Medical Diagnosis on PMC-Patients
62.6Similarity ScoreMedExAgent-8B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MedExAgent-8BVariant=full2026.05 | 62.6 | 34.4 | 31.4 | |
| MedExAgent-8BVariant=w/o noise injection2026.05 | 62.4 | 26.9 | 25.4 | |
| MedExAgent-8BVariant=w/o tool-match reward2026.05 | 59.3 | 21.9 | 20.3 | |
| Aloe-Beta-70BModel=Aloe-Beta-70B2026.05 | 59 | 17.1 | 21.9 | |
| Qwen3-14BModel=Qwen3-14B2026.05 | 56.5 | 11.3 | 14.3 | |
| Qwen3-32BModel=Qwen3-32B2026.05 | 55.1 | 12.7 | 15.7 | |
| HuatuoGPT-o1-70BModel=HuatuoGPT-o1-70B2026.05 | 53.4 | 12 | 12.4 | |
| Qwen3-8BModel=Qwen3-8B2026.05 | 53.2 | 9.2 | 12 | |
| Llama-3.1-8B-InstructModel=Llama-3.1-8B-Instruct2026.05 | 50.7 | 12.2 | 15.7 | |
| MedExAgent-8BVariant=SFT-only2026.05 | 50.6 | 23.1 | 22.1 | |
| MedGemma-27B-text-itModel=MedGemma-27B-text-it2026.05 | 46.9 | 10.8 | 12.3 | |
| MedExAgent-8BVariant=w/o cost penalty2026.05 | 16 | 10.2 | 9.4 |