Lightning Unified Video Editing via In-Context Sparse Attention

Hugging Face Daily Papers Papers
video-editing sparse-attention in-context-learning efficiency generative-ai computer-vision

Summary

This paper introduces In-context Sparse Attention (ISA), a framework that significantly reduces computational costs in video editing by pruning redundant context and using dynamic query grouping. The authors demonstrate the method's effectiveness with LIVEditor, achieving near-lossless acceleration and state-of-the-art results on multiple video editing benchmarks.

Video editing has evolved toward In-Context Learning (ICL) paradigms, yet the resulting quadratic attention costs create a critical computational bottleneck. In this work, we propose In-context Sparse Attention (ISA), the first near-lossless empirical sparse framework tailored for ICL video editing. Our design is grounded in two key insights: first, context tokens exhibit significantly lower saliency than source tokens; second, we theoretically prove and empirically validate that Query sharpness correlates with approximation error. Motivated by these findings, ISA implements an efficient pre-selection strategy to prune redundant context, followed by a dynamic query grouping mechanism that routes high-error queries to full attention and low-error ones to a computationally efficient 0-th order Taylor sparse attention. Furthermore, we build \texttt{LIVEditor} , a novel lightning video editing model via ISA and a proposed video-editing data pipeline that curated a 1.7M high-quality dataset. Extensive experiments demonstrate that LIVEditor achieves a sim60% reduction in attention-module latency while surpassing state-of-the-art methods across EditVerseBench, IVE-Bench, and VIE-Bench, delivering near-lossless acceleration without compromising visual fidelity.
Original Article
View Cached Full Text

Cached at: 05/08/26, 08:07 AM

Paper page - Lightning Unified Video Editing via In-Context Sparse Attention

Source: https://huggingface.co/papers/2605.04569

Abstract

In-context sparse attention framework enables efficient video editing with reduced computational costs while maintaining visual quality.

Video editinghas evolved towardIn-Context Learning(ICL) paradigms, yet the resulting quadratic attention costs create a critical computational bottleneck. In this work, we propose In-contextSparse Attention(ISA), the first near-lossless empirical sparse framework tailored for ICLvideo editing. Our design is grounded in two key insights: first, context tokens exhibit significantly lower saliency than source tokens; second, we theoretically prove and empirically validate thatQuery sharpnesscorrelates with approximation error. Motivated by these findings, ISA implements an efficient pre-selection strategy to prune redundant context, followed by a dynamic query grouping mechanism that routes high-error queries to full attention and low-error ones to a computationally efficient0-th order Taylor sparse attention. Furthermore, we build \texttt{LIVEditor} , a novel lightningvideo editingmodel via ISA and a proposed video-editing data pipeline that curated a 1.7M high-quality dataset. Extensive experiments demonstrate that LIVEditor achieves a sim60% reduction inattention-module latencywhile surpassing state-of-the-art methods acrossEditVerseBench,IVE-Bench, andVIE-Bench, delivering near-lossless acceleration without compromising visual fidelity.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2605\.04569

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.04569 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.04569 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.04569 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Faster Video Diffusion with Trainable Sparse Attention

Papers with Code Trending

This paper introduces Trainable Sparse Attention (VSA), a hardware-efficient sparse attention mechanism that reduces computational costs in video diffusion transformers without compromising performance, enabling more efficient scaling and faster generation.

FVAttn: Adaptive Sparse Attention with Runtime Load Balancing for Video Generation

Hugging Face Daily Papers

FVAttn is a training-free sparse attention system that uses runtime load balancing to improve distributed execution efficiency of adaptive sparse attention under multi-GPU sequence parallelism for video generation, achieving up to 4.41x attention speedup and 2.11x inference speedup over FlashAttention on step-distilled Wan2.2 I2V while maintaining competitive video quality.