Light Forcing: Accelerating Autoregressive Video Diffusion via Sparse Attention 事件

Light Forcing: Accelerating Autoregressive Video Diffusion via Sparse Attention arXiv:2602.04789v2 Announce Type: replace Abstract: Advanced autoregressive (AR) video generation models have improved visual fidelity and interactivity, but the quadratic complexity of attention remains a primary bottleneck for efficient deployment. While existing sparse attention solutions have shown promise on bidirectional models, we identify that applying these solutions to AR models leads to considerable perfo