Mathematical Reasoning on OlympiadBench (Avg@k, Pass@k)
39.18Avg@kIW
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| IWBackbone=Qwen2.5-7B-Instruct2026.05 | 39.18 | 56.3 | |
| RPSFTBackbone=Qwen2.5-7B-Instruct2026.05 | 38 | 56.15 | |
| BaseBackbone=Qwen2.5-7B-Instruct2026.05 | 37.41 | 54.96 | |
| SFTBackbone=Qwen2.5-7B-Instruct2026.05 | 37.41 | 54.96 | |
| DFTBackbone=Qwen2.5-7B-Instruct2026.05 | 37.41 | 50.07 | |
| RPSFTBackbone=Qwen2.5-3B-Instruct2026.05 | 27.67 | 43.85 | |
| SFTBackbone=Qwen2.5-3B-Instruct2026.05 | 26.25 | 42.37 | |
| DFTBackbone=Llama-3.1-8B-Instruct2026.05 | 26.22 | 26.22 | |
| BaseBackbone=Qwen2.5-3B-Instruct2026.05 | 26.22 | 41.48 | |
| IWBackbone=Qwen2.5-3B-Instruct2026.05 | 26 | 42.67 | |
| IWBackbone=Llama-3.1-8B-Instruct2026.05 | 25.71 | 41.63 | |
| SFTBackbone=Llama-3.1-8B-Instruct2026.05 | 25.59 | 41.33 | |
| RPSFTBackbone=Llama-3.1-8B-Instruct2026.05 | 25.52 | 42.22 | |
| DFTBackbone=Qwen2.5-3B-Instruct2026.05 | 25 | 38.37 | |
| BaseBackbone=Llama-3.1-8B-Instruct2026.05 | 14.63 | 29.78 |