ResearchBenchmarksSemantic Robustness Evaluation on Flowers102 (synthetic)Follow2.3ID Color ErrorI-Spec2.0563.7035.356.997Jun 17, 2026Evaluation ResultsMethodMethodLinksID Color ErrorID Viewpoint ErrorOOD Color ErrorOOD Shape ErrorI-SpecDescription=Image-spec...Description=Image-specified transformation2026.062.347.56.6T-SpecDescription=Text-speci...Description=Text-specified transformation2026.066.58.29.48.7ExactLine2026.068.412.610.119.3