# Attention Sparsity is Input-Stable: Training-Free Sparse Attention for Video Generation via Offline Sparsity Profiling and Online QK Co-Clustering

arXiv: https://arxiv.org/abs/2603.18636 · HTML: https://mlsanity.com/paper/2603.18636/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- Placed in Video generation, one of 14 directions in Diffusion models.
- First posted on arXiv 19 Mar 2026.

## Placed in

- Diffusion models / [Video generation](https://mlsanity.com/diffusion/video-generation/)

## Built on by (most cited first, 10 of 10)

- [Ride the Wave: Precision-Allocated Sparse Attention for Smooth Video Generation](https://mlsanity.com/paper/2604.12219/) (2026)
- [EcoVideo: Entropy-Orchestrated Video Generation Paradigm in Cloud-Edge Dynamics](https://mlsanity.com/paper/2606.30557/) (2026)
- [Partition the Support, Reconstruct the Residual: Training-Free Sparse Attention for Video Generation and World Models](https://mlsanity.com/paper/2608.18484/) (2026)
- [Token Radius Attention for Efficient Video Generation](https://mlsanity.com/paper/2608.02504/) (2026)
- [SCOPE: Subspace Clustering with Online Per-Head Top-K Estimation for Sparse Video Attention](https://mlsanity.com/paper/2608.12780/) (2026)
- [ClusterAttention: A training-free speedup of bidirectional attention](https://mlsanity.com/paper/2608.26965/) (2026)
- [DraftAttention2: Fast Video Diffusion with Low-Resolution-Guided Mixed-Precision Attention](https://mlsanity.com/paper/2609.32628/) (2026)
- [PARK: Accurate Block Retrieval for Sparse Attention in Video Diffusion Transformers](https://mlsanity.com/paper/2609.38978/) (2026)
- [MC-Sparse: Deconstructing and Closing the Dense-Sparse Attention Gap in Diffusion Transformers](https://mlsanity.com/paper/2610.06801/) (2026)
- [iCATS: Fast Video Generation via Interaction-Aware Sparse Attention and Timestep-Adaptive Sparsity](https://mlsanity.com/paper/2610.11302/) (2026)

## Abstract (arXiv)

Diffusion Transformers (DiTs) achieve strong video generation quality but suffer from high inference cost due to dense 3D attention, motivating sparse attention techniques for improving efficiency. However, existing training-free sparse attention methods for video generation still face two unresolved limitations: ignoring layer heterogeneity in attention pruning and ignoring query-key coupling in block partitioning, which hinder a better quality-speedup trade-off. In this work, we uncover a critical insight: attention sparsity is an intrinsic layer-wise property, with only minor variation across different inputs. Motivated by this observation, we propose SVOO, a training-free sparse attention framework for fast video generation via offline layer-wise sparsity profiling and online bidirectional co-clustering. Specifically, SVOO adopts a two-stage paradigm: (i) offline layer-wise sensitivity profiling to derive intrinsic per-layer pruning levels, and (ii) online block-wise sparse attention via a bidirectional co-clustering algorithm. Extensive experiments on seven widely used video generation models demonstrate that SVOO achieves a superior quality-speedup trade-off over state-of-the-art methods, delivering up to 1.93x speedup while maintaining a PSNR of up to 29 dB on Wan2.1. Code is available at: https://github.com/Mutual-Luo/SVOO.