ResearchBenchmarksImage-to-Text Generation Evaluation on COCOFollow6.65AccuracyCMPC5.32925.67216.0156.3579Jan 3, 2025Evaluation ResultsMethodMethodLinksAccuracyDetailednessCMPCModel=LLaVA 1.5, Base...Model=LLaVA 1.5, Base Decoding=Baseline2025.016.656.45CMPCModel=LLaVA 1.5, Base...Model=LLaVA 1.5, Base Decoding=ICD2025.016.616.52CMPCModel=LLaVA 1.5, Base...Model=LLaVA 1.5, Base Decoding=VCD2025.016.576.54ICDModel=LLaVA 1.5Model=LLaVA 1.52025.016.036.48VCDModel=LLaVA 1.5Model=LLaVA 1.52025.015.835.93BaselineModel=LLaVA 1.5Model=LLaVA 1.52025.015.385.88