<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Understanding the Vulnerability of CLIP to Image Compression · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/2311.14029/</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/2311.14029/feed.xml" rel="self" type="application/rss+xml" /><item><title>Understanding the Vulnerability of CLIP to Image Compression — mentioned by heyitsguay</title><link>https://news.ycombinator.com/item?id=38449307</link><guid isPermaLink="false">urn:hackerpapers:mention:2311.14029:38449307</guid><pubDate>Tue, 28 Nov 2023 18:31:36 GMT</pubDate><description>&lt;p&gt;Kind of an odd blog post -- is this AI generated? Or maybe written by someone who isn&amp;#x27;t a native English speaker? The writing feels off. In any case, maybe worth linking directly to the paper, it&amp;#x27;s an easy read and more informative: https://arxiv.org/abs/2311.14029 The tl;dr is they tested CLIP variants for zero-shot classification on CIFAR-10 and STL-10 at different levels of image compression and observed how performance dropped with more compression. They claim the purpose is more about demonstrating performance change attribution using this Integrated Gradients method but imo it doesn&amp;#x27;t…&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/2311.14029/"&gt;Paper and discussion history&lt;/a&gt;. A mention is not an endorsement.&lt;/p&gt;</description><category>cs.CV</category></item></channel></rss>
