Mathematical Problem Solving on OlympiadBench (Accuracy, Refinement Gap)
71AccuracyselfRef@4
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| selfRef@4backbone=GSR-7B2025.08 | 71 | 2.7 | |
| BoN@4backbone=GSR-7B, Aggregation Model=RM-R1-7B2025.08 | 69.1 | — | |
| maj@5backbone=GSR-7B2025.08 | 68.8 | — | |
| maj@4backbone=GSR-7B2025.08 | 68.3 | — | |
| BoN@5backbone=GSR-7B, Aggregation Model=SkyworkRM-27B2025.08 | 67.7 | — | |
| Ref@4backbone=GSR-7B, Aggregation Model=Synthesizer-8B2025.08 | 67.7 | — | |
| BoN@4backbone=GSR-7B, Aggregation Model=SkyworkRM-27B2025.08 | 67.6 | — | |
| BoN@4backbone=GSR-7B, Aggregation Model=RRM-7B2025.08 | 64.6 | — | |
| pass@1backbone=GSR-7B2025.08 | 64.4 | — | |
| QwQ-Previewmode=pass@12025.08 | 61.2 | — | |
| selfRef@4backbone=Qwen2.5-7B-Instruct (Base)2025.08 | 43.8 | 0.3 | |
| maj@4backbone=Qwen2.5-7B-Instruct (Base)2025.08 | 43.5 | — | |
| BoN@4backbone=Qwen2.5-7B-Instruct (Base), Aggregation Model=SkyworkRM-27B2025.08 | 43 | — | |
| pass@1backbone=Qwen2.5-7B-Instruct (Base)2025.08 | 39.5 | — |