PartHackBench: Certified Equal-Progress Stress Tests for Partial-Credit Tool-Agent Evaluation

T1Covered by 1 source · first reported by arXiv at 25 Sept, 04:00 UTC

arXiv:2609.29578v1 Announce Type: new Abstract: Long-horizon tool agents often make useful progress without reaching terminal success, motivating partial-credit evaluation. Yet evaluators may reward milestones that were temporary, later reversed, or not attributable to the evaluated agent…

science

Sources

• T1PartHackBench: Certified Equal-Progress Stress Tests for Partial-Credit Tool-Agent EvaluationarXiv25 Sept, 04:00