Medical VQA on In-house (Held-out)
80.5AccuracyVITAL
Evaluation Results
| Method | Links | |
|---|---|---|
| VITAL2026.05 | 80.5 | |
| GPT-5.42026.05 | 73.8 | |
| Fleming-VL2026.05 | 73.6 | |
| Claude-Opus-4.62026.05 | 72.5 | |
| HuatuoGPT-Vision2026.05 | 66.1 | |
| Qwen3-VL-8BThinking=true2026.05 | 55.3 | |
| SIM-CoTBackbone=Qwen3-VL-8B + LoRA2026.05 | 52.9 | |
| LVRWeights=Fine-tuned2026.05 | 52.7 | |
| Qwen3-VL-8BZero-shot=true2026.05 | 52.6 | |
| MCOUTWeights=Fine-tuned2026.05 | 51.5 | |
| CODIBackbone=Qwen3-VL-8B + LoRA2026.05 | 48.4 | |
| VividMed2026.05 | 47.1 | |
| LLaVA-Med2026.05 | 45.7 | |
| CoconutBackbone=Qwen3-VL-8B + LoRA2026.05 | 41.5 | |
| LVRWeights=Original2026.05 | 39.7 | |
| MCOUTWeights=Original2026.05 | 38.2 |