Closed-format Question Answering on Abdomen (n=143)
84.6AccuracyGPT-5.4
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-5.42026.07 | 84.6 | |
| HR1.52026.07 | 80.4 | |
| Gemini-3-Flash-Preview2026.07 | 79 | |
| HR1.5+2026.07 | 74.8 | |
| HR12026.07 | 74.1 | |
| Claude Opus 4.72026.07 | 60.8 | |
| MedGemma-4B2026.07 | 60.5 | |
| LLaMa-4-Scout2026.07 | 58.7 | |
| MedGemma-27B2026.07 | 44.4 |