ResearchBenchmarksMathematical Reasoning on MGSM (English)Follow0.968Pass@1DyCon0.955520.958760.9620.96524Jun 5, 2026Evaluation ResultsMethodMethodLinksPass@1Avg Output TokensDyConModel=Qwen3-4B-Thinkin...Model=Qwen3-4B-Thinking-25072026.060.9681,116BaselineModel=Qwen3-4B-Thinkin...Model=Qwen3-4B-Thinking-25072026.060.9561,483