Mathematical Reasoning on AIME 25 (Pass@1, #Tok)
80Pass@1 AccuracyCoD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoDBackbone=Qwen3-4B-Thinking-25072026.06 | 80 | 21,246 | |
| BaselineBackbone=Qwen3-4B-Thinking-25072026.06 | 76.7 | 22,708 | |
| DyConBackbone=Qwen3-4B-Thinking-25072026.06 | 76.7 | 21,100 | |
| NothinkingBackbone=Qwen3-4B-Thinking-25072026.06 | 73.3 | 19,177 | |
| DEERBackbone=Qwen3-4B-Thinking-25072026.06 | 70 | 13,342 | |
| NothinkingBackbone=QwQ-32B2026.06 | 70 | 15,312 | |
| BaselineBackbone=Qwen3-14B2026.06 | 70 | 16,613 | |
| DyConBackbone=Qwen3-14B2026.06 | 70 | 14,537 | |
| DyConBackbone=QwQ-32B2026.06 | 66.7 | 13,640 | |
| DEERBackbone=Qwen3-14B2026.06 | 66.7 | 11,135 | |
| BaselineBackbone=QwQ-32B2026.06 | 60 | 16,462 | |
| Dynasor-CoTBackbone=Qwen3-14B2026.06 | 60 | 12,159 | |
| NoWaitBackbone=Qwen3-14B2026.06 | 60 | 12,276 | |
| SEALBackbone=QwQ-32B2026.06 | 56.7 | 12,089 | |
| NoWaitBackbone=QwQ-32B2026.06 | 56.7 | 11,871 | |
| NoWaitBackbone=Qwen3-4B-Thinking-25072026.06 | 53.3 | 13,322 | |
| DEERBackbone=QwQ-32B2026.06 | 50 | 11,598 | |
| CoDBackbone=QwQ-32B2026.06 | 46.7 | 13,289 | |
| CoDBackbone=Qwen3-14B2026.06 | 46.7 | 12,391 | |
| TrimRBackbone=QwQ-32B2026.06 | 43.3 | 8,827 | |
| Controlling Thinking SpeedBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 40 | 10,997 | |
| FlashThinkBackbone=QwQ-32B2026.06 | 40 | 11,861 | |
| BaselineBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 36.7 | 15,245 | |
| DEERBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 36.7 | 7,257 | |
| DyConBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 36.7 | 12,415 | |
| CoDBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 33.3 | 14,333 | |
| ThinkpilotBackbone=Qwen3-4B-Thinking-25072026.06 | 30 | 8,814 | |
| SEALBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 26.7 | 11,092 | |
| NoWaitBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 26.7 | 9,302 | |
| NothinkingBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 23.3 | 4,385 | |
| NothinkingBackbone=Qwen3-14B2026.06 | 23.3 | 5,115 | |
| ThinkpilotBackbone=Qwen3-14B2026.06 | 23.3 | 3,488 | |
| ThinkpilotBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 10 | 1,961 |