Metraipolis › archive › dc_1702ad2b6816

Sparse Attention Is Matrix Approximation, Not Choosing from a Bag of Values — arXiv:2610.10871

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.10871
License (as archived)
CC0-1.0
Archived
2026-10-09T06:04:02.613Z
SHA-256 of the archived text
49c6724b017624461cc000b2bafdaecee92c14383e831e2e12f3d7831c62e81a
Stable URL of this page
https://metraipolis.com/documento/dc_1702ad2b6816

How to cite

Sparse Attention Is Matrix Approximation, Not Choosing from a Bag of Values — arXiv:2610.10871 — arXiv (en), https://arxiv.org/abs/2610.10871 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_1702ad2b6816 (Metraipolis archive dc_1702ad2b6816, sha256 49c6724b017624461cc000b2bafdaecee92c14383e831e2e12f3d7831c62e81a)
BibTeX
@misc{dc1702ad2b6816,
  title = {Sparse Attention Is Matrix Approximation, Not Choosing from a Bag of Values — arXiv:2610.10871},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.10871},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_1702ad2b6816; sha256 49c6724b017624461cc000b2bafdaecee92c14383e831e2e12f3d7831c62e81a},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.10871. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large Language Models (LLMs) achieve strong performance across many domains, but their efficiency is limited by the quadratic cost of attention with respect to prompt length. Sparse attention reduces this cost by retaining only a small fraction of query-key interactions to approximate the full attention matrix. However, existing methods are trapped in a mathematically wrong view: they simply keep large scalar entries or high-mass regions of the attention matrix. This treats the attention matrix as a bag of values, ignoring that it is used as a structured matrix whose entries jointly determine the attention output through multiplication with value vectors. We argue that this is the core conceptual issue: sparse attention should be formulated as matrix approximation, not as blindly choosing the largest values from a bag of entries. Based on this view, we propose Matrix Approximation Sparse Attention (MASA). MASA replaces raw attention-mass ranking with a closed-form score that measures how much each sparse unit reduces matrix-product approximation error. As a theory-grounded plug-in correction, MASA can be added to existing sparse attention frameworks without changing their sparse kernels or budgets. Extensive experiments across multiple sparse attention methods, benchmarks, and LLM backbones show consistent accuracy gains, supporting both MASA and the matrix-approximation view of sparse attention.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.10871 — 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 2026-10-09T05:47:15.099Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).