ResearchBenchmarksPrototypicality Bias Evaluation on ProtoBias DemographyFollow0.515Correct Ranking MarginGPT-4o0.051160.171580.2920.41242Jan 8, 2026Evaluation ResultsMethodMethodLinksCorrect Ranking MarginIncorrect Ranking MarginGPT-4o2026.010.5150.272GPT-52026.010.420.048PROTOSCORE2026.010.3580.057PickScore2026.010.1860.217CLIPScore2026.010.1510.127VQAScore2026.010.0690.124