Medical VQA on BiomedParse In-Domain (test)
78.68AccuracyVITAL
Evaluation Results
| Method | Links | |
|---|---|---|
| VITAL2026.05 | 78.68 | |
| Fleming-VL2026.05 | 61.49 | |
| Claude-Opus-4.62026.05 | 55.12 | |
| GPT-5.42026.05 | 53.52 | |
| HuatuoGPT-Vision2026.05 | 50.27 | |
| LLaVA-Med2026.05 | 45.31 | |
| LVRWeights=Fine-tuned2026.05 | 44.97 | |
| SIM-CoTBackbone=Qwen3-VL-8B + LoRA2026.05 | 43.37 | |
| MCOUTWeights=Fine-tuned2026.05 | 38.32 | |
| CoconutBackbone=Qwen3-VL-8B + LoRA2026.05 | 38.15 | |
| VividMed2026.05 | 37.74 | |
| CODIBackbone=Qwen3-VL-8B + LoRA2026.05 | 36.77 | |
| Qwen3-VL-8BThinking=true2026.05 | 33.19 | |
| Qwen3-VL-8BZero-shot=true2026.05 | 32.01 | |
| LVRWeights=Original2026.05 | 28.85 | |
| MCOUTWeights=Original2026.05 | 22.88 |