<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Universal Zero-shot Embedding Inversion · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/2504.00147/</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/2504.00147/feed.xml" rel="self" type="application/rss+xml" /><item><title>Universal Zero-shot Embedding Inversion — mentioned by luke-stanley</title><link>https://news.ycombinator.com/item?id=43965837</link><guid isPermaLink="false">urn:hackerpapers:mention:2504.00147:43965837</guid><pubDate>Mon, 12 May 2025 18:01:16 GMT</pubDate><description>&lt;p&gt;&amp;quot;starting from an embedding and working backwards to generate a piece of text that is semantically close by?&amp;quot; Apparently this is called embedding inversion and Universal Zero-shot Embedding Inversion https://arxiv.org/abs/2504.00147 Going incrementally closer and closer to the target with some means to vary seems to be the most general way, there are lots of ways to be more optimal though. Image diffusion with CLIP embeddings and such is kinda related too.&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/2504.00147/"&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>
