Mathematical Reasoning on GSM8K (Acc, Len, CR)
96.2AccuracyASAG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ASAGBackbone=Qwen3-14B2026.06 | 96.2 | 1,039 | 49.9 | |
| ASAGBackbone=Qwen3-8B2026.06 | 95.9 | 1,063 | 45.5 | |
| DynasorBackbone=Qwen3-14B2026.06 | 95.6 | 1,364 | 65.5 | |
| DEERBackbone=Qwen3-14B2026.06 | 95.3 | 863 | 41.5 | |
| VanillaBackbone=Qwen3-14B2026.06 | 95.1 | 2,082 | 100 | |
| DEERBackbone=Qwen3-8B2026.06 | 95 | 1,096 | 46.9 | |
| TALEBackbone=Qwen3-14B2026.06 | 94.7 | 963 | 46.3 | |
| VanillaBackbone=Qwen3-8B2026.06 | 94.6 | 2,338 | 100 | |
| DEERBackbone=Qwen3-4B2026.06 | 94.2 | 1,286 | 60 | |
| ASAGBackbone=Qwen3-4B2026.06 | 94.2 | 1,139 | 53.2 | |
| DynasorBackbone=Qwen3-8B2026.06 | 94.2 | 1,024 | 43.8 | |
| VanillaBackbone=Qwen3-4B2026.06 | 93.8 | 2,142 | 100 | |
| NoThinkingBackbone=Qwen3-14B2026.06 | 93.5 | 447 | 21.5 | |
| DynasorBackbone=Qwen3-4B2026.06 | 92.9 | 938 | 43.8 | |
| TALEBackbone=Qwen3-8B2026.06 | 92.5 | 1,142 | 48.8 | |
| TALEBackbone=Qwen3-4B2026.06 | 91.3 | 926 | 43.2 | |
| NoThinkingBackbone=Qwen3-8B2026.06 | 91.3 | 685 | 29.3 | |
| ASAGBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 91 | 1,009 | 69.5 | |
| DEERBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 90.3 | 967 | 66.6 | |
| DynasorBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 89.7 | 1,258 | 86.6 | |
| NoThinkingBackbone=Qwen3-4B2026.06 | 89.6 | 624 | 29.1 | |
| VanillaBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 89.5 | 1,452 | 100 | |
| TALEBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 88.2 | 909 | 62.6 | |
| NoThinkingBackbone=DeepSeek-R1-Distill-Qwen-7B2026.06 | 86.3 | 328 | 22.6 |