<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Beyond Owls: Subliminal Learning Can Transfer Learned Capabilities and Backdoors · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/2610.10657/</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/2610.10657/feed.xml" rel="self" type="application/rss+xml" /><item><title>Beyond Owls: Subliminal Learning Can Transfer Learned Capabilities and Backdoors — mentioned by khafra</title><link>https://news.ycombinator.com/item?id=50163957</link><guid isPermaLink="false">urn:hackerpapers:mention:2610.10657:50163957</guid><pubDate>Tue, 20 Oct 2026 10:25:40 GMT</pubDate><description>&lt;p&gt;Unfortunately, frontier models are all trained in part via distillation, and distillation can produce sleeper agents without ever mentioning the trigger phrase or the behavior (https://arxiv.org/html/2610.10657v1).&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/2610.10657/"&gt;Paper and discussion history&lt;/a&gt;. A mention is not an endorsement.&lt;/p&gt;</description><category>cs.LG</category></item></channel></rss>
