<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Safety Evaluation | Samuele Poppi</title><link>https://seppia978.github.io/tags/safety-evaluation/</link><atom:link href="https://seppia978.github.io/tags/safety-evaluation/index.xml" rel="self" type="application/rss+xml"/><description>Safety Evaluation</description><generator>Hugo Blox Builder (https://hugoblox.com)</generator><language>en-us</language><lastBuildDate>Tue, 09 Jun 2026 00:00:00 +0000</lastBuildDate><image><url>https://seppia978.github.io/media/icon_hu14756676106600458814.png</url><title>Safety Evaluation</title><link>https://seppia978.github.io/tags/safety-evaluation/</link></image><item><title>When the Chain of Thought Knows Better: Failure Modes in Multi-Turn Reasoning Models</title><link>https://seppia978.github.io/publication/2606-arxiv-cot_knows_better/</link><pubDate>Tue, 09 Jun 2026 00:00:00 +0000</pubDate><guid>https://seppia978.github.io/publication/2606-arxiv-cot_knows_better/</guid><description/></item><item><title>Not All Attackers Are Malicious: When Safety Degrades Without Harmful Intent</title><link>https://seppia978.github.io/event/mbzuai_sec_symposium_2602/</link><pubDate>Mon, 09 Feb 2026 00:00:00 +0000</pubDate><guid>https://seppia978.github.io/event/mbzuai_sec_symposium_2602/</guid><description>&lt;p>Slides from the talk are available below.&lt;/p></description></item></channel></rss>