Veda

Veda: Scalable Video Diffusion via Distilled Sparse Attention

ICML 2026 路 Video Diffusion 路 Sparse Attention

馃搫 Paper 路 馃寪 Project Page 路 馃捇 GitHub 路 馃摎 BibTeX

Veda is a learned sparse-attention method for video DiTs. A lightweight tile-score predictor, distilled from full attention, keeps only the attention tiles that matter and skips the rest with a hardware-aligned tile-sparse kernel.