Saved in:
Bibliographic Details
Main Authors: Liberty, Edo, Andoni, Alexandr, Kleiner, Eldar
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.05602
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913097362440192
author Liberty, Edo
Andoni, Alexandr
Kleiner, Eldar
author_facet Liberty, Edo
Andoni, Alexandr
Kleiner, Eldar
contents We consider the problem of estimating the Attention mechanism in small space, and prove the existence of coresets for it of nearly optimal size. Specifically, we show that for any set of unit-norm keys and values $(K,V)$ in $\mathbb{R}^d$, there exists a subset $(K',V')$ of size at most $O({\sqrt{d} e^{ρ+o(ρ)}/\varepsilon})$ such that \[ \left\| \operatorname{Attn}(q,K,V)- \operatorname{Attn}(q,K',V') \right\| \le \varepsilon \] simultaneously for all queries whose norm is bounded by $ρ$. This outperforms the best known results for this problem. We also offer an improved lower bound showing that $\varepsilon$-coresets must have size $Ω({\sqrt{d} e^ρ/ε})$.
format Preprint
id arxiv_https___arxiv_org_abs_2605_05602
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Nearly Optimal Attention Coresets
Liberty, Edo
Andoni, Alexandr
Kleiner, Eldar
Data Structures and Algorithms
Artificial Intelligence
We consider the problem of estimating the Attention mechanism in small space, and prove the existence of coresets for it of nearly optimal size. Specifically, we show that for any set of unit-norm keys and values $(K,V)$ in $\mathbb{R}^d$, there exists a subset $(K',V')$ of size at most $O({\sqrt{d} e^{ρ+o(ρ)}/\varepsilon})$ such that \[ \left\| \operatorname{Attn}(q,K,V)- \operatorname{Attn}(q,K',V') \right\| \le \varepsilon \] simultaneously for all queries whose norm is bounded by $ρ$. This outperforms the best known results for this problem. We also offer an improved lower bound showing that $\varepsilon$-coresets must have size $Ω({\sqrt{d} e^ρ/ε})$.
title Nearly Optimal Attention Coresets
topic Data Structures and Algorithms
Artificial Intelligence
url https://arxiv.org/abs/2605.05602