1 paper
Min Zhao, Hongzhou Zhu, Yingze Wang +6
Despite advances, video diffusion transformers still struggle to generalize beyond their training length, a challenge we term video length extrapolation. We identify two failure mo…