue data, positive revenue, excluding "Financial Services", "Real Estate", and "Conglomerates"), assume a corporate income tax rate of 25%; R&D tax benefits include 100% additional deduction. What is the sum of the declines (in percentage points) for the three industries with the largest average net profit margin decline, where net profit margin impact = R&D × 100% × 25% / operating revenue × 100%?
+
+Output guidelines:
+Answer format: a numeric value (2 decimal places). For example, 15.67 means the sum of the three industries' declines is 15.67 percentage points. If relevant data cannot be found, answer "No relevant data found".
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Single-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`601.08`
+
+Scoring rules:
+- Judge semantic equivalence between the model final answer and the gold answer.
+- Return `scores` with one key `match` as 1 or 0.
+- Return `total` as 1.0 if equivalent, otherwise 0.0.
+
diff --git a/tasks/task_485_risk_assessment_medium_medium004.md b/tasks/task_485_risk_assessment_medium_medium004.md
new file mode 100644
index 0000000000000000000000000000000000000000..ba74737bd83183b0730a7e4a74d7dc7ae90d3fb1
--- /dev/null
+++ b/tasks/task_485_risk_assessment_medium_medium004.md
@@ -0,0 +1,119 @@
+---
+id: task_485_risk_assessment_medium_medium004
+name: risk_assessment-medium-medium004
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/medium004.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，《建材行业碳达峰实施方案》由多部委联合印发，各省也相继出台了碳达峰或节能减排地方行动方案，建材类企业由此面临双层合规压力。在同时处于上述两级政策约束之下的省份中，进一步筛选非金属矿物制品业样本：省内上市企业总数至少5家，且整体营业利润为正。对于符合条件的省份，若将碳排放合规成本设定为各企业运营成本（运营成本=营业收入-营业利润）的5%，并全部计入利润扣减项，那么，哪个省份的非金属矿物制品业总营业利润下降幅度最大？该省合规成本冲击后，总营业利润究竟会下降多少个百分点？
+
+Output guidelines:
+依次回答省份名称和营业利润下降百分比。百分比保留2位小数。如["湖南省", 28.63]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["安徽省", 45.74]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_486_risk_assessment_medium_medium005.md b/tasks/task_486_risk_assessment_medium_medium005.md
new file mode 100644
index 0000000000000000000000000000000000000000..d4d1d6cdc9a428db214945809854e0326602781f
--- /dev/null
+++ b/tasks/task_486_risk_assessment_medium_medium005.md
@@ -0,0 +1,119 @@
+---
+id: task_486_risk_assessment_medium_medium005
+name: risk_assessment-medium-medium005
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/medium005.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，相关部委出台了促进钢铁工业高质量发展的指导意见，明确提出金属冶炼行业研发投入强度力争达到1.5%的政策目标。现聚焦以下范围：同时受到国家层面金属冶炼产业发展政策和地方金属冶炼和压延加工业相关政策覆盖的省份，且该省上市企业总数（以营业收入、营业利润和研发投入数据均不为空、营业收入大于零为准）不低于6家，同时全省总营业利润为正值。在此范围内，模拟地方政府出台强制合规要求——凡研发投入强度（研发投入÷营业收入）未达1.5%门槛的企业，须将研发投入强制补足至1.5%，补足部分直接计入成本并从营业利润中扣除。请问：在符合条件的省份中，哪个省份的企业需要补足的研发投入缺口总量最大？执行该合规要求后，该省金属冶炼和压延加工业的总营业利润预计下降多少个百分点？
+
+Output guidelines:
+依次回答省份名称和营业利润下降比例。下降比例以百分数表示，保留2位小数。如["山东省", 3.75]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["江西省", 5.91]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_487_risk_assessment_medium_medium006.md b/tasks/task_487_risk_assessment_medium_medium006.md
new file mode 100644
index 0000000000000000000000000000000000000000..113b1673b7424eee4dec2d8af48bfa298a30af62
--- /dev/null
+++ b/tasks/task_487_risk_assessment_medium_medium006.md
@@ -0,0 +1,119 @@
+---
+id: task_487_risk_assessment_medium_medium006
+name: risk_assessment-medium-medium006
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/medium006.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，国家将橡胶制品业列入新污染物治理试点，同时也要求塑料制品行业加快推进绿色低碳转型。在同时具备国家层面新污染物治理政策约束和地方橡胶和塑料制品政策支持、且省内上市企业总数不低于5家、总营业利润为正的省份中，平均营业利润率（以总营业利润除以总营业收入的比值衡量）最低的是哪个省份？鉴于该省企业本身盈利空间有限，一旦面临新污染物合规落地——以各企业运营成本（=营业收入-营业利润）的3%估算额外环保支出，且这部分费用全额从营业利润中扣减——请问该省橡胶和塑料制品业的总营业利润将因此下降多大比例？
+
+Output guidelines:
+依次回答省份名称和总营业利润下降的百分比。百分比保留2位小数。如["山东省", 56.12]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["上海市", 83.39]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_488_risk_assessment_medium_medium007.md b/tasks/task_488_risk_assessment_medium_medium007.md
new file mode 100644
index 0000000000000000000000000000000000000000..e37f4fbeaf19d268769a19ddc29ed306d7012c0f
--- /dev/null
+++ b/tasks/task_488_risk_assessment_medium_medium007.md
@@ -0,0 +1,119 @@
+---
+id: task_488_risk_assessment_medium_medium007
+name: risk_assessment-medium-medium007
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/medium007.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，若要评估各省半导体行业的市场结构风险，需重点关注市场集中程度——集中度越高，单一龙头企业的经营波动对全省产业的冲击越大。请在满足以下两个条件的省份范围内作答：一是该省已出台半导体行业政策；二是省内半导体业上市企业总数不少于10家（仅统计营业收入不为空且大于零的企业）。基于赫芬达尔-赫希曼指数（HHI=各企业营业收入占省内总营业收入百分比的平方和），哪个省份市场集中度最高？在此基础上，进一步模拟：若该省营业收入体量最大的企业遭遇外部供应链冲击，营业收入收缩30%而省内其余企业保持不变，新的HHI指数会变为多少？
+
+Output guidelines:
+依次回答省份名称、原HHI指数和冲击后HHI指数，HHI指数保留2位小数。如["上海市", 1285.07, 1223.31]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["浙江省", 1596.76, 1302.75]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_489_risk_assessment_medium_medium008.md b/tasks/task_489_risk_assessment_medium_medium008.md
new file mode 100644
index 0000000000000000000000000000000000000000..67d28fd03aa9b313c2bac11553e2ac9515ccd6f2
--- /dev/null
+++ b/tasks/task_489_risk_assessment_medium_medium008.md
@@ -0,0 +1,119 @@
+---
+id: task_489_risk_assessment_medium_medium008
+name: risk_assessment-medium-medium008
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/medium008.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，随着新能源汽车补贴政策进入退坡阶段，各省汽车制造业对政府补贴的依存程度成为衡量产业抗风险能力的关键变量。在已出台地方汽车产业专项扶持政策、且本省汽车制造业上市企业数量不低于10家的省份中（仅纳入政府奖励资金补贴、营业利润、营业收入三项数据均完整的企业，政府补贴依赖度定义为省内补贴总额与营业利润总额之比），哪个省份的补贴依赖度最高？如果对该省所有汽车制造企业同步实施50%补贴削减，且营业利润随之等额下降，则该省汽车制造业的整体营业利润将下降多少？
+
+Output guidelines:
+依次回答省份名称、补贴依赖度和营业利润下降比例。补贴依赖度和下降比例均以百分数表示，保留2位小数。如["广东省", 45.20, 22.60]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["上海市", 66.53, 33.27]`
+
+Scoring rules:
+- The gold answer is a list with N=3 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_2` each as 0 or 1.
+- Return `total = (sum(part_i)) / 3` exactly.
+- If the model output is missing or cannot be parsed into 3 comparable parts, score all parts 0.
+
diff --git a/tasks/task_490_risk_assessment_medium_medium009.md b/tasks/task_490_risk_assessment_medium_medium009.md
new file mode 100644
index 0000000000000000000000000000000000000000..30f6fce8c9945e72bf079c44b01f66651d0090ed
--- /dev/null
+++ b/tasks/task_490_risk_assessment_medium_medium009.md
@@ -0,0 +1,119 @@
+---
+id: task_490_risk_assessment_medium_medium009
+name: risk_assessment-medium-medium009
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/medium009.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，纺织鞋服业受出口需求放缓与国内消费压力的双重影响，结构性风险进一步暴露。在同时受到国家层面纺织产业发展政策和地方纺织行业相关政策双重覆盖、且省内拥有正营业收入的上市企业总数不低于10家的省份范围内，哪个省份的纺织鞋服业市场结构最为集中——即按各企业营业收入占省内总营业收入的百分比计算市场份额后，所有市场份额百分比的平方加总（HHI）最大？进一步地，若该省第一大企业突遭外部需求骤降冲击，营业收入萎缩40%，而省内其余企业营收保持稳定，请重新计算此时的HHI指数。
+
+Output guidelines:
+依次回答省份名称和冲击后的HHI指数。HHI指数保留2位小数。如["广东省", 1538.67]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["福建省", 2069.92]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_491_risk_assessment_medium_medium010.md b/tasks/task_491_risk_assessment_medium_medium010.md
new file mode 100644
index 0000000000000000000000000000000000000000..52c7ba77ac01291e720e6825a92c8eb75d44c049
--- /dev/null
+++ b/tasks/task_491_risk_assessment_medium_medium010.md
@@ -0,0 +1,119 @@
+---
+id: task_491_risk_assessment_medium_medium010
+name: risk_assessment-medium-medium010
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/medium010.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，食品饮料业上市企业在受到国家轻工业高质量发展政策指引的同时，部分省份也出台了针对性的地方食品产业政策。在同时满足两项条件的省份中（条件一：省内有国家轻工业政策与地方食品政策的双重覆盖；条件二：省内有营业收入记录的食品饮料业上市企业不少于6家），哪个省份企业的平均资产负债率水平最高，意味着其对外部融资依赖最深？在此前提下，若货币政策收紧、基准利率上升2个百分点，以各企业总负债额乘以2%估算新增利息负担，所有企业的额外利息成本加总后，这一总额相当于该省食品饮料业总营业利润的多少？
+
+Output guidelines:
+依次回答省份名称和额外利息成本占总营业利润的比例。比例以百分数表示，保留2位小数。如["湖北省", 35.20]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["湖南省", 50.47]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+
diff --git a/tasks/task_492_risk_assessment_medium_medium011.md b/tasks/task_492_risk_assessment_medium_medium011.md
new file mode 100644
index 0000000000000000000000000000000000000000..4940b5e6c72a35766faabcc4fe766a5a50a01348
--- /dev/null
+++ b/tasks/task_492_risk_assessment_medium_medium011.md
@@ -0,0 +1,119 @@
+---
+id: task_492_risk_assessment_medium_medium011
+name: risk_assessment-medium-medium011
+category: risk_assessment
+grading_type: llm_judge
+timeout_seconds: 1200
+gold_file: qa_gold/risk_assessment/medium011.json
+workspace_files: [
+  {
+    "source": "database/bilingual_translation_english_chinese.json",
+    "dest": "database/bilingual_translation_english_chinese.json"
+  },
+  {
+    "source": "database/enterprise/company_core.csv",
+    "dest": "database/enterprise/company_core.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status.csv",
+    "dest": "database/enterprise/company_operation_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_status_detail.csv",
+    "dest": "database/enterprise/company_operation_status_detail.csv"
+  },
+  {
+    "source": "database/enterprise/company_operation_yearly_status.csv",
+    "dest": "database/enterprise/company_operation_yearly_status.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile.csv",
+    "dest": "database/enterprise/company_profile.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_as.csv",
+    "dest": "database/enterprise/company_profile_as.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_eu.csv",
+    "dest": "database/enterprise/company_profile_eu.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_na.csv",
+    "dest": "database/enterprise/company_profile_na.csv"
+  },
+  {
+    "source": "database/enterprise/company_profile_oc.csv",
+    "dest": "database/enterprise/company_profile_oc.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status.csv",
+    "dest": "database/industry/national_industry_status.csv"
+  },
+  {
+    "source": "database/industry/national_industry_status_detail.csv",
+    "dest": "database/industry/national_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/national_industry_yearly_status.csv",
+    "dest": "database/industry/national_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status.csv",
+    "dest": "database/industry/regional_industry_status.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_status_detail.csv",
+    "dest": "database/industry/regional_industry_status_detail.csv"
+  },
+  {
+    "source": "database/industry/regional_industry_yearly_status.csv",
+    "dest": "database/industry/regional_industry_yearly_status.csv"
+  },
+  {
+    "source": "database/internal_metrics.csv",
+    "dest": "database/internal_metrics.csv"
+  },
+  {
+    "source": "database/policy/policy_release_status.csv",
+    "dest": "database/policy/policy_release_status.csv"
+  },
+  {
+    "source": "database/policy/policy_resource.csv",
+    "dest": "database/policy/policy_resource.csv"
+  }
+]
+---
+
+## Prompt
+
+2022年，国家发布一系列采矿业相关产业政策，同时各省也出台了采矿业相关的产业发展政策。在双重政策覆盖之下，且有正营业收入的上市企业总数不低于6家、全省采矿业总营业利润大于零的省份中，哪个省份的采矿业市场份额分布最为集中（市场集中度以营业收入HHI衡量）？此外，鉴于采矿业对国际大宗商品价格高度敏感，若该省营业收入体量最大的采矿企业因国际市场剧烈波动遭遇营收下滑40%（其余企业营收保持不变），重新计算各企业市场份额后，全省采矿业的HHI指数将变动至多少？
+
+Output guidelines:
+依次回答省份名称和冲击后的HHI指数。HHI指数保留2位小数。如["内蒙古自治区", 3200.50]。如果无法找到相关数据，请回答"未查询到相关数据"
+
+Only use files under `./database/`.
+
+## Expected Behavior
+
+Agent should read the provided `database/` files, compute the result, and return the final answer. The final answer must follow the required output format.
+
+## Grading Criteria
+
+- [ ] Final answer semantically matches the gold `answer`.
+- [ ] Output format follows `guidelines`.
+
+## LLM Judge Rubric
+
+### Criterion 1: Multi-answer Correctness (Weight: 100%)
+
+Gold answer JSON:
+`["新疆维吾尔自治区", "4444.61"]`
+
+Scoring rules:
+- The gold answer is a list with N=2 parts.
+- Judge each predicted part against the corresponding gold part by semantic equivalence.
+- Return `scores` with `part_0 ... part_1` each as 0 or 1.
+- Return `total = (sum(part_i)) / 2` exactly.
+- If the model output is missing or cannot be parsed into 2 comparable parts, score all parts 0.
+