ResearchBenchmarksMathematical Reasoning on MATH (Final Accuracy and Correction Uplift)Follow9.8Final AccuracyROSA7.5128.1068.79.294Sep 27, 2025Evaluation ResultsMethodMethodLinksFinal AccuracyCorrection UpliftROSAModel=DeepSeek-R1-Dist...Model=DeepSeek-R1-Distill-Qwen-8B2025.099.85.65BaselineModel=DeepSeek-R1-Dist...Model=DeepSeek-R1-Distill-Qwen-8B2025.097.63.14