<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>From Language Models over Tokens to Language Models over Characters · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/2412.03719/</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/2412.03719/feed.xml" rel="self" type="application/rss+xml" /><item><title>From Language Models over Tokens to Language Models over Characters — mentioned by kcarnold</title><link>https://news.ycombinator.com/item?id=42667503</link><guid isPermaLink="false">urn:hackerpapers:mention:2412.03719:42667503</guid><pubDate>Sat, 11 Jan 2025 17:45:10 GMT</pubDate><description>&lt;p&gt;This was the subject of https://arxiv.org/abs/2412.03719. (I suspect you can do simpler than the paper&amp;#x27;s solution if you&amp;#x27;re only interested in the top-k.) A related topic is &amp;quot;token healing&amp;quot;, although some implementations (unfortunately including the one in HuggingFace Transformers) make some big assumptions that aren&amp;#x27;t always true (like treating spaces as special).&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/2412.03719/"&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>
