遇见数据集

ProofBench: official evaluation runs

收藏
Zenodo2026-11-03 更新2026-11-05 收录
官方服务:

资源简介:

Archive of the official ProofBench evaluation runs, the data behind the accompanying paper. Each version adds one run (or one part of a run executed in parts): the zipped run folder (manifest, Inspect AI logs, results) and its SHA-256 checksums. All runs use the final harness (commit 737d089 or later of github.com/proof-bench/QEDRank). Files are restricted until the accompanying paper is published. Pilot runs: 10.5281/zenodo.23606289; a superseded first part of the Fable run with web search: 10.5281/zenodo.23837497.

提供机构:
Zenodo
创建时间:
2026-11-03
二维码
社区交流群
二维码
科研交流群
商业服务