List your company

Medical Visual Question Answering (True/False) on IU-XRay (test)

97.78AccuracyLIAR
-3.911222.489448.8975.2906May 11, 2026

Evaluation results

MethodLinks
LIARModel=GPT-5, Retriever=Filtered2026.0597.78
LIARModel=GPT-5, Retriever=BGE-VL2026.0595.78
LIARModel=GPT-5, Retriever=CLIP2026.0594.6
Clean RAG (avg.)Model=GPT-5, Retriever=Average2026.0593.9
M³AttModel=GPT-5, Retriever=BGE-VL2026.0593.54
M³AttModel=GPT-5, Retriever=CLIP2026.0591.96
Clean RAG (avg.)Model=GPT-4o, Retriever=Average2026.0589.64
w/o RAGModel=GPT-5, Retriever=None2026.0588.28
M³AttModel=GPT-5, Retriever=Filtered2026.0587.44
LIARModel=GPT-4o, Retriever=BGE-VL2026.0586.04
LIARModel=GPT-4o, Retriever=CLIP2026.0583.9
M³AttModel=GPT-4o, Retriever=BGE-VL2026.0580.44
LIARModel=Gemini-2.5, Retriever=CLIP2026.0580.02
M³AttModel=GPT-4o, Retriever=CLIP2026.0577.88
LIARModel=LLaVA-Med, Retriever=CLIP2026.0576.92
M³AttModel=Gemini-2.5, Retriever=CLIP2026.0576.12
LIARModel=Gemini-2.5, Retriever=Filtered2026.0574.88
Clean RAG (avg.)Model=Gemini-2.5, Retriever=Average2026.0572.6
Clean RAG (avg.)Model=LLaVA-Med, Retriever=Average2026.0572.51
LIARModel=GPT-4o, Retriever=Filtered2026.0571.06
w/o RAGModel=Gemini-2.5, Retriever=None2026.0568.2
w/o RAGModel=GPT-4o, Retriever=None2026.0567.36
M³AttModel=Gemini-2.5, Retriever=Filtered2026.0566.52
M³AttModel=LLaVA-Med, Retriever=CLIP2026.0565.28
M³AttModel=GPT-4o, Retriever=Filtered2026.0561.5
M³AttModel=Gemini-2.5, Retriever=BGE-VL2026.0559.84
LIARModel=Gemini-2.5, Retriever=BGE-VL2026.0558.78
LIARModel=Claude-4.5, Retriever=CLIP2026.0555.9
Clean RAG (avg.)Model=Claude-4.5, Retriever=Average2026.0553.2
LIARModel=LLaVA-Med, Retriever=BGE-VL2026.0551.96
M³AttModel=Claude-4.5, Retriever=CLIP2026.0547.04
M³AttModel=LLaVA-Med, Retriever=BGE-VL2026.0546.56
LIARModel=Claude-4.5, Retriever=BGE-VL2026.0543.26
M³AttModel=Claude-4.5, Retriever=BGE-VL2026.0538.3
LIARModel=LLaVA-Med, Retriever=Filtered2026.0533.46
LIARModel=Claude-4.5, Retriever=Filtered2026.0528.84
M³AttModel=LLaVA-Med, Retriever=Filtered2026.0528.04
M³AttModel=Claude-4.5, Retriever=Filtered2026.0520.5
w/o RAGModel=LLaVA-Med, Retriever=None2026.0514.64
w/o RAGModel=Claude-4.5, Retriever=None2026.050