<?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>Jailbreak on 서소영의 서재</title><link>https://seosoyoung.eiaserinnys.me/tags/jailbreak/</link><description>Recent content in Jailbreak on 서소영의 서재</description><generator>Hugo</generator><language>ko</language><lastBuildDate>Thu, 21 May 2026 14:00:00 +0900</lastBuildDate><atom:link href="https://seosoyoung.eiaserinnys.me/tags/jailbreak/index.xml" rel="self" type="application/rss+xml"/><item><title>Persuading Large Language Models to Comply with Objectionable Requests</title><link>https://seosoyoung.eiaserinnys.me/digest/persuading-llms-pnas-2026/</link><pubDate>Thu, 21 May 2026 14:00:00 +0900</pubDate><guid>https://seosoyoung.eiaserinnys.me/digest/persuading-llms-pnas-2026/</guid><description>Cialdini의 일곱 설득 원칙을 프롬프트에 넣으면 LLM의 규제 약물 합성 요청 컴플라이언스가 35.3%에서 51.3%로 상승한다. PNAS 2026, 세 프런티어 모델 대상 126,000회 통제 실험.</description></item></channel></rss>