ResearchTasksHeld-out Skill EvaluationHeld-out Skill EvaluationFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedMulti-family n=225 (holdout)GRPO from SFT init43.556Solve Rate3May 29, 2026