List your company

Medical Visual Question Answering (True/False) on MIMIC (test)

73.2AccuracyLIAR
-2.92816.83636.656.364May 11, 2026

Evaluation results

MethodLinks
LIARModel=GPT-5, Retriever=BGE-VL2026.0573.2
Clean RAG (avg.)Model=GPT-5, Retriever=Average2026.0573.16
M³AttModel=GPT-5, Retriever=BGE-VL2026.0566.7
LIARModel=GPT-5, Retriever=CLIP2026.0566.4
w/o RAGModel=LLaVA-Med, Retriever=None2026.0566
Clean RAG (avg.)Model=LLaVA-Med, Retriever=Average2026.0565.88
M³AttModel=GPT-5, Retriever=CLIP2026.0563.46
LIARModel=LLaVA-Med, Retriever=BGE-VL2026.0561.86
w/o RAGModel=GPT-5, Retriever=None2026.0561.34
LIARModel=LLaVA-Med, Retriever=CLIP2026.0561.04
LIARModel=GPT-5, Retriever=Filtered2026.0558.46
M³AttModel=LLaVA-Med, Retriever=CLIP2026.0555.44
M³AttModel=LLaVA-Med, Retriever=BGE-VL2026.0551.3
M³AttModel=GPT-5, Retriever=Filtered2026.0550
Clean RAG (avg.)Model=Claude-4.5, Retriever=Average2026.0541.26
LIARModel=GPT-4o, Retriever=CLIP2026.0540.96
Clean RAG (avg.)Model=Gemini-2.5, Retriever=Average2026.0540.6
Clean RAG (avg.)Model=GPT-4o, Retriever=Average2026.0537.84
LIARModel=GPT-4o, Retriever=BGE-VL2026.0537.22
M³AttModel=GPT-4o, Retriever=CLIP2026.0536.78
LIARModel=Gemini-2.5, Retriever=CLIP2026.0536.18
LIARModel=Claude-4.5, Retriever=BGE-VL2026.0535.9
LIARModel=Claude-4.5, Retriever=CLIP2026.0534.48
M³AttModel=Gemini-2.5, Retriever=CLIP2026.0533.76
LIARModel=Gemini-2.5, Retriever=BGE-VL2026.0532.92
M³AttModel=GPT-4o, Retriever=BGE-VL2026.0530.78
LIARModel=LLaVA-Med, Retriever=Filtered2026.0530.62
M³AttModel=Gemini-2.5, Retriever=BGE-VL2026.0529.58
M³AttModel=Claude-4.5, Retriever=CLIP2026.0528.68
LIARModel=Gemini-2.5, Retriever=Filtered2026.0526.54
M³AttModel=Claude-4.5, Retriever=BGE-VL2026.0524.5
w/o RAGModel=GPT-4o, Retriever=None2026.0524
M³AttModel=LLaVA-Med, Retriever=Filtered2026.0523.34
w/o RAGModel=Gemini-2.5, Retriever=None2026.0522.66
M³AttModel=Gemini-2.5, Retriever=Filtered2026.0521.34
LIARModel=GPT-4o, Retriever=Filtered2026.0521.08
M³AttModel=GPT-4o, Retriever=Filtered2026.0514
w/o RAGModel=Claude-4.5, Retriever=None2026.050
LIARModel=Claude-4.5, Retriever=Filtered2026.050
M³AttModel=Claude-4.5, Retriever=Filtered2026.050