SVG-EAR: Parameter-Free Linear Compensation for Sparse Video Generation via Error-aware Routing

TL;DR

SVG-EAR introduces a parameter-free, error-aware linear compensation method for sparse video generation that recovers information lost due to sparse attention, significantly improving efficiency without sacrificing quality.

Contribution

The paper proposes SVG-EAR, a novel parameter-free approach that uses centroid-based approximation and error-aware routing to recover skipped attention contributions in sparse video diffusion models.

Findings

Achieves up to 1.93× speedup in video generation.

Maintains high fidelity with PSNR up to 31.043.

Establishes a Pareto frontier over prior sparse attention methods.

Abstract

Diffusion Transformers (DiTs) have become a leading backbone for video generation, yet their quadratic attention cost remains a major bottleneck. Sparse attention reduces this cost by computing only a subset of attention blocks. However, prior methods often either drop the remaining blocks, which incurs information loss, or rely on learned predictors to approximate them, introducing training overhead and potential output distribution shifting. In this paper, we show that the missing contributions can be recovered without training: after semantic clustering, keys and values within each block exhibit strong similarity and can be well summarized by a small set of cluster centroids. Based on this observation, we introduce SVG-EAR, a parameter-free linear compensation branch that uses the centroid to approximate skipped blocks and recover their contributions. While centroid compensation is accurate for most blocks, it can fail on a small subset. Standard sparsification typically selects blocks by attention scores, which indicate where the model places its attention mass, but not where the approximation error would be largest. SVG-EAR therefore performs error-aware routing: a lightweight probe estimates the compensation error for each block, and we compute exactly the blocks with the highest error-to-cost ratio while compensating for skipped blocks. We provide theoretical guarantees that relate attention reconstruction error to clustering quality, and empirically show that SVG-EAR improves the quality-efficiency trade-off and increases throughput at the same generation fidelity on video diffusion tasks. Overall, SVG-EAR establishes a clear Pareto frontier over prior approaches, achieving up to 1.77$\times$ and 1.93$\times$ speedups while maintaining PSNRs of up to 29.759 and 31.043 on Wan2.2 and HunyuanVideo, respectively.