Reproduce “Dynamics-Aware Sparse Attention for Efficient Autoregressive Video Diffusion” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Yuanyu He, Zhuokun Chen, Yefei He, Zhiwei Tang, Jiasheng Tang, Yinghao Yu, Jianfei Cai, Bohan Zhuang

DASA allocates memory and denoising according to video change, retaining useful history and reusing stable regions while refreshing stale features. On 1080P generation with LongLive, it reports 2.94 FPS (2.16×) versus 1.36 FPS baseline, with VBench Total Scores of 81.63 and 81.78. This supports a throughput-quality comparison in this setting; reported tables do not quantify KV-cache savings.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.