Mathematical Reasoning on AIME 2025 (Mean@8, Pass@8, Length)
76.25Mean@8Art
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ArtBackbone=Qwen3-4B-Thinking-2507, Training Step=step 2002026.02 | 76.25 | 86.67 | 16 | |
| VanillaBackbone=Qwen3-4B-Thinking-25072026.02 | 75.83 | 90 | 20.9 | |
| ArtBackbone=Qwen3-8B, Training Step=step 1002026.02 | 67.08 | 83.33 | 12.8 | |
| VanillaBackbone=Qwen3-8B2026.02 | 65.83 | 86.67 | 17.9 | |
| VanillaBackbone=Qwen3-30B-A3B-Instruct-25072026.02 | 60.83 | 83.33 | 6.9 | |
| ArtBackbone=Qwen3-30B-A3B-Instruct-2507, Training Step=step 6002026.02 | 60.83 | 76.67 | 5.1 | |
| ArtBackbone=Qwen3-4B-Instruct-2507, Training Step=step 14402026.02 | 46.67 | 70 | 4.8 | |
| VanillaBackbone=Qwen3-4B-Instruct-25072026.02 | 45.42 | 66.67 | 9.1 | |
| ArtBackbone=Qwen3-1.7B, Training Step=step 5602026.02 | 38.75 | 60 | 11.2 | |
| VanillaBackbone=Qwen3-1.7B2026.02 | 35 | 60 | 17.7 | |
| ArtBackbone=Qwen3-0.6B, Training Step=step 6402026.02 | 24.58 | 36.67 | 8.9 | |
| VanillaBackbone=Qwen3-0.6B2026.02 | 13.33 | 26.67 | 14.9 |