<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Do LLMs Know about Hallucination? An Empirical Investigation of LLM's Hidden States · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/2402.09733/</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/2402.09733/feed.xml" rel="self" type="application/rss+xml" /><item><title>Do LLMs Know about Hallucination? An Empirical Investigation of LLM's Hidden States — mentioned by cornel_io</title><link>https://news.ycombinator.com/item?id=43420049</link><guid isPermaLink="false">urn:hackerpapers:mention:2402.09733:43420049</guid><pubDate>Thu, 20 Mar 2025 05:09:37 GMT</pubDate><description>&lt;p&gt;There are various results that suggest that LLMs do internally have everything they&amp;#x27;d need to know that they&amp;#x27;re hallucinating/wrong: https://arxiv.org/abs/2402.09733 https://arxiv.org/abs/2305.18248 https://www.ox.ac.uk/news/2024-06-20-major-research-hallucin... So I don&amp;#x27;t think it&amp;#x27;s that they have no concept of correctness, they do, but it&amp;#x27;s not strong enough. We&amp;#x27;re probably just not training them in ways that optimize for that over other desirable qualities, at least aggressively enough. It&amp;#x27;s also clear to anyone who has used many different models over the years that the amount of…&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/2402.09733/"&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>
