MMCOMPOSITION: Revisiting the Compositionality of Pre-trained Vision-Language Models Paper • 2410.09733 • Published 5 days ago • 6