README.md

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

<div align="center">

<img src="https://veda-sparse.github.io/logo.svg" alt="Veda" width="96"/>

# Veda: Scalable Video Diffusion via Distilled Sparse Attention

**ICML 2026** · Video Diffusion · Sparse Attention

[📄 Paper](https://arxiv.org/abs/2605.30325) · [🌐 Project Page](https://veda-sparse.github.io/) · [💻 GitHub](https://github.com/veda-sparse) · [📚 BibTeX](https://veda-sparse.github.io/#bibtex)

</div>

**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.