Visual Misinterpretation Hallucination Evaluation on MM-VisHal (test)
88.3Accuracy (A)MedQwen
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MedQwenevaluation_mode=close-ended2026.04 | 88.3 | 79.8 | 89.6 | 84.7 | 85.4 | |
| GPT-4oevaluation_mode=close-ended2026.04 | 77.5 | 69.7 | 70.8 | 84.6 | 74.1 | |
| LLaVA-Med-1.5evaluation_mode=close-ended2026.04 | 61.9 | 39.7 | 49.9 | 48.3 | 49.9 | |
| LLaVA-NeXT 13Bevaluation_mode=close-ended2026.04 | 57.7 | 43 | 55.1 | 44.5 | 51 | |
| LLaVA-NeXT 7Bevaluation_mode=close-ended2026.04 | 57.6 | 42.6 | 50.7 | 45.1 | 49.4 | |
| LLaVA-Medevaluation_mode=close-ended2026.04 | 52.5 | 35.7 | 58.4 | 48.5 | 49.9 | |
| CheXagentevaluation_mode=close-ended2026.04 | 52.4 | 51.6 | 57.2 | 46.4 | 52.9 | |
| Med-Flamingoevaluation_mode=close-ended2026.04 | 52.3 | 49.7 | 58.8 | 32.7 | 50.7 | |
| LLM-CXRevaluation_mode=close-ended2026.04 | 48.6 | 46 | 51.3 | 31.4 | 46.1 | |
| MiniGPT-4evaluation_mode=close-ended2026.04 | 48.3 | 53.7 | 55.3 | 43 | 51.2 |