ResearchBenchmarksMath Reasoning on BRUMO 2025 (pass@1, pass@4)Follow4.79Pass@1GRPO3.493.82754.1654.5025Oct 8, 2025Evaluation ResultsMethodMethodLinksPass@1Pass@4GRPOModel=Llama-3.2-3BModel=Llama-3.2-3B2025.104.797.82XRPOModel=Llama-3.2-3BModel=Llama-3.2-3B2025.103.549.48