Mathematical Reasoning on AIME 24 (Avg@k, Pass@k)
15.21Avg@kSFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SFTBackbone=Qwen2.5-7B-Instruct2026.05 | 15.21 | 50 | |
| RPSFTBackbone=Qwen2.5-7B-Instruct2026.05 | 14.38 | 43.33 | |
| IWBackbone=Qwen2.5-7B-Instruct2026.05 | 14.17 | 46.67 | |
| DFTBackbone=Qwen2.5-7B-Instruct2026.05 | 12.29 | 23.33 | |
| BaseBackbone=Qwen2.5-7B-Instruct2026.05 | 12.08 | 36.67 | |
| IWBackbone=Qwen2.5-3B-Instruct2026.05 | 6.67 | 26.66 | |
| RPSFTBackbone=Qwen2.5-3B-Instruct2026.05 | 6.46 | 33.33 | |
| SFTBackbone=Llama-3.1-8B-Instruct2026.05 | 5 | 20 | |
| BaseBackbone=Qwen2.5-3B-Instruct2026.05 | 4.79 | 20 | |
| RPSFTBackbone=Llama-3.1-8B-Instruct2026.05 | 4.38 | 20 | |
| DFTBackbone=Llama-3.1-8B-Instruct2026.05 | 4.17 | 13.33 | |
| SFTBackbone=Qwen2.5-3B-Instruct2026.05 | 3.96 | 30 | |
| BaseBackbone=Llama-3.1-8B-Instruct2026.05 | 3.13 | 20 | |
| IWBackbone=Llama-3.1-8B-Instruct2026.05 | 3.13 | 16.67 | |
| DFTBackbone=Qwen2.5-3B-Instruct2026.05 | 2.71 | 23.33 |