ResearchTasksSkill-Augmented Agent Performance EvaluationFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedSkillsBench 86 runnable tasks (frozen-policy transfer)Policy-selected79Valid Task Count15Jun 9, 2026SkillsBench 20-task balanced panel (held-out evaluation)Rule/formula anchored19Valid Task Count5Jun 9, 2026
SkillsBench 20-task balanced panel (held-out evaluation)Rule/formula anchored19Valid Task Count5Jun 9, 2026