cs.CVNov 16, 2026

QASK-Edit: Adaptive Sparse Attention Across Queries and Key Sources for Efficient Video Editing

Authors: Hengrui Liu, Wenbo Xu, Ziqi Sheng, Wei lu

Organizations: School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou, China

Abstract

Diffusion Transformers (DiTs) have achieved strong performance in generation based video editing, but suffer from high inference latency due to the long token sequences formed by conditioning inputs and generated videos. Although training-free sparse attention reduces this cost, existing methods typically apply a fixed retention ratio or a single threshold shared across all tokens. In this work, we view video editing as conditional generation and identify a key insight: attention demand varies significantly both across different key sources (e.g., source videos, reference images, audio, text, and generated videos) and across queries within each source. Motivated by this, we propose QASK-Edit (Query-Adaptive Sparsity over Key Sources for Video Editing), a training-free sparse attention framework that allocates computation independently for each query and key source. Specifically, for each query, QASK-Edit considers both the attention mass and approximation risk of each block. It then evaluates candidate routing thresholds against each key source's risk budget, choosing exact computation or approximations to dynamically adjust sparsity for each query. Extensive experiments on three video editing models demonstrate that QASK-Edit achieves up to 1.46×1.46\times end-to-end speedup over dense attention pipeline while preserving both editing quality and unedited region fidelity. Furthermore, QASK-Edit generalizes directly to Reference-to-Video generation, delivering up to 1.36×1.36\times speedup while maintaining generation quality.

Figures & tables

Appendix figures & tables7 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

Nov 7, 2026

ESTRA: Enabling Extreme Attention Sparsity for Training-Free Video Diffusion Acceleration

Video diffusion transformers generate high-quality videos, but attention over long spatiotemporal sequences makes inference computationally expensive. Sparse attention reduces this cost by computing only a subset of query-key interactions exactly, and recent training-free methods further approximate the remaining interactions. These methods, however, degrade quickly at high sparsity, which limits the achievable speedup. We present ESTRA (Extreme Sparsity Through Refined Attention), a training-free acceleration method that preserves generation quality even at extreme sparsity. ESTRA concentrates exact attention computation on a small set of important key-value tokens and approximates the remaining contributions with cluster representatives. To identify these tokens efficiently, it groups similar queries and uses their representatives to score individual keys. A variance-based correction further refines the attention weights of the cluster representatives, improving the accuracy of the approximation. Experiments on Wan2.1-14B and HunyuanVideo-13B show that ESTRA establishes a new quality-efficiency frontier for training-free sparse attention, achieving 3.26×3.26\times and 4.71×4.71\times end-to-end speedups at 5% attention density while surpassing Sol-Attn and SVG-EAR in fidelity to dense generation and on most VBench dimensions. Even at 1% density, ESTRA remains more faithful to dense generation than Sol-Attn at 12.5%, while reaching up to 5.39×5.39\times speedup.
Oct 15, 2026

Sparse Attention for Video Generation Acceleration via Growing Sparsity and Reduced Search

Diffusion Transformers (DiTs) dominate video generation but suffer from high inference cost due to the quadratic complexity of self-attention. Sparse attention mitigates this by skipping unnecessary computation in the attention matrix, but existing methods apply a fixed sparsity ratio throughout denoising and struggle to accurately determine which computations to skip. To address these limitations, we analyze attention dynamics through attention entropy and mean attention distance, and find that attention becomes increasingly concentrated across denoising steps. We further observe that blocks selected early continue to capture a large fraction of the total attention weight at later steps. These findings suggest two acceleration opportunities: the sparsity ratio should grow as denoising progresses, and earlier search results can be reused to reduce the search space. Based on these findings, we propose Growing Sparsity and Reduced Search (GSRS), a training-free framework that retains more computation early and progressively reduces it later, while restricting subsequent searches to an early candidate set to reduce search overhead. GSRS adopts Top-p selection for adaptive sparsity control, supported by a precise evaluation scheme that computes exact attention weights within the FlashAttention computation at negligible overhead. Experiments on HunyuanVideo and Wan 2.1 validate that GSRS achieves 1.55×\times--2.21×\times speedup while maintaining generation quality comparable to full attention.
Jul 27, 2026

Sol-Attn: Accelerating Video Generation Inference via On-the-Fly Attention Sparsification

Diffusion transformers are essential for high-fidelity video generation, but long token sequences make attention a dominant inference bottleneck. Training-free dynamic sparse attention alleviates this bottleneck by computing only selected key-value blocks, yet existing methods struggle to sparsify attention both efficiently and accurately for two reasons: (1) Rigid, unpredictable, and costly routing: selecting a fixed fraction of top-ranked blocks by proxy score imposes fixed budgets, whereas retaining blocks to reach a target cumulative proxy probability mass yields dynamic but potentially imbalanced budgets; both incur non-negligible overhead from computing and materializing proxy scores. (2) Lossy keep-or-drop sparsification: unselected blocks are discarded entirely, degrading accuracy under aggressive sparsity. These limitations motivate cheaper dynamic-budget routing while limiting accuracy degradation. In this paper, we introduce training-free Sol-Attn (Sparsifying online attention), which unifies dynamic routing, sparse computation, and approximation correction in a single online-softmax pass, achieving a better accuracy-efficiency trade-off in sparse attention. The core of Sol-Attn is on-the-fly block thresholding with proxy-score reuse, which selects critical blocks by comparing block proxy scores against a threshold during online softmax. This design enables dynamic yet controllable block budgets without materializing the proxy map, while directly reusing the proxy scores of unselected blocks to approximate their contribution. Experiments across image and video generation tasks show that Sol-Attn advances the quality-efficiency frontier of training-free sparse attention, delivering 2.1 times and 2.3 times end-to-end speedups for video generation and editing, respectively, while preserving visual quality.