data_source stringclasses 4
values | prompt listlengths 2 2 | ability stringclasses 2
values | reward_model dict | extra_info dict | sub_source stringclasses 21
values |
|---|---|---|---|---|---|
taco | [
{
"content": "Please reason step by step, then provide your complete solution as a single Python code block (```python ... ```). The program must read from standard input and write the answer to standard output.",
"role": "system"
},
{
"content": "You are given a permutation p of n elements. A permu... | code | {
"ground_truth": "{\"inputs\": [\"1 1\\n1\\n2 1 1\\n\", \"100 10\\n71 72 20 93 38 81 19 44 88 57 3 37 86 64 63 97 15 85 53 99 61 66 75 96 48 67 6 31 92 70 27 54 74 43 7 5 84 33 60 65 95 76 11 14 49 24 77 21 50 98 46 25 83 45 94 8 68 91 28 4 80 18 17 1 47 34 40 26 69 35 58 9 36 32 22 16 10 79 13 90 78 29 23 41 100 51... | {
"id": "taco-test-140",
"rating": null,
"sub_source": "VERY_HARD"
} | VERY_HARD |
taco | [{"content":"Please reason step by step, then provide your complete solution as a single Python code(...TRUNCATED) | code | {"ground_truth":"{\"inputs\": [[\"3\", \"4 2\", \"abcb\", \"7 1\", \"aayzbaa\", \"12 80123123\", \"a(...TRUNCATED) | {
"id": "taco-test-158",
"rating": null,
"sub_source": "MEDIUM_HARD"
} | MEDIUM_HARD |
codeforces | [{"content":"Please reason step by step, then provide your complete solution as a single Python code(...TRUNCATED) | code | {"ground_truth":"{\"inputs\": [\"++-+-\\n+-+-+\", \"+-+-\\n+-??\", \"+++\\n??-\", \"++++++++++\\n+++(...TRUNCATED) | {
"id": "476B",
"rating": 1300,
"sub_source": "rating_1300"
} | rating_1300 |
taco | [{"content":"Please reason step by step, then provide your complete solution as a single Python code(...TRUNCATED) | code | {"ground_truth":"{\"inputs\": [\"4\\n22 73 9\\n45 64 6\\n47 55 7\\n2 62 4\\n\", \"4\\n82 94 6\\n56 6(...TRUNCATED) | {
"id": "taco-test-66",
"rating": null,
"sub_source": "EASY"
} | EASY |
taco | [{"content":"Please reason step by step, then provide your complete solution as a single Python code(...TRUNCATED) | code | {"ground_truth":"{\"inputs\": [[\"2\", \"3\", \"0 1 1\", \"1 1 0\", \"3\", \"0 1 5\", \"5 1 0\"], [\(...TRUNCATED) | {
"id": "taco-test-26",
"rating": null,
"sub_source": "HARD"
} | HARD |
codeforces | [{"content":"Please reason step by step, then provide your complete solution as a single Python code(...TRUNCATED) | code | {"ground_truth":"{\"inputs\": [\"3\\n5 10 1\", \"4\\n1 1 10 11\", \"5\\n0 3 1 100 1\", \"40\\n10 100(...TRUNCATED) | {
"id": "614B",
"rating": 1400,
"sub_source": "rating_1400"
} | rating_1400 |
taco | [{"content":"Please reason step by step, then provide your complete solution as a single Python code(...TRUNCATED) | code | {"ground_truth":"{\"inputs\": [\"2\\n2\\n2 2\\n1 1\\n2\\n2 3\\n1 2\", \"1000\\n2\\n335 501 \\n75 194(...TRUNCATED) | {
"id": "taco-test-112",
"rating": null,
"sub_source": "MEDIUM"
} | MEDIUM |
math_box | [{"content":"Please reason step by step, and put your final answer within \\boxed{}.","role":"system(...TRUNCATED) | math | {
"ground_truth": "0",
"style": "rule"
} | {
"id": "numina-test-65",
"rating": null,
"sub_source": "synthetic_math"
} | synthetic_math |
codeforces | [{"content":"Please reason step by step, then provide your complete solution as a single Python code(...TRUNCATED) | code | {"ground_truth":"{\"inputs\": [\"5\\n270\\n250\\n250\\n230\\n250\", \"5\\n250\\n250\\n250\\n250\\n25(...TRUNCATED) | {
"id": "99B",
"rating": 1300,
"sub_source": "rating_1300"
} | rating_1300 |
codeforces | [{"content":"Please reason step by step, then provide your complete solution as a single Python code(...TRUNCATED) | code | {"ground_truth":"{\"inputs\": [\"3 2\\nab\\nac\\ncd\\n1 2\", \"5 3\\ntest\\ntezt\\ntest.\\n.est\\nte(...TRUNCATED) | {
"id": "730H",
"rating": 1300,
"sub_source": "rating_1300"
} | rating_1300 |
End of preview. Expand in Data Studio
rl-value-eval
Held-out evaluation set (774 prompts) for the Block-3 confidence / correctness-estimation
comparison. Pairs with YangyiYY/rl-value-confidence-train
(training) and the nct-ppo actor/critic models. Each row is in the verl RL schema so the shared
grader scores it directly.
Single test split (the former validation+test splits are merged into one).
Composition & subset annotation
Every row carries a fine-grained subset label in both extra_info.sub_source and a top-level
sub_source column:
| data_source | n | sub_source = |
|---|---|---|
taco |
300 | TACO difficulty — EASY / MEDIUM / MEDIUM_HARD / HARD / VERY_HARD |
codeforces |
300 | rating_1300, rating_1400 |
math_box (numina) |
98 | NuminaMath origin — cn_k12 / synthetic_math / orca_math / olympiads / gsm8k / synthetic_amc / aops_forum / amc_aime / math |
codecontests |
76 | rating_800 … rating_1200 |
Grader / ground_truth
reward_model.ground_truth is built for the shared reference grader (reference/ in the
project repo): math → {"answer": …}; code → input_output JSON with the full hidden test
suites (CodeContests public+private+generated, Codeforces full suite, TACO raw
input_output incl. call-based fn_name), no per-problem test-case cap. Correctness =
last \boxed{} match (math) or all stdio/functional cases pass (code, 6s/case · 60s/problem).
Schema
data_source · prompt (chat messages) · ability · reward_model.ground_truth · sub_source ·
extra_info (id, rating, sub_source).
- Downloads last month
- -