ResearchBenchmarksVision-Language Evaluation Metric Sensitivity Analysis on Pascal-50S cultural subsetFollow-0.5Median Sensitivity (%)FLEUR-2.164-1.732-1.3-0.868May 23, 2026Evaluation ResultsMethodMethodLinksMedian Sensitivity (%)FLEUR2026.05-0.5Judge2026.05-0.7UMIC2026.05-0.9PAC-S2026.05-1.7CLIPScore2026.05-2.1