ResearchTasksReward rollout alignmentReward rollout alignmentFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updated10-task benchmark T1: Folding ShortsSARM20.889Rollout ρ8Jun 11, 202610-task benchmark T2: Cleaning WhiteboardSARM20.667Rollout ρ7Jun 11, 2026