ResearchBenchmarksMathematical Reasoning on MATH-500 (Pass@8)Follow82.6Pass@8POPO55.35262.42669.576.574May 7, 2026Evaluation ResultsMethodMethodLinksPass@8POPOModel=DeepSeek-Math-7BModel=DeepSeek-Math-7B2026.0582.6GRPOModel=DeepSeek-Math-7BModel=DeepSeek-Math-7B2026.0580.2POPOModel=Llama-3.1-8BModel=Llama-3.1-8B2026.0559.8GRPOModel=Llama-3.1-8BModel=Llama-3.1-8B2026.0556.4