Chain-of-Thought Reasoning on AIME 25
81.04AccuracyVanilla
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| VanillaModel=Qwen3-4B-Thinking-25072026.06 | 81.04 | 22,613 | — | |
| CUSUMModel=Qwen3-4B-Thinking-25072026.06 | 78.13 | 20,663 | 8.6 | |
| DEERModel=Qwen3-4B-Thinking-25072026.06 | 76.2 | 21,056 | 6.9 | |
| DynasorModel=Qwen3-4B-Thinking-25072026.06 | 74 | 20,702 | 8.4 | |
| VanillaModel=Qwen3-14B2026.06 | 71.67 | 16,727 | — | |
| CUSUMModel=Qwen3-14B2026.06 | 68.96 | 14,653 | 12.4 | |
| DEERModel=Qwen3-14B2026.06 | 68.89 | 15,438 | 7.7 | |
| DynasorModel=Qwen3-14B2026.06 | 68.27 | 14,305 | 14.5 | |
| VanillaModel=DeepSeek-R1-Distill-Qwen-7B2026.06 | 41.04 | 14,556 | — | |
| CUSUMModel=DeepSeek-R1-Distill-Qwen-7B2026.06 | 39.17 | 11,740 | 19.3 | |
| DEERModel=DeepSeek-R1-Distill-Qwen-7B2026.06 | 37.5 | 12,148 | 16.5 | |
| DynasorModel=DeepSeek-R1-Distill-Qwen-7B2026.06 | 37.11 | 12,822 | 11.9 |