<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Fast Attention Requires Bounded Entries · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/2302.13214/</link><description>Collected HN comments and submissions linking this paper.</description><language>en</language><lastBuildDate>Mon, 02 Nov 2026 11:16:22 GMT</lastBuildDate><atom:link href="https://hackerpaper.kaipfister.de/p/2302.13214/feed.xml" rel="self" type="application/rss+xml" /><item><title>Fast Attention Requires Bounded Entries — mentioned by thomasahle</title><link>https://news.ycombinator.com/item?id=46887069</link><guid isPermaLink="false">urn:hackerpapers:mention:2302.13214:46887069</guid><pubDate>Wed, 04 Feb 2026 15:33:26 GMT</pubDate><description>&lt;p&gt;There&amp;#x27;s a graveyard of 100s of papers with &amp;quot;approximate near linear time attention.&amp;quot; They always hope the speed increase makes up for the lower quality, but it never does. The quadratic time seems inherent to the problem. Indeed, there are lower bounds showing that sub n^2 algorithms can&amp;#x27;t work: https://arxiv.org/pdf/2302.13214&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/2302.13214/"&gt;Paper and discussion history&lt;/a&gt;. A mention is not an endorsement.&lt;/p&gt;</description><category>cs.LG</category></item></channel></rss>
