End-to-End Training for Autoregressive Video Diffusion via Self-Resampling
IntermediateYuwei Guo, Ceyuan Yang et al.Dec 17arXiv
This paper fixes a common problem in video-making AIs where tiny mistakes snowball over time and ruin long videos.
#autoregressive video diffusion#exposure bias#teacher forcing