ResearchTasksOut-of-Distribution ReasoningFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedOOD Reasoning Datasets (MMLU-S, GPQA, ARC, BBH) (test)DaGRPO (w/o Off-policy)45.9GPQA20Feb 26, 2026OOD Reasoning Suite GPQA-D, Connect4 (test)RFT49.1GPQA-D Score10Mar 4, 2026ARC-c, MMLU-Pro, GPQA-diamond OOD Evaluation SuiteLPO38.4ARC-c Accuracy (Avg@32)6May 8, 2026