1 paper
Geunhyeok Yu, Sunjae Jeong, Yoonyoung Choi +2
Vision Transformers are widely adopted as the backbone of vision foundation models, but they are known to produce high-norm artifacts that degrade representation quality. When know…