<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Transformers Can Do Arithmetic with the Right Embeddings · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/2405.17399/</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/2405.17399/feed.xml" rel="self" type="application/rss+xml" /><item><title>Transformers Can Do Arithmetic with the Right Embeddings — mentioned by riku_iki</title><link>https://news.ycombinator.com/item?id=41703362</link><guid isPermaLink="false">urn:hackerpapers:mention:2405.17399:41703362</guid><pubDate>Tue, 01 Oct 2024 00:03:04 GMT</pubDate><description>&lt;p&gt;&amp;gt; https://arxiv.org/abs/2405.17399 they built specialized model which is after bunch of trickery still has 99% accuracy(naive model had very low accuracy) on very simple deterministic algo. I also think most of the accuracy came from memorization of training set(model didn&amp;#x27;t provide intermediate results, and started failing significantly at sligtly larger input). In my book it is fundamental inability to learn and reproduce algorithm. They also demonstrated that transformer can&amp;#x27;t learn sorting.&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/2405.17399/"&gt;Paper and discussion history&lt;/a&gt;. A mention is not an endorsement.&lt;/p&gt;</description><category>cs.LG</category></item><item><title>Transformers Can Do Arithmetic with the Right Embeddings — mentioned by famouswaffles</title><link>https://news.ycombinator.com/item?id=41703300</link><guid isPermaLink="false">urn:hackerpapers:mention:2405.17399:41703300</guid><pubDate>Mon, 30 Sep 2024 23:47:38 GMT</pubDate><description>&lt;p&gt;&amp;gt; there are some benchmarks which show fundamental inability of LLM perform certain tasks which human can, for example add 100 digits numbers. fundamental inability ? No. Current Sota LLM (4o, claude, gemini) woes with arithmetic is not a transformer weakness never mind a large language modelling one. Those benchmarks show that those particular models have problems with accuracy on that many digits, not that LLMs fundamentally cannot be accurate on that many digits. https://arxiv.org/abs/2405.17399 https://arxiv.org/abs/2307.03381 https://arxiv.org/abs/2310.02989 Ultimately, numbers are…&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/2405.17399/"&gt;Paper and discussion history&lt;/a&gt;. A mention is not an endorsement.&lt;/p&gt;</description><category>cs.LG</category></item><item><title>Transformers Can Do Arithmetic with the Right Embeddings — HN submission</title><link>https://news.ycombinator.com/item?id=40502881</link><guid isPermaLink="false">urn:hackerpapers:submission:2405.17399:40502881</guid><pubDate>Tue, 28 May 2024 17:03:20 GMT</pubDate><description>&lt;p&gt;Transformers Can Do Arithmetic with the Right Embeddings&lt;/p&gt;&lt;p&gt;&lt;/p&gt;&lt;p&gt;1 point; 0 discussion replies on HN. &lt;a href="https://hackerpaper.kaipfister.de/p/2405.17399/"&gt;Paper and source history&lt;/a&gt;. A submission is not an endorsement.&lt;/p&gt;</description><category>cs.LG</category></item><item><title>Transformers Can Do Arithmetic with the Right Embeddings — HN submission</title><link>https://news.ycombinator.com/item?id=40497379</link><guid isPermaLink="false">urn:hackerpapers:submission:2405.17399:40497379</guid><pubDate>Tue, 28 May 2024 04:39:39 GMT</pubDate><description>&lt;p&gt;Transformers Can Do Arithmetic with the Right Embeddings&lt;/p&gt;&lt;p&gt;&lt;/p&gt;&lt;p&gt;207 points; 211 discussion replies on HN. &lt;a href="https://hackerpaper.kaipfister.de/p/2405.17399/"&gt;Paper and source history&lt;/a&gt;. A submission is not an endorsement.&lt;/p&gt;</description><category>cs.LG</category></item></channel></rss>
