Mathematical Reasoning on AIME 2025 (Acc, Len, CR)
66.7AccuracyVanilla
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| VanillaModel=Qwen3-32B2026.06 | 66.7 | 12,398 | 100 | |
| ASAGModel=Qwen3-32B2026.06 | 66.7 | 9,211 | 74.3 | |
| ASAGModel=Qwen3-14B2026.06 | 63.3 | 11,290 | 90.7 | |
| ASAGModel=Qwen3-8B2026.06 | 60 | 11,826 | 90.7 | |
| VanillaModel=Qwen3-14B2026.06 | 60 | 12,451 | 100 | |
| ASAGModel=Qwen3-4B2026.06 | 53.3 | 11,490 | 89.9 | |
| VanillaModel=Qwen3-8B2026.06 | 53.3 | 13,041 | 100 | |
| VanillaModel=Qwen3-4B2026.06 | 46.7 | 12,774 | 100 | |
| ASAGModel=DeepSeek-R1-Distill-Qwen-7B2026.06 | 36.7 | 7,547 | 63.6 | |
| ASAGModel=DeepSeek-R1-Distill-Llama-8B2026.06 | 33.3 | 7,721 | 63.6 | |
| VanillaModel=DeepSeek-R1-Distill-Qwen-7B2026.06 | 26.7 | 11,861 | 100 | |
| VanillaModel=DeepSeek-R1-Distill-Llama-8B2026.06 | 26.7 | 12,142 | 100 |