1 paper
Yonghao Yu, Lang Huang, Runyi Li +2
Causal video generators must predict from the past, but they need not learn only from it. In streaming autoregressive video diffusion, each emitted segment becomes a commitment tha…