ResearchTasksReasoning and KnowledgeFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedReasoning and Knowledge Suite (MMLU, ARC-C, ARC-E, BoolQ, CSQA, HSwag, PIQA, SocIQ, Wino) (various)Qwen3-4B75.78MMLU14Feb 26, 2026Out-of-Distribution Benchmarks SummaryICPO†75.3Average Score12Apr 16, 2026GSM8K, Math, AIME, HumanEval, LiveCodeBench, ARC-C, ARC-E, MMLU, GPQAReasoning95.41GSM8K9Feb 27, 2026
Reasoning and Knowledge Suite (MMLU, ARC-C, ARC-E, BoolQ, CSQA, HSwag, PIQA, SocIQ, Wino) (various)Qwen3-4B75.78MMLU14Feb 26, 2026
GSM8K, Math, AIME, HumanEval, LiveCodeBench, ARC-C, ARC-E, MMLU, GPQAReasoning95.41GSM8K9Feb 27, 2026