Multiple-choice VQA on WBCAtt
68.4AccuracyInstFree
Evaluation results
| Method | Links | |
|---|---|---|
| InstFree2026.03 | 68.4 | |
| InstFreeresponse shuffling=true2026.03 | 65.2 | |
| LLaMA-3.2-11B-Visionfine-tuning=true2026.03 | 61.7 | |
| MedGemma-4Bfine-tuning=true2026.03 | 45.9 | |
| Qwen2.5-VL-3Bfine-tuning=true2026.03 | 39.4 | |
| MedGemma-4Bfine-tuning=false2026.03 | 37.3 | |
| PubMedVision-7Bfine-tuning=false2026.03 | 36.3 | |
| BLIP-2fine-tuning=false2026.03 | 35.2 | |
| LLaMA-3.2-11B-Visionfine-tuning=false2026.03 | 34.6 | |
| Qwen2.5-VL-3Bfine-tuning=false2026.03 | 32.8 |