4M-21: An Any-to-Any Vision Model for Tens of Tasks and Modalities Paper • 2406.09406 • Published Jun 13 • 13 • 2