Clinical Diagnosis on MedQA
13.5Exact Match (EM)AutoRefine-Embedding
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| AutoRefine-EmbeddingBackbone=Qwen2.5-7B2026.05 | 13.5 | 22.1 | 17.8 | |
| IGPOBackbone=Qwen2.5-7B2026.05 | 13 | 22.6 | 17.8 | |
| AutoRefineBackbone=Qwen2.5-7B2026.05 | 12 | 19.8 | 15.9 | |
| Search-R1Backbone=Qwen2.5-7B2026.05 | 11 | 20.8 | 15.9 | |
| AutoRefine-ICDTreeBackbone=Qwen2.5-7B2026.05 | 11 | 23.7 | 17.35 | |
| C-MIGBackbone=Qwen2.5-7B2026.05 | 11 | 20.4 | 15.7 | |
| AutoRefine-HardSearchBackbone=Qwen2.5-7B2026.05 | 10.5 | 23.4 | 16.95 | |
| AutoRefine-HardDocBackbone=Qwen2.5-7B2026.05 | 10.5 | 20.2 | 15.35 | |
| Qwen2.5-7BBackbone=Qwen2.5-7B2026.05 | 0 | 2.9 | 1.45 |