Mathematical Reasoning on MATH 500 (Avg@k, Pass@k)
74.95Avg@kRPSFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RPSFTBackbone=Qwen2.5-7B-Instruct2026.05 | 74.95 | 87.4 | |
| DFTBackbone=Qwen2.5-7B-Instruct2026.05 | 73.35 | 84.2 | |
| BaseBackbone=Qwen2.5-7B-Instruct2026.05 | 72.85 | 84.6 | |
| SFTBackbone=Qwen2.5-7B-Instruct2026.05 | 72.85 | 87.6 | |
| IWBackbone=Qwen2.5-7B-Instruct2026.05 | 72.6 | 89 | |
| BaseBackbone=Qwen2.5-3B-Instruct2026.05 | 63.35 | 79.2 | |
| RPSFTBackbone=Qwen2.5-3B-Instruct2026.05 | 63.35 | 79.8 | |
| DFTBackbone=Qwen2.5-3B-Instruct2026.05 | 62.05 | 75.8 | |
| IWBackbone=Qwen2.5-3B-Instruct2026.05 | 61.05 | 79.6 | |
| SFTBackbone=Qwen2.5-3B-Instruct2026.05 | 60.95 | 80.6 | |
| RPSFTBackbone=Llama-3.1-8B-Instruct2026.05 | 60.45 | 77.2 | |
| SFTBackbone=Llama-3.1-8B-Instruct2026.05 | 59.45 | 78.6 | |
| IWBackbone=Llama-3.1-8B-Instruct2026.05 | 58.5 | 76.2 | |
| DFTBackbone=Llama-3.1-8B-Instruct2026.05 | 46.4 | 61.2 | |
| BaseBackbone=Llama-3.1-8B-Instruct2026.05 | 43.65 | 65 |