Medical Visual Question Answering on MIMIC-CXR-VQA (test)
65.2AccuracyMeissa
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MeissaFramework=Framework I, Inference Mode=Agent-based, Agent System=MedRAX2026.03 | 65.2 | — | — | |
| Gemini-3-flashFramework=Framework I, Inference Mode=Agent-based, Agent System=MedRAX2026.03 | 65 | — | — | |
| Gemini-3-flashFramework=Framework III, Inference Mode=Agent-based, Agent System=MDAgents2026.03 | 64 | — | — | |
| M4CXR2024.08 | 62.3 | 70.3 | 66.4 | |
| MeissaFramework=Framework III, Inference Mode=Agent-based, Agent System=MDAgents2026.03 | 59.4 | — | — | |
| InstFreeResponse Shuffling (RS)=true2026.03 | 59.3 | — | — | |
| CheXagent2024.08 | 59 | 72.8 | 62.5 | |
| InstFree2026.03 | 58.9 | — | — | |
| GPT-4VFramework=Framework III, Inference Mode=Agent-based, Agent System=MDAgents2026.03 | 55.9 | — | — | |
| GPT-4oFramework=Framework I, Inference Mode=Agent-based, Agent System=MedRAX2026.03 | 55.6 | — | — | |
| Qwen3-VL-4BFramework=Framework III, Inference Mode=Direct Inference, Training Protocol=SFT2026.03 | 55.2 | — | — | |
| Qwen3-VL-4BFramework=Framework III, Inference Mode=Agent-based, Agent System=MDAgents2026.03 | 54.9 | — | — | |
| Qwen3-VL-4BFramework=Framework I, Inference Mode=Agent-based, Agent System=MedRAX2026.03 | 51.4 | — | — | |
| Qwen3-VL-4BFramework=Framework I, Inference Mode=Direct Inference, Training Protocol=SFT2026.03 | 49.7 | — | — | |
| LLaMA-3.2-11B-VisionFine-tuning=w/o FT2026.03 | 46.9 | — | — | |
| MedGemma-4B2026.03 | 44.7 | — | — | |
| Gemini-3-flashFramework=Framework I, Inference Mode=Direct Inference2026.03 | 43.6 | — | — | |
| Gemini-3-flashFramework=Framework III, Inference Mode=Direct Inference2026.03 | 43.6 | — | — | |
| GPT-4oFramework=Framework I, Inference Mode=Direct Inference2026.03 | 40 | — | — | |
| GPT-4VFramework=Framework III, Inference Mode=Direct Inference2026.03 | 40 | — | — | |
| RadFM2024.08 | 11.2 | 36.7 | 13.8 | |
| RaDialogvariant=LLaVA version2024.08 | 0 | 43 | 5 |