<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>At Which Training Stage Does Code Data Help LLMs Reasoning? · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/2309.16298/</link><description>Collected HN comments and submissions linking this paper.</description><language>en</language><lastBuildDate>Sat, 07 Nov 2026 05:36:49 GMT</lastBuildDate><atom:link href="https://hackerpaper.kaipfister.de/p/2309.16298/feed.xml" rel="self" type="application/rss+xml" /><item><title>At Which Training Stage Does Code Data Help LLMs Reasoning? — mentioned by reasonableklout</title><link>https://news.ycombinator.com/item?id=50400599</link><guid isPermaLink="false">urn:hackerpapers:mention:2309.16298:50400599</guid><pubDate>Sat, 07 Nov 2026 01:05:46 GMT</pubDate><description>&lt;p&gt;This isn&amp;#x27;t really comforting to me for 2 reasons: 1. A key insight of foundation models is that capabilities are correlated. For example, training on code improved general reasoning capabilities: https://arxiv.org/abs/2309.16298 2. A profound capabilities explosion in easily measured domains will still feel very different to today. Like, the moment we come up with a systematic, objective way to verify a task, we become obsolete at it?&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/2309.16298/"&gt;Paper and discussion history&lt;/a&gt;. A mention is not an endorsement.&lt;/p&gt;</description><category>cs.CL</category></item></channel></rss>
