HyperVAttention：基于时空聚类的高效稀疏注意力视频扩散方法
HyperVAttention: Efficient Sparse Attention with Spatio-Temporal Clustering for Video Diffusion
AI 导读
HyperVAttention（HVA）是一种免训练的视频扩散稀疏注意力框架，通过 3D 局部窗口聚类、混合聚类策略和硬件感知聚类合并，降低聚类开销并提升 GPU 利用率。在 HunyuanVideo 上实现比全注意力快 2.13 倍的速度，HVA-Flash 变体达 2.23 倍，且在相近延迟下保真度超过 SVG2。该工作已被 NeurIPS 2026 接收。
正文 · 原文
Abstract:Video Diffusion Transformers (VDiTs) have demonstrated significant capabilities in high-fidelity video generation. However, their ability to produce long-duration videos is fundamentally constrained by the quadratic complexity of the self-attention mechanism. Recent clustering-based sparse attention methods improve the quality-speed trade-off by grouping semantically similar tokens, but their practical efficiency remains limited by two bottlenecks: substantial clustering overhead and low CTA utilization caused by irregular cluster-induced blocks. We propose HyperVAttention (HVA), a training-free sparse attention framework that addresses both bottlenecks jointly. To reduce clustering overhead, we introduce 3D local-window clustering, which exploits the spatio-temporal locality of video tokens to restrict centroid search to fixed local neighborhoods, and implement it with a custom Triton kernel for efficient execution. We further propose a hybrid clustering strategy that performs full clustering only at anchor steps and updates only subset tokens at intermediate steps, leveraging the temporal stability of cluster assignments across denoising steps. To improve CTA utilization, we present hardware-aware cluster merging that minimizes CTA-aligned execution cost through parallel agglomerative merging, improving block density and approximation fidelity by utilizing idle tile capacity. Together, these components reduce clustering overhead, avoid redundant updates, and better align sparse attention with the fixed tile structure of modern GPU kernels. Experiments on video generation show that HVA advances the Pareto frontier of training free sparse attention, achieving higher fidelity than SVG2 at comparable latency and a $2.13\times$ speedup over full attention on HunyuanVideo, or $2.23\times$ with its faster variant HVA-Flash.
| Comments: | Accepted by NeurIPS 2026. Project Page: this https URL |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2607.03012 [cs.CV] |
| (or arXiv:2607.03012v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2607.03012 arXiv-issued DOI via DataCite |
Submission history
From: Dongyeun Lee [view email]
[v1]
Fri, 3 Jul 2026 06:46:35 UTC (3,889 KB)
[v2]
Thu, 5 Nov 2026 14:37:08 UTC (3,891 KB)
来源：arXiv cs.AI · arxiv.org