ResearchBenchmarksMath Reasoning on AIME 2025 (pass@1 and pass@4)Follow0.63Pass@1XRPO0.51560.54530.5750.6047Oct 8, 2025Evaluation ResultsMethodMethodLinksPass@1Pass@4XRPOModel=Llama-3.2-3BModel=Llama-3.2-3B2025.100.632.38GRPOModel=Llama-3.2-3BModel=Llama-3.2-3B2025.100.521.97