1 paper · 1 filter
Xiaoyan Zhang, Zechen Bai, Haofan Wang +1
Recent unified models such as Bagel demonstrate that paired image-edit data can effectively align multiple visual tasks within a single diffusion transformer. However, these models…