Visual Question Answering on SLAKE Open
89.9Token RecallMEDSIGHT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MEDSIGHTFine-tuning protocol=LoRA, LoRA rank=322026.06 | 89.9 | — | |
| MedVInT-TEParams=8.6B2026.05 | 88.2 | — | |
| FAVP (Vicuna)Params=8.5B2026.05 | 87.2 | — | |
| LLaVA-MedParams=7B2026.05 | 87.1 | — | |
| LLaVA-MedBackbone=BioMedCLIP2025.08 | 87 | — | |
| RoiMAMParams=1.7B2026.05 | 86.5 | — | |
| HuatuoGPT-Vision 7BFine-tuning protocol=LoRA, LoRA rank=322026.06 | 85.6 | — | |
| ExGra-MedFine-tuning protocol=LoRA, LoRA rank=322026.06 | 85.1 | — | |
| LLaVA-MedBackbone=LLaVA2025.08 | 85 | — | |
| MedVInT-TDParams=8.6B2026.05 | 84.5 | — | |
| MMed-RAGBackbone=LLaVA-Med2025.08 | 84 | — | |
| CLAREBackbone=Qwen2-vl2025.08 | 84 | 84 | |
| CLAREBackbone=Pixtral2025.08 | 84 | 82 | |
| CLAREBackbone (Size)=Qwen2-vl (7B)2025.08 | 84 | — | |
| CLAREBackbone (Size)=Pixtral (12B)2025.08 | 84 | — | |
| LLaVA-Med 1.5Fine-tuning protocol=LoRA, LoRA rank=322026.06 | 83.3 | — | |
| LLaVA-MedBackbone=Vicuna2025.08 | 83 | — | |
| MMed-RAGBackbone (Size)=LLaVA (7B)2025.08 | 81 | — | |
| FT RAGBackbone (Size)=Qwen2-vl (7B)2025.08 | 81 | — | |
| CLARE + PDistBackbone (Size)=Qwen2-vl (7B)2025.08 | 81 | — | |
| FT RAGBackbone (Size)=Pixtral (12B)2025.08 | 81 | — | |
| CLARE + PDistBackbone (Size)=Pixtral (12B)2025.08 | 81 | — | |
| ReaderBackbone (Size)=Qwen2-vl (7B)2025.08 | 80 | — | |
| ReaderBackbone (Size)=Pixtral (12B)2025.08 | 80 | — | |
| PubMedCLIPParams=1.6B2026.05 | 78.4 | — | |
| PMC-CLIPParams=1.6B2026.05 | 72.7 | — | |
| BiomedGPT-B2025.08 | — | 85 |