Mathematical Reasoning on AIME25 (Avg@k, Pass@k)
19.38Avg@kSFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SFTBackbone=Qwen2.5-7B-Instruct2026.05 | 19.38 | 36.67 | |
| RPSFTBackbone=Qwen2.5-7B-Instruct2026.05 | 18.75 | 43.33 | |
| IWBackbone=Qwen2.5-7B-Instruct2026.05 | 18.54 | 36.67 | |
| DFTBackbone=Qwen2.5-7B-Instruct2026.05 | 9.38 | 33.33 | |
| BaseBackbone=Qwen2.5-7B-Instruct2026.05 | 7.7 | 36.67 | |
| IWBackbone=Llama-3.1-8B-Instruct2026.05 | 6.88 | 26.67 | |
| RPSFTBackbone=Llama-3.1-8B-Instruct2026.05 | 6.67 | 20 | |
| IWBackbone=Qwen2.5-3B-Instruct2026.05 | 6.04 | 30 | |
| SFTBackbone=Llama-3.1-8B-Instruct2026.05 | 5 | 13.33 | |
| SFTBackbone=Qwen2.5-3B-Instruct2026.05 | 5 | 23.33 | |
| RPSFTBackbone=Qwen2.5-3B-Instruct2026.05 | 5 | 23.33 | |
| DFTBackbone=Qwen2.5-3B-Instruct2026.05 | 4.79 | 23.33 | |
| BaseBackbone=Qwen2.5-3B-Instruct2026.05 | 2.29 | 20 | |
| DFTBackbone=Llama-3.1-8B-Instruct2026.05 | 1.88 | 16.67 | |
| BaseBackbone=Llama-3.1-8B-Instruct2026.05 | 0.83 | 13.33 |