Multiple-choice VQA on CBIS
68.3AccuracyInstFree
Evaluation results
| Method | Links | |
|---|---|---|
| InstFreeresponse shuffling=true2026.03 | 68.3 | |
| InstFree2026.03 | 65 | |
| LLaMA-3.2-11B-Visionfine-tuning=true2026.03 | 63.9 | |
| LLaMA-3.2-11B-Visionfine-tuning=false2026.03 | 43 | |
| MedGemma-4Bfine-tuning=true2026.03 | 42.5 | |
| PubMedVision-7Bfine-tuning=false2026.03 | 40.6 | |
| Qwen2.5-VL-3Bfine-tuning=true2026.03 | 39.3 | |
| MedGemma-4Bfine-tuning=false2026.03 | 39.2 | |
| Qwen2.5-VL-3Bfine-tuning=false2026.03 | 36 | |
| BLIP-2fine-tuning=false2026.03 | 4.6 |