<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Learning with Latent Language · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/1711.00482/</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/1711.00482/feed.xml" rel="self" type="application/rss+xml" /><item><title>Learning with Latent Language — mentioned by gsjbjt</title><link>https://news.ycombinator.com/item?id=29601783</link><guid isPermaLink="false">urn:hackerpapers:mention:1711.00482:29601783</guid><pubDate>Sat, 18 Dec 2021 04:54:51 GMT</pubDate><description>&lt;p&gt;Nice post! I work on NLP and I think a lot of ideas in this post resonate with what I find exciting about working on the intersection of language + the real world: large text datasets as sources of abundant prior knowledge about the world, structure of language ~ structure of concepts that matter to humans, etc. I feel like the bottleneck is getting access to paired (language, other modality) data though (if your other modality isn&amp;#x27;t images). i.e. &amp;quot;bolt on generalization&amp;quot; is an intuitively appealing concept, but then it reduces to the hard problem of &amp;quot;how do I learn to ground language to e.g.…&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/1711.00482/"&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>
