ResearchBenchmarksMathematical Reasoning on AIME OpenR1-Math Harder 2024Follow72.7AccuracyQwen-4B61.05264.07667.170.124Feb 11, 2026Evaluation ResultsMethodMethodLinksAccuracyQwen-4BBackbone=Qwen, Paramet...Backbone=Qwen, Parameters=4B2026.0272.7RePOBackbone=Qwen, Paramet...Backbone=Qwen, Parameters=4B2026.0272.1LUFFYBackbone=Qwen, Paramet...Backbone=Qwen, Parameters=4B2026.0261.5