Datasets:
component stringclasses 3
values | split stringclasses 13
values | item_id stringlengths 2 32 | input stringlengths 0 361k | output stringlengths 9 1.81M | metadata_json stringlengths 66 1.81M | record_hash stringlengths 64 64 |
|---|---|---|---|---|---|---|
planconditions | goodplan | HB_A_HELLO_000228 | {
"goal": {
"id": "g0",
"task_type": "constrained_long_form_generation",
"deliverable": "screenplay",
"audience": "",
"primary_objective": "Write a well-structured, engaging, and dynamic screenplay based on the prompt of a woman being attacked and overpowered in a pool by a man in a wet suit.",
... | {"condition": "GoodPlan", "condition_id": "good_plan_R0_json", "corruption_type": null, "lambda": null, "plan_ir": {"constraints": [{"binds_to": ["whole_output"], "constraint_type": "structure", "id": "c1", "severity": "hard", "text": "Start with a slugline (e.g., INT. HOUSE - NIGHT) that sets the scene."}, {"binds_to"... | {"condition": "GoodPlan", "condition_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000228/condition_family/good_plan_R0_json/condition.json", "condition_id": "good_plan_R0_json", "condition_plan_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000228/condition_family/good_plan_R0_json/condition... | d44ef58ffb0024b0089f5badef6c3de1910aeb0dc1e6bccd730500f63789989b |
planconditions | goodplan | HB_A_HELLO_000228 | {
"goal": {
"id": "g0",
"task_type": "constrained_long_form_generation",
"deliverable": "screenplay",
"audience": "",
"primary_objective": "Write a well-structured, engaging, and dynamic screenplay based on the prompt of a woman being attacked and overpowered in a pool by a man in a wet suit.",
... | {"condition": "GoodPlan", "condition_id": "good_plan_self_refine", "corruption_type": null, "lambda": null, "plan_ir": {"constraints": [{"binds_to": ["whole_output"], "constraint_type": "structure", "id": "c1", "severity": "hard", "text": "Start with a slugline (e.g., INT. HOUSE - NIGHT) that sets the scene."}, {"binds... | {"condition": "GoodPlan", "condition_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000228/condition_family/good_plan_self_refine/condition.json", "condition_id": "good_plan_self_refine", "condition_plan_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000228/condition_family/good_plan_self_refi... | a84c5708351c7eaff006dd1689bc6710a37b81a211067d7e6f81d6650928d623 |
planconditions | goodplan | HB_A_HELLO_000251 | {
"goal": {
"id": "g0",
"task_type": "constrained_long_form_generation",
"deliverable": "engaging story",
"audience": "",
"primary_objective": "Write an engaging, creative story based on the prompt that humans are not Space Orcs but Space Dwarves.",
"success_criteria": [
"Story is clearl... | {"condition": "GoodPlan", "condition_id": "good_plan_R0_json", "corruption_type": null, "lambda": null, "plan_ir": {"constraints": [{"binds_to": ["whole_output"], "constraint_type": "structure", "id": "c1", "severity": "soft", "text": "Use creativity to expand on the prompt and create an interesting and captivating nar... | {"condition": "GoodPlan", "condition_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000251/condition_family/good_plan_R0_json/condition.json", "condition_id": "good_plan_R0_json", "condition_plan_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000251/condition_family/good_plan_R0_json/condition... | f300f8fc4262c03e494600a3c17a88e60287987aee3aa0dbfc649bf4da2885a0 |
planconditions | goodplan | HB_A_HELLO_000251 | {
"goal": {
"id": "g0",
"task_type": "constrained_long_form_generation",
"deliverable": "engaging story",
"audience": "",
"primary_objective": "Write an engaging, creative story based on the prompt that humans are not Space Orcs but Space Dwarves.",
"success_criteria": [
"Story is clearl... | {"condition": "GoodPlan", "condition_id": "good_plan_self_refine", "corruption_type": null, "lambda": null, "plan_ir": {"constraints": [{"binds_to": ["whole_output"], "constraint_type": "structure", "id": "c1", "severity": "soft", "text": "Use creativity to expand on the prompt and create an interesting and captivating... | {"condition": "GoodPlan", "condition_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000251/condition_family/good_plan_self_refine/condition.json", "condition_id": "good_plan_self_refine", "condition_plan_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000251/condition_family/good_plan_self_refi... | c112dcffe65e1fc0227777fb12ed1585c5773419a5b72268ba310c778665ce68 |
planconditions | goodplan | HB_A_HELLO_000332 | {
"goal": {
"id": "g0",
"task_type": "constrained_long_form_generation",
"deliverable": "detailed written response to a question on food",
"audience": "",
"primary_objective": "Answer whether gold is actually edible and safe to eat or mostly used as decor for extravagance in certain foods, providi... | {"condition": "GoodPlan", "condition_id": "good_plan_R0_json", "corruption_type": null, "lambda": null, "plan_ir": {"constraints": [{"binds_to": ["whole_output"], "constraint_type": "length", "id": "c1", "severity": "hard", "text": "The answer should be long enough to provide a comprehensive response."}], "derivation_m... | {"condition": "GoodPlan", "condition_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000332/condition_family/good_plan_R0_json/condition.json", "condition_id": "good_plan_R0_json", "condition_plan_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000332/condition_family/good_plan_R0_json/condition... | 2533b9ee0107f2f8e34f73354f55f7135e37519218ebc98675fc9cde4be4eabd |
planconditions | goodplan | HB_A_HELLO_000332 | {
"goal": {
"id": "g0",
"task_type": "constrained_long_form_generation",
"deliverable": "detailed written response to a question on food",
"audience": "",
"primary_objective": "Answer whether gold is actually edible and safe to eat or mostly used as decor for extravagance in certain foods, providi... | {"condition": "GoodPlan", "condition_id": "good_plan_self_refine", "corruption_type": null, "lambda": null, "plan_ir": {"constraints": [{"binds_to": ["whole_output"], "constraint_type": "length", "id": "c1", "severity": "hard", "text": "The answer should be long enough to provide a comprehensive response."}], "derivati... | {"condition": "GoodPlan", "condition_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000332/condition_family/good_plan_self_refine/condition.json", "condition_id": "good_plan_self_refine", "condition_plan_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000332/condition_family/good_plan_self_refi... | 40341585cca475397c78f817d8e7e97b3e2df73925ea0781d0de46327a923614 |
planconditions | goodplan | HB_A_HELLO_000368 | {
"goal": {
"id": "g0",
"task_type": "constrained_long_form_generation",
"deliverable": "detailed written response to a question about movie and literature classification",
"audience": "",
"primary_objective": "Answer whether there is a classification for fiction movies and literature that disting... | {"condition": "GoodPlan", "condition_id": "good_plan_R0_json", "corruption_type": null, "lambda": null, "plan_ir": {"constraints": [{"binds_to": ["whole_output"], "constraint_type": "length", "id": "c1", "severity": "hard", "text": "The answer should be long enough to provide a comprehensive response."}], "derivation_m... | {"condition": "GoodPlan", "condition_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000368/condition_family/good_plan_R0_json/condition.json", "condition_id": "good_plan_R0_json", "condition_plan_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000368/condition_family/good_plan_R0_json/condition... | 354e17d7aca26be3ddd7b92d5544021c1aeaab3a18b17830b4c6f320a371249d |
planconditions | goodplan | HB_A_HELLO_000368 | {
"goal": {
"id": "g0",
"task_type": "constrained_long_form_generation",
"deliverable": "detailed written response to a question about movie and literature classification",
"audience": "",
"primary_objective": "Answer whether there is a classification for fiction movies and literature that disting... | {"condition": "GoodPlan", "condition_id": "good_plan_self_refine", "corruption_type": null, "lambda": null, "plan_ir": {"constraints": [{"binds_to": ["whole_output"], "constraint_type": "length", "id": "c1", "severity": "hard", "text": "The answer should be long enough to provide a comprehensive response."}], "derivati... | {"condition": "GoodPlan", "condition_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000368/condition_family/good_plan_self_refine/condition.json", "condition_id": "good_plan_self_refine", "condition_plan_file": "datasets/harmonicbench_main/domain1/runs/HB_A_HELLO_000368/condition_family/good_plan_self_refi... | e4a2f07c652abd958ec1e766f5a3dfa65870479f87f671116c68d6b017a55d8d |
planconditions | goodplan | HB_A_HELLO_000546 | "{\n \"goal\": {\n \"id\": \"g0\",\n \"task_type\": \"constrained_long_form_generation\",\n (...TRUNCATED) | "{\"condition\": \"GoodPlan\", \"condition_id\": \"good_plan_R0_json\", \"corruption_type\": null, \(...TRUNCATED) | "{\"condition\": \"GoodPlan\", \"condition_file\": \"datasets/harmonicbench_main/domain1/runs/HB_A_H(...TRUNCATED) | e67cfe95787cffd9979479a3c2331896d0e995b327221fdc5db1bad864626863 |
planconditions | goodplan | HB_A_HELLO_000546 | "{\n \"goal\": {\n \"id\": \"g0\",\n \"task_type\": \"constrained_long_form_generation\",\n (...TRUNCATED) | "{\"condition\": \"GoodPlan\", \"condition_id\": \"good_plan_self_refine\", \"corruption_type\": nul(...TRUNCATED) | "{\"condition\": \"GoodPlan\", \"condition_file\": \"datasets/harmonicbench_main/domain1/runs/HB_A_H(...TRUNCATED) | a41cca21e3dca55040f03a502ae661cc4c863277b118fd7008c914d4eb3ec435 |
End of preview. Expand in Data Studio
HARMONICBench Croissant Release
This Hugging Face Dataset repository hosts the Croissant-generation release for HARMONICBench v1.0.0. The files follow the Hugging Face Dataset Viewer best-practice layout:
train.parquet
validation.parquet
test.parquet
README.md
LICENSE
metadata.json
Each row uses a stable schema:
{
"component": "core | planconditions | scoring | run_index | source_pools | release_metadata",
"split": "component-specific split name",
"item_id": "stable row identifier",
"input": "prompt, plan text, source text, or serialized table row",
"output": "reference/evidence/condition metadata or serialized table row",
"metadata_json": "JSON string with provenance, hashes, source paths, and release metadata",
"record_hash": "sha256 of the canonical row"
}
Paper-facing artifact scope
Load-bearing evidence in the submitted v1 artifact consists of fixed-target measurements, benchmark-native score agreement, numeric scorer-route ablation, released scorer/configuration artifacts, and a reviewer smoke path. Powered human preference outcomes and completed open-weight aggregate results are not load-bearing in this release.
Dataset files
| File | Records | SHA256 |
|---|---|---|
train.parquet |
22029 | ddbad4d9a2cd71277ee538eefae6c38a5203c69d4a65a27facf569e35cce93c4 |
validation.parquet |
78 | d399f5e118670321e66e70acab0038dd6fcc00557dcd929c07dd1b828bf45174 |
test.parquet |
811 | 12270c2c35a9fd91a1df25ba37635677d98145fddc39eccf9815a511066fa8d3 |
GitHub release: https://github.com/AlexJJJChen/HARMONIC-Bench
- Downloads last month
- 7