Mathematical Reasoning on AMC (acc@1, pass@8)
43.4Accuracy@1InfoSFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| InfoSFTBackbone=Qwen-Math-7B2026.05 | 43.4 | 69.1 | |
| InfoSFTBackbone=Qwen-Math-1.5B2026.05 | 41 | 67.5 | |
| SFTBackbone=Qwen-Math-7B2026.05 | 39.9 | 62.1 | |
| baseBackbone=Qwen-Math-7B2026.05 | 39.8 | 68 | |
| baseBackbone=Qwen-Math-1.5B2026.05 | 34.9 | 63.4 | |
| DFTBackbone=Qwen-Math-7B2026.05 | 34.1 | 58.9 | |
| SFTBackbone=Qwen-Math-1.5B2026.05 | 31.3 | 54.4 | |
| DFTBackbone=Qwen-Math-1.5B2026.05 | 30.1 | 56.4 | |
| InfoSFTBackbone=Llama-3.1-8B2026.05 | 13.2 | 35.2 | |
| SFTBackbone=Llama-3.1-8B2026.05 | 8.4 | 31.7 | |
| DFTBackbone=Llama-3.1-8B2026.05 | 6 | 15.2 | |
| baseBackbone=Llama-3.1-8B2026.05 | 0.9 | 6.8 |