<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>What is New in Stochastic Rounding: a Survey on Theory, Hardware, and Applications · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/2603.06060/</link><description>Collected HN comments and submissions linking this paper.</description><language>en</language><lastBuildDate>Thu, 05 Nov 2026 05:34:29 GMT</lastBuildDate><atom:link href="https://hackerpaper.kaipfister.de/p/2603.06060/feed.xml" rel="self" type="application/rss+xml" /><item><title>What is New in Stochastic Rounding: a Survey on Theory, Hardware, and Applications — mentioned by martincsweiss</title><link>https://news.ycombinator.com/item?id=50374271</link><guid isPermaLink="false">urn:hackerpapers:mention:2603.06060:50374271</guid><pubDate>Thu, 05 Nov 2026 02:17:10 GMT</pubDate><description>&lt;p&gt;Yep, Stochastic Rounding [1]. I actually had &amp;quot;round-to-nearest&amp;quot; in the title of the piece originally. Usually people do that if they don&amp;#x27;t have the space for fp32 master weights, but it doesn&amp;#x27;t fix the second moment part of the learning process. [1] What is New in Stochastic Rounding: a Survey on Theory, Hardware, and Applications, https://arxiv.org/pdf/2603.06060v2&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/2603.06060/"&gt;Paper and discussion history&lt;/a&gt;. A mention is not an endorsement.&lt;/p&gt;</description><category>math.NA</category></item></channel></rss>
