A benchmark to evaluate vision and language models' understanding of figurative language.