1 paper
Yongwoo Kim, Sungmin Cha, Hyunsoo Kim +2
With the increasing versatility of text-to-image diffusion models, the ability to selectively erase undesirable concepts (e.g., harmful content) has become indispensable. However,…