Medical Visual Question Answering on VQA-RAD 10
78.8Closed AccuracyMed-Gemini
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Med-Gemini2026.06 | 78.8 | — | — | |
| MedG. (4B)Parameters=4B2026.06 | 69.1 | 49.9 | 20.8 | |
| RadGrounderTraining Data=Report, VQA, Detect.2026.06 | 68.1 | 47.9 | 28.6 | |
| RadGrounderTraining Data=Report, VQA, Slake, VQA-RAD, Detect.2026.06 | 65.9 | 50.4 | 46.6 | |
| RadGrounderTraining Data=Report, VQA, Slake, VQA-RAD, Detect., Vision Encoder=Frozen SigLIP (adapted)2026.06 | 64.7 | 50.7 | 44.3 | |
| RadGrounderTraining Data=Report, VQA, Slake, VQA-RAD, Detect., Vision Encoder=Unfreeze orig. SL.2026.06 | 64.5 | 50.2 | 44.9 | |
| RadGrounderTraining Data=Report, VQA, Slake, VQA-RAD2026.06 | 63.2 | 49.2 | 42.8 | |
| RadGrounderTraining Data=Report, VQA, Slake, VQA-RAD, Detect., Vision Encoder=Unfreeze finet. SL.2026.06 | 62.7 | 48.7 | 44.3 | |
| PaliG. 2 (3B)Parameters=3B2026.06 | 55.6 | 37.9 | 26 | |
| RadGrounderTraining Data=Report, VQA2026.06 | 52.9 | 39.6 | 33 | |
| RadGrounderTraining Data=Report, VQA, Segment2026.06 | 52.9 | 40.2 | 35.9 | |
| Gemma 3 (4B)Parameters=4B2026.06 | 33.6 | 33.6 | 11.1 |