<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Mllm on The Engine Room</title><link>/en/tags/mllm/</link><description>Recent content in Mllm on The Engine Room</description><generator>Hugo -- 0.155.3</generator><language>en-us</language><lastBuildDate>Sat, 22 Nov 2025 15:00:00 +0300</lastBuildDate><atom:link href="/en/tags/mllm/index.xml" rel="self" type="application/rss+xml"/><item><title>Defending MLLMs from Implicit Jailbreak Attacks</title><link>/en/notes/defence_mllm_from_jailbreak/</link><pubDate>Sat, 22 Nov 2025 15:00:00 +0300</pubDate><guid>/en/notes/defence_mllm_from_jailbreak/</guid><description>A new class of attacks where text and image look safe separately, but their combination carries malicious meaning</description></item></channel></rss>