Math Reasoning on AIME 2024 (Avg@12)
79.2Avg@12 ScorePAINT
Evaluation Results
| Method | Links | |
|---|---|---|
| PAINTBackbone=Qwen3-8B Instruct2026.04 | 79.2 | |
| OPSDBackbone=Qwen3-8B Instruct2026.04 | 77.8 | |
| PAINTBackbone=Qwen3-4B Instruct2026.04 | 76.9 | |
| GRPOBackbone=Qwen3-8B Instruct2026.04 | 76.4 | |
| OPSDBackbone=Qwen3-4B Instruct2026.04 | 76.4 | |
| No updateBackbone=Qwen3-8B Instruct2026.04 | 75.8 | |
| GRPOBackbone=Qwen3-4B Instruct2026.04 | 75.6 | |
| No updateBackbone=Qwen3-4B Instruct2026.04 | 74.9 | |
| SFTBackbone=Qwen3-8B Instruct2026.04 | 72.3 | |
| SFTBackbone=Qwen3-4B Instruct2026.04 | 70.2 | |
| PAINTBackbone=Qwen3-1.7B Instruct2026.04 | 59.7 | |
| OPSDBackbone=Qwen3-1.7B Instruct2026.04 | 57.2 | |
| No updateBackbone=Qwen3-1.7B Instruct2026.04 | 51.5 | |
| GRPOBackbone=Qwen3-1.7B Instruct2026.04 | 51.1 | |
| SFTBackbone=Qwen3-1.7B Instruct2026.04 | 48.4 |