List your company

Medical factuality evaluation on IU-Chest X-ray (test)

68.31AccuracyStandard RAG + BAIR
63.57864.806566.03567.2635May 7, 2026

Evaluation results

MethodLinks
Standard RAG + BAIRBackbone=MedGemma-4B2026.0568.3124.258.4
MAD-RAG + BAIRBackbone=MedGemma-4B2026.0568.2524.198.45
MAD-RAGBackbone=MedGemma-4B2026.0568.0324.589.01
MS-PoE + BAIRBackbone=MedGemma-4B2026.0567.621.017.43
MS-PoEBackbone=MedGemma-4B2026.0567.1720.687.85
MS-PoE + BAIRBackbone=CheXagent-8B2026.0566.611.725.05
MS-PoEBackbone=CheXagent-8B2026.0566.4514.196.62
Standard RAGBackbone=MedGemma-4B2026.0566.332813.56
LongLLMLingua + BAIRBackbone=CheXagent-8B2026.0566.2416.648.36
Standard RAG + BAIRBackbone=CheXagent-8B2026.0566.1712.386.04
MAD-RAG + BAIRBackbone=CheXagent-8B2026.0566.0811.835.87
LongLLMLingua + BAIRBackbone=MedGemma-4B2026.0565.9125.2112.46
Visual-focus Instr. + BAIRBackbone=MedGemma-4B2026.0565.6924.4812.33
LongLLMLinguaBackbone=MedGemma-4B2026.0565.6625.1412.79
Visual-focus Instr. + BAIRBackbone=CheXagent-8B2026.0565.5812.897.19
BaselineBackbone=CheXagent-8B2026.0565.51——
Visual-focus Instr.Backbone=MedGemma-4B2026.0565.3826.6814.13
LongLLMLinguaBackbone=CheXagent-8B2026.0565.3323.6313.63
Visual-focus Instr.Backbone=CheXagent-8B2026.0564.9415.899.81
MAD-RAGBackbone=CheXagent-8B2026.0564.7916.1610.19
Standard RAGBackbone=CheXagent-8B2026.0564.5615.8810.36
BaselineBackbone=MedGemma-4B2026.0563.76——