ResearchTasksMulti-Task ReasoningMulti-Task ReasoningFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedAverage (2WikiMultiHop, MMLU, GSM8k) (in-distribution)CONCUR75.2Accuracy29Apr 21, 2026Average Out-of-domainDFT49.57Accuracy (OOD)24Apr 21, 2026MMLU-ProComposition-RL69.3Pass@118May 12, 2026GPQA DiamondStandard RLVR55Pass@115Mar 18, 2026BigBench HardGPT-4o31.1Score5Feb 26, 2026MMLUDCRL56.7Pass@13Mar 18, 2026