ResearchBenchmarksCommonsense Reasoning on GPQA (pass@1)Follow24.2pass@1GTPO21.622.27522.9523.625Nov 18, 2025Evaluation ResultsMethodMethodLinkspass@1GTPOBase Model=Qwen2.5-7BBase Model=Qwen2.5-7B2025.1124.2GRPOBase Model=Qwen2.5-7BBase Model=Qwen2.5-7B2025.1121.7