Paper Detail

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

Hao Liu, Chenghuan Huang, Ye Huang, Zhiying Wen, Mohan Zhang, Chen Li, Ziyang Ma, Jing Lyu, Jiangsu Du

huggingface Score 5.4

Published 2026-07-17 · First seen 2026-07-23

General AI

Abstract

Video Diffusion Transformers process long spatio-temporal sequences, making self-attention the main bottleneck in high-resolution video generation. Training-free sparse attention reduces this cost, but adaptive Top-p routing creates uneven per-head workloads under multi-GPU sequence parallelism. The resulting workload heterogeneity turns sparse attention into a rank-level straggler problem. We present , a training-free sparse-attention system that improves the distributed execution efficiency of adaptive sparse attention under multi-GPU sequence parallelism. uses Top-p routing, a Top-k safety floor, and video-aware block organization as the sparse-routing frontend, then repairs the materialized mask at runtime. Runtime Load Balancing migrates a small number of heavy heads via P2P communication to shorten the current critical path. Slack-Aware Sparse Augmentation fills residual non-critical-rank slack with additional high-value blocks, while overlap hides scheduling and migration overhead behind existing computation. On step-distilled Wan2.2 I2V, reduces average load imbalance from 1.34 to 1.08 and delivers a 4.41times attention speedup over FlashAttention, while achieving a 2.02--2.11times DiT inference speedup with competitive video quality.

Workflow Status

Review status
pending
Role
unreviewed
Read priority
later
Vote
Not set.
Saved
no
Collections
Not filed yet.
Next action
Not filled yet.

Reading Brief

No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.

Why It Surfaced

No ranking explanation is available yet.

Tags

No tags.

BibTeX

@misc{liu2026fvattn,
  title = {FVAttn: Adaptive Sparse Attention with Runtime Load Balancing for Video Generation},
  author = {Hao Liu and Chenghuan Huang and Ye Huang and Zhiying Wen and Mohan Zhang and Chen Li and Ziyang Ma and Jing Lyu and Jiangsu Du},
  year = {2026},
  abstract = {Video Diffusion Transformers process long spatio-temporal sequences, making self-attention the main bottleneck in high-resolution video generation. Training-free sparse attention reduces this cost, but adaptive Top-p routing creates uneven per-head workloads under multi-GPU sequence parallelism. The resulting workload heterogeneity turns sparse attention into a rank-level straggler problem. We present , a training-free sparse-attention system that improves the distributed execution efficiency of},
  url = {https://huggingface.co/papers/2607.16190},
  keywords = {huggingface daily},
  eprint = {2607.16190},
  archiveprefix = {arXiv},
}

Metadata

{}