List your company
ResearchFormerly WizWand

Medical Visual Question Answering (Multiple-Choice) on MIMIC (test)

82.56AccuracyLIAR
1.367222.446143.52564.6039May 11, 2026
Updated 5mo ago

Evaluation results

MethodLinks
LIARModel=GPT-5, Retriever=BGE-VL2026.0582.56
Clean RAG (avg.)Model=GPT-5, Retriever=Average2026.0581.24
w/o RAGModel=GPT-5, Retriever=None2026.0574.69
LIARModel=GPT-5, Retriever=CLIP2026.0574.5
M³AttModel=GPT-5, Retriever=BGE-VL2026.0572.26
M³AttModel=GPT-5, Retriever=CLIP2026.0570.33
Clean RAG (avg.)Model=GPT-4o, Retriever=Average2026.0569.57
LIARModel=GPT-4o, Retriever=BGE-VL2026.0567.83
Clean RAG (avg.)Model=Claude-4.5, Retriever=Average2026.0566.41
LIARModel=Claude-4.5, Retriever=CLIP2026.0566.38
LIARModel=Claude-4.5, Retriever=BGE-VL2026.0564.68
LIARModel=GPT-4o, Retriever=CLIP2026.0564.09
M³AttModel=Claude-4.5, Retriever=CLIP2026.0561.41
M³AttModel=GPT-4o, Retriever=CLIP2026.0559.98
w/o RAGModel=Claude-4.5, Retriever=None2026.0559.26
M³AttModel=GPT-4o, Retriever=BGE-VL2026.0558.84
w/o RAGModel=GPT-4o, Retriever=None2026.0558.02
LIARModel=Gemini-2.5, Retriever=BGE-VL2026.0555.62
M³AttModel=Claude-4.5, Retriever=BGE-VL2026.0555.43
Clean RAG (avg.)Model=Gemini-2.5, Retriever=Average2026.0552.66
LIARModel=GPT-5, Retriever=Filtered2026.0547.03
M³AttModel=Gemini-2.5, Retriever=BGE-VL2026.0546.34
M³AttModel=GPT-5, Retriever=Filtered2026.0544.44
LIARModel=Gemini-2.5, Retriever=CLIP2026.0542.09
M³AttModel=Gemini-2.5, Retriever=CLIP2026.0539.21
LIARModel=GPT-4o, Retriever=Filtered2026.0537.32
M³AttModel=GPT-4o, Retriever=Filtered2026.0534.57
w/o RAGModel=Gemini-2.5, Retriever=None2026.0533.96
LIARModel=Claude-4.5, Retriever=Filtered2026.0532.97
M³AttModel=Claude-4.5, Retriever=Filtered2026.0530.24
LIARModel=Gemini-2.5, Retriever=Filtered2026.0522.37
M³AttModel=Gemini-2.5, Retriever=Filtered2026.0520.99
LIARModel=LLaVA-Med, Retriever=Filtered2026.0512.9
M³AttModel=LLaVA-Med, Retriever=Filtered2026.0511.73
M³AttModel=LLaVA-Med, Retriever=CLIP2026.0510.77
LIARModel=LLaVA-Med, Retriever=CLIP2026.0510.74
w/o RAGModel=LLaVA-Med, Retriever=None2026.058.64
Clean RAG (avg.)Model=LLaVA-Med, Retriever=Average2026.058.5
M³AttModel=LLaVA-Med, Retriever=BGE-VL2026.056.36
LIARModel=LLaVA-Med, Retriever=BGE-VL2026.054.49