Medical Reasoning on MedQA-USMLE Medical (val)
83Accuracy+GPT-4o
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GPT-4o2025.10 | 83 | 99.2 | 86.3 | |
| CLARITYModel Size=7B2025.10 | 63.3 | 89.5 | 66.2 | |
| Data Reformulation2025.10 | 62.4 | 89 | 65.3 | |
| PRM RL2025.10 | 60.2 | 96.4 | 60.8 | |
| Refine-then-Monitor2025.10 | 59.7 | 89.3 | 62.3 | |
| Standard RL2025.10 | 57.8 | 81.3 | 63.6 | |
| CLARITYModel Size=3B2025.10 | 50.1 | 86 | 55.1 | |
| Qwen2.5Model Size=7B2025.10 | 42.5 | 88.7 | 49.4 | |
| Qwen2.5Model Size=3B2025.10 | 34.7 | 77.7 | 38.2 |