Mathematical Reasoning on Minerva Math (Avg@k, Pass@k)
37.09Avg@kBase
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BaseBackbone=Qwen2.5-7B-Instruct2026.05 | 37.09 | 55.17 | |
| IWBackbone=Qwen2.5-7B-Instruct2026.05 | 34.05 | 55.51 | |
| RPSFTBackbone=Qwen2.5-7B-Instruct2026.05 | 33.82 | 56.62 | |
| DFTBackbone=Qwen2.5-7B-Instruct2026.05 | 33.73 | 48.53 | |
| SFTBackbone=Qwen2.5-7B-Instruct2026.05 | 31.52 | 58.09 | |
| RPSFTBackbone=Llama-3.1-8B-Instruct2026.05 | 27.48 | 50.74 | |
| SFTBackbone=Llama-3.1-8B-Instruct2026.05 | 26.88 | 47.06 | |
| IWBackbone=Llama-3.1-8B-Instruct2026.05 | 26.75 | 48.16 | |
| DFTBackbone=Qwen2.5-3B-Instruct2026.05 | 26.56 | 42.28 | |
| BaseBackbone=Qwen2.5-3B-Instruct2026.05 | 26.24 | 50 | |
| RPSFTBackbone=Qwen2.5-3B-Instruct2026.05 | 23.85 | 47.06 | |
| IWBackbone=Qwen2.5-3B-Instruct2026.05 | 23.71 | 45.22 | |
| SFTBackbone=Qwen2.5-3B-Instruct2026.05 | 22.75 | 47.42 | |
| BaseBackbone=Llama-3.1-8B-Instruct2026.05 | 22.33 | 44.85 | |
| DFTBackbone=Llama-3.1-8B-Instruct2026.05 | 20.17 | 37.5 |