<?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>ExploitBench on TechLLM</title><link>https://techllm.github.io/tags/exploitbench/</link><description>Recent content in ExploitBench on TechLLM</description><generator>Hugo</generator><language>ko</language><lastBuildDate>Sun, 17 May 2026 11:27:00 +0900</lastBuildDate><atom:link href="https://techllm.github.io/tags/exploitbench/index.xml" rel="self" type="application/rss+xml"/><item><title>AI가 진짜로 브라우저 해킹을?! — ExploitBench 벤치마크 이야기</title><link>https://techllm.github.io/posts/exploitbench-ai-browser-security/</link><pubDate>Sun, 17 May 2026 11:27:00 +0900</pubDate><guid>https://techllm.github.io/posts/exploitbench-ai-browser-security/</guid><description>Carnegie Mellon 대학의 ExploitBench 벤치마크가 Claude Mythos와 GPT-5.5의 실제 브라우저 취약점 악용 능력을 测试했다. Mythos가 압도적 성능을 보였지만, 비용도 12배 더 들어갔다.</description></item></channel></rss>