Visual Data-Type Understanding does not emerge from scaling Vision-Language Models.

Vishaal Udandarao, Max F. Burg, Samuel Albanie, Matthias Bethge

Browse the full ICLR paper archive.