Math Reasoning on AIME 2025 (Avg@12)
72.2Average ScorePAINT
Evaluation Results
| Method | Links | |
|---|---|---|
| PAINTBackbone=Qwen3-8B Instruct2026.04 | 72.2 | |
| OPSDBackbone=Qwen3-8B Instruct2026.04 | 70.8 | |
| PAINTBackbone=Qwen3-4B Instruct2026.04 | 69.7 | |
| GRPOBackbone=Qwen3-8B Instruct2026.04 | 68.9 | |
| OPSDBackbone=Qwen3-4B Instruct2026.04 | 68.3 | |
| GRPOBackbone=Qwen3-4B Instruct2026.04 | 68.1 | |
| No updateBackbone=Qwen3-4B Instruct2026.04 | 66.4 | |
| No updateBackbone=Qwen3-8B Instruct2026.04 | 65.6 | |
| SFTBackbone=Qwen3-8B Instruct2026.04 | 64.2 | |
| SFTBackbone=Qwen3-4B Instruct2026.04 | 62.3 | |
| PAINTBackbone=Qwen3-1.7B Instruct2026.04 | 44.2 | |
| OPSDBackbone=Qwen3-1.7B Instruct2026.04 | 43.9 | |
| GRPOBackbone=Qwen3-1.7B Instruct2026.04 | 38.3 | |
| No updateBackbone=Qwen3-1.7B Instruct2026.04 | 36.7 | |
| SFTBackbone=Qwen3-1.7B Instruct2026.04 | 36.3 |