DropPos: Pre-Training Vision Transformers by Reconstructing Dropped Positions Paper • 2309.03576 • Published Sep 7, 2023 • 2