README.md

metadata

title: README
emoji: ⚡
colorFrom: purple
colorTo: blue
sdk: static
pinned: false

Veda: Scalable Video Diffusion via Distilled Sparse Attention

ICML 2026 · Video Diffusion · Sparse Attention

📄 Paper · 🌐 Project Page · 💻 GitHub · 📚 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.