Visual Question Answering on PMC-VQA (test)
64.9AccuracyLiteMedCoT-VL
Evaluation Results
| Method | Links | |
|---|---|---|
| LiteMedCoT-VL2026.05 | 64.9 | |
| Qwen3-VL-4B2026.05 | 53.9 | |
| Qwen3-VL-2B2026.05 | 48.7 | |
| TMA-AllComponBase Model=Gemma-3-4B, Image Resolution=3-5x reduced2025.08 | 45.3 | |
| SmolVLM2-2.2B2026.05 | 41.5 | |
| Single-Agent BestBase Model=Gemma-3-4B, Image Resolution=3-5x reduced2025.08 | 40 | |
| Phi-3.5-vision2026.05 | 39.2 | |
| ReinPathParams=8B, Samples=22K2026.01 | 38.6 | |
| SFTPathParams=8B, Samples=13K2026.01 | 38.1 | |
| MedICaP-PMCVQA-GPT-42026.05 | 35.9 | |
| ReConcileBase Model=Gemma-3-4B, Image Resolution=3-5x reduced2025.08 | 35.3 | |
| LLaVA-1.5Params=13B, Samples=665K2026.01 | 35.1 | |
| LLaVA-Med2026.05 | 34.8 | |
| MedAgentsBase Model=Gemma-3-4B, Image Resolution=3-5x reduced2025.08 | 34.7 | |
| DyLanBase Model=Gemma-3-4B, Image Resolution=3-5x reduced2025.08 | 32.7 | |
| Quilt-LLaVAParams=7B, Samples=107K2026.01 | 32.4 | |
| InternVL2-2B2026.05 | 31.7 | |
| MDAgentsBase Model=Gemma-3-4B, Image Resolution=3-5x reduced2025.08 | 31.3 | |
| MedICap-GPT-42026.05 | 27.2 | |
| Open-Flamingo2026.05 | 26.4 | |
| LLaVA-Med_{MedITok}Evaluation Protocol=free-form2025.05 | 25.55 | |
| LLaVA-Med_{MedITok} w/o S1Evaluation Protocol=free-form2025.05 | 24.74 | |
| PMC-CLIP2026.05 | 24.7 | |
| BLIP-22026.05 | 24.3 | |
| LLaVA-Med_{UniTok}Evaluation Protocol=free-form2025.05 | 22.6 | |
| LLaVA-MedEvaluation Protocol=free-form2025.05 | 21.55 | |
| LLaVA-MedParams=7B, Samples=60K2026.01 | 1.3 |