Mathematical Reasoning on OlympiadBench (Acc, Len, CR)
64.6AccuracyASAG
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ASAGBackbone=Qwen3-4B2026.06 | 64.6 | 7,536 | 84.5 | |
| ASAGBackbone=Qwen3-14B2026.06 | 64.6 | 6,819 | 80 | |
| ASAGBackbone=Qwen3-8B2026.06 | 63.9 | 7,296 | 78.7 | |
| DynasorBackbone=Qwen3-4B2026.06 | 63.6 | 6,863 | 77 | |
| DynasorBackbone=Qwen3-14B2026.06 | 63 | 7,336 | 86.1 | |
| DEERBackbone=Qwen3-4B2026.06 | 62.9 | 7,344 | 82.4 | |
| DEERBackbone=Qwen3-14B2026.06 | 62.7 | 7,020 | 82.4 | |
| DynasorBackbone=Qwen3-8B2026.06 | 62.4 | 7,640 | 82.4 | |
| VanillaBackbone=Qwen3-14B2026.06 | 61.9 | 8,519 | 100 | |
| DEERBackbone=Qwen3-8B2026.06 | 61.7 | 7,367 | 79.5 | |
| VanillaBackbone=Qwen3-8B2026.06 | 59.9 | 9,268 | 100 | |
| VanillaBackbone=Qwen3-4B2026.06 | 59 | 8,915 | 100 | |
| TALEBackbone=Qwen3-14B2026.06 | 58.4 | 7,250 | 85.1 | |
| TALEBackbone=Qwen3-8B2026.06 | 56.1 | 7,306 | 78.8 | |
| TALEBackbone=Qwen3-4B2026.06 | 54.7 | 6,068 | 68.1 | |
| ASAGBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 52.3 | 5,325 | 63.8 | |
| DEERBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 51.1 | 5,420 | 64.9 | |
| NoThinkingBackbone=Qwen3-14B2026.06 | 50.4 | 5,462 | 64.1 | |
| DynasorBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 50 | 5,973 | 71.6 | |
| TALEBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 48.9 | 5,711 | 68.4 | |
| NoThinkingBackbone=Qwen3-8B2026.06 | 48.7 | 5,219 | 56.3 | |
| VanillaBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 48.6 | 8,347 | 100 | |
| NoThinkingBackbone=Qwen3-4B2026.06 | 40.6 | 4,358 | 48.9 | |
| NoThinkingBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 37.3 | 3,140 | 37.6 |