3 ms·MSFT Research: Making Visual Representations Matter in Vision-Language Models3 points by panabee 6y ago