ResearchBenchmarksMathematical Reasoning on GSM-Symbolic-P1Follow89Pass@1 AccuracyDiRL68.273.67984.4Jun 9, 2026Evaluation ResultsMethodMethodLinksPass@1 AccuracyDiRLModel=Qwen3-4BModel=Qwen3-4B2026.0689EVOL-RLModel=Qwen3-4BModel=Qwen3-4B2026.0681G2RLModel=Qwen3-4BModel=Qwen3-4B2026.0680DiRLModel=Qwen3-1.7BModel=Qwen3-1.7B2026.0671EVOL-RLModel=Qwen3-1.7BModel=Qwen3-1.7B2026.0669G2RLModel=Qwen3-1.7BModel=Qwen3-1.7B2026.0669