ResearchBenchmarksMathematical Reasoning on MATH-500 (Final Acc., Correction Uplift)Follow22.2Final AccuracyROSA6.80810.80414.818.796Sep 27, 2025Evaluation ResultsMethodMethodLinksFinal AccuracyCorrection UpliftROSAModel=DeepSeek-R1-Dist...Model=DeepSeek-R1-Distill-Qwen-8B2025.0922.218.62BaselineModel=DeepSeek-R1-Dist...Model=DeepSeek-R1-Distill-Qwen-8B2025.097.46.09