RewardBench
Attribution: Nathan Lambert et al.; Allen Institute for AI; LLMBar, HumanEvalPack/HumanEval, PRM800K and MATH creators.
Original source: https://huggingface.co/datasets/allenai/reward-bench
Source revision: 168d848cdbbea9764fae4a544dc9ca1e6cca4931.
Declared terms for selected material: ODC-By-1.0 AND MIT (selected components).
Sharing decision: include. All 12 selected entries are mapped to LLMBar, HumanEvalPack, or PRM800K. The original RewardBench component table and publisher notices identify MIT terms for these sources. Retain collection ODC-By attribution and all applicable component notices, including HumanEval and MATH.
Changes: selected excerpts and choice-format normalization; capability labels are ours. Released records disclose any adapted choice representations. No endorsement by source authors or publishers is implied.
Bundled original notices:
Per-question attribution details identify the cleared selections. Checked evidence URLs and hashes are recorded in sources.json.