ResearchBenchmarksMathematical Reasoning on AIME 2024 (Avg@30, Token Stats)Follow85.3Avg@30DyCon51.81260.50669.277.894Jun 5, 2026Evaluation ResultsMethodMethodLinksAvg@30Std Dev (Avg@30)Std Dev (#Tok)Token CountDyConBackbone=Qwen3-4B-Thin...Backbone=Qwen3-4B-Thinking-25072026.0685.30.037264118,632BaselineBackbone=Qwen3-4B-Thin...Backbone=Qwen3-4B-Thinking-25072026.0683.10.038465721,279DyConBackbone=DeepSeek-R1-D...Backbone=DeepSeek-R1-Distill-Qwen-7B2026.0654.70.046198110,183BaselineBackbone=DeepSeek-R1-D...Backbone=DeepSeek-R1-Distill-Qwen-7B2026.0653.10.06431,20513,358