ResearchBenchmarksVision-Language Reasoning on SelfEval Benchmark (test)Vision-Language Reasoning on SelfEval Benchmark (test)Follow55.4Attribute BindingCLIP49.78451.24252.754.158Nov 17, 2023Evaluation resultsMethodMethodLinksAttribute BindingColorCountShapeSpatialText CorruptionCLIPBackbone=ViT-L/14, zer...Backbone=ViT-L/14, zero-shot=true2023.1155.485.267.891.140.551Randommode=chance accuracymode=chance accuracy2023.11502525332520