Video diffusion transformers (DiTs) suffer from prohibitive inference latency due to quadratic attention complexity. Existing sparse attention methods either overlook semantic similarity or fail to adapt to heterogeneous token distributions across layers, leading to model performance degradation. We propose AdaCluster, a training-free adaptive clustering framework that accelerates the generation of DiTs while preserving accuracy. AdaCluster applies an angle-similarity-preserving clustering method to query vectors for higher compression, and designs a euclidean-similarity-preserving clustering method for keys, covering cluster number assignment, threshold-wise adaptive clustering, and efficient critical cluster selection. Experiments on CogVideoX-2B, HunyuanVideo, and Wan-2.1 on one A40 GPU demonstrate up to 1.67-4.31x speedup with negligible quality degradation.
@article{arxiv.2604.18348,
title = {AdaCluster: Adaptive Query-Key Clustering for Sparse Attention in Video Generation},
author = {Haoyue Tan and Shengnan Wang and Yulin Qiao and Juncheng Zhang and Youhui Bai and Ping Gong and Zewen Jin and Cheng Li},
journal= {arXiv preprint arXiv:2604.18348},
year = {2026}
}