Fetching the paper…

Training-free and Adaptive Sparse Attention for Efficient Long Video Generation · Around