Focused Forcing: Content-Aware Per-Frame KV Selection for Efficient Autoregressive Video Diffusion
arXiv:2605.18346v2 Announce Type: replace-cross Abstract: Recent advances in autoregressive video diffusion have enabled sequential and streaming video generation. However, long-horizon generation requires increasingly large KV caches, making efficient compression without sacrificing quality challenging. Existing methods mostly select historical frames based on attention scores, but their context decisions remain coarse. When multiple frames are generated in the same chunk, these methods often…