Metraipolis › archive › dc_d55309ab347a
Improving Video Sparse Attention with Fine-grained Router and Sparse Rebasing — arXiv:2609.32882
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.32882
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T23:51:08.496Z
- SHA-256 of the archived text
225159ce5a2de9aed5f67d02eec1cd291ed5d951cf838f0511acb740f6ccbecb- Stable URL of this page
https://metraipolis.com/documento/dc_d55309ab347a
How to cite
Improving Video Sparse Attention with Fine-grained Router and Sparse Rebasing — arXiv:2609.32882 — arXiv (en), https://arxiv.org/abs/2609.32882 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d55309ab347a (Metraipolis archive dc_d55309ab347a, sha256 225159ce5a2de9aed5f67d02eec1cd291ed5d951cf838f0511acb740f6ccbecb)
BibTeX
@misc{dcd55309ab347a,
title = {Improving Video Sparse Attention with Fine-grained Router and Sparse Rebasing — arXiv:2609.32882},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.32882},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d55309ab347a; sha256 225159ce5a2de9aed5f67d02eec1cd291ed5d951cf838f0511acb740f6ccbecb},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.32882. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
We present VSA2, a frontier trainable sparse attention for video DiTs. VSA2 includes a variety of new architectural features and training procedures that we apply across all stages of the DiT development cycle, including pretraining, RL, and inference, to produce a DiT with comparable or better quality than a full attention counterpart. Architecturally, VSA2 introduces a fine-grained router that improves the precision of identifying critical tokens and supports dynamic computation by allowing each query to attend to a variable number of key-value pairs. In training, we identify a Hard-to-Easy Curriculum, where models trained under high sparsity and later evaluated with lower sparsity during inference not only generalize effectively, but also outperform models trained with full attention in motion quality. VSA2 is also flexible: it can replace full attention during the middle of progressive low-to-high resolution pretraining, rebasing early-stage full-attention checkpoints. Experiments show that VSA2 reduces attention computation by half over VSA with lower loss. On 720p videos, it accelerates attention by 8.9x and end-to-end generation by 4.62x compared to the FlashAttention-3 baseline, while achieving comparable or better video quality. Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-29 in cs.CV). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2609.32882 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-07T23:29:27.114Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).