Mathematical Reasoning on AIME 2024 (Accuracy, Length, CR)
73.3AccuracyDEER
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DEERBackbone=Qwen3-14B2026.06 | 73.3 | 8,402 | 79.7 | |
| ASAGBackbone=Qwen3-14B2026.06 | 73.3 | 7,602 | 72.1 | |
| ASAGBackbone=Qwen3-4B2026.06 | 70 | 8,768 | 73.6 | |
| VanillaBackbone=Qwen3-14B2026.06 | 70 | 10,537 | 100 | |
| TALEBackbone=Qwen3-14B2026.06 | 70 | 10,358 | 98.3 | |
| DynasorBackbone=Qwen3-14B2026.06 | 70 | 11,156 | 105.9 | |
| ASAGBackbone=Qwen3-8B2026.06 | 66.7 | 8,683 | 71.7 | |
| VanillaBackbone=Qwen3-4B2026.06 | 63.3 | 11,916 | 100 | |
| DynasorBackbone=Qwen3-4B2026.06 | 63.3 | 11,387 | 95.6 | |
| VanillaBackbone=Qwen3-8B2026.06 | 63.3 | 12,101 | 100 | |
| DEERBackbone=Qwen3-8B2026.06 | 63.3 | 8,930 | 73.8 | |
| TALEBackbone=Qwen3-4B2026.06 | 60 | 11,469 | 96.2 | |
| DEERBackbone=Qwen3-4B2026.06 | 60 | 8,745 | 73.4 | |
| TALEBackbone=Qwen3-8B2026.06 | 60 | 11,847 | 97.9 | |
| DynasorBackbone=Qwen3-8B2026.06 | 60 | 11,365 | 93.9 | |
| ASAGBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 46.7 | 9,154 | 66.7 | |
| DEERBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 43.3 | 8,669 | 63.2 | |
| VanillaBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 40 | 13,722 | 100 | |
| DynasorBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 40 | 11,462 | 83.5 | |
| TALEBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 33.3 | 11,627 | 84.7 | |
| NoThinkingBackbone=Qwen3-14B2026.06 | 30 | 7,726 | 73.3 | |
| NoThinkingBackbone=Qwen3-4B2026.06 | 23.3 | 6,458 | 54.2 | |
| NoThinkingBackbone=Qwen3-8B2026.06 | 23.3 | 7,121 | 58.8 | |
| NoThinkingBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 16.7 | 7,365 | 53.7 |