1 paper
Xiaoyan Liu, Jiaxin Liu, Kangrui Li +1
Current 4D generation paradigms are often bottlenecked by a sequential decoupling design: video is generated first, followed by 3D reconstruction, leading to high interaction laten…