<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[We fact-checked 41 debates between different LLMs and struck through every fabrication, without editing a word]]></title><description><![CDATA[<p dir="auto">I built H2AI Chat, an AGPL platform where several different models — from different vendors — debate a topic in turns while a human moderates. Disclosure up front: this is my project.</p>
<p dir="auto">Over the past two days we hand-verified 41 of those debates, claim by claim: 141 statements marked, 44 of them flatly false.</p>
<p dir="auto">We don't delete or correct them. The sentence stays, struck through, and you can still read it by selecting it — with the reason and the source underneath. Editing what a model said would break the only promise the site makes.</p>
<p dir="auto">Three patterns we didn't expect:</p>
<ul>
<li>Fabricated authority shows up exactly where an argument is challenged. One debate answers a budget objection with three invented citations in a single turn.</li>
<li>Fabrications spread between models. One invents a figure, a second treats it as established, a third does arithmetic on it.</li>
<li>One claim contradicts itself inside its own sentence: "62% voted Remain on a 67% turnout, meaning roughly 22% of the electorate" — which is 41.5%.</li>
</ul>
<p dir="auto">Debates: <a href="https://h2aichat.com/" rel="nofollow ugc">https://h2aichat.com/</a><br />
Code and the fact-check register: <a href="https://github.com/Tonterias/h2aichat" rel="nofollow ugc">https://github.com/Tonterias/h2aichat</a></p>


<div class="row mt-3"><div class="col-12 mt-3"><img class="img-thumbnail" src="https://lemmy.world/pictrs/image/3101c3d6-83d6-4ad2-9505-db4d69245745.png" alt="Link Preview Image" /></div></div>]]></description><link>https://forum.ieu.app/topic/65b9a91e-a9b3-45b8-b39a-7bca581624be/we-fact-checked-41-debates-between-different-llms-and-struck-through-every-fabrication-without-editing-a-word</link><generator>RSS for Node</generator><lastBuildDate>Sat, 05 Sep 2026 18:36:17 GMT</lastBuildDate><atom:link href="https://forum.ieu.app/topic/65b9a91e-a9b3-45b8-b39a-7bca581624be.rss" rel="self" type="application/rss+xml"/><pubDate>Fri, 21 Aug 2026 18:57:31 GMT</pubDate><ttl>60</ttl></channel></rss>