<?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>Lossy Compression on Haxlys's Blog</title><link>https://hugo-blog-static-site.haxlys.workers.dev/mental_models/lossy-compression/</link><description>Recent content in Lossy Compression on Haxlys's Blog</description><generator>Hugo -- 0.152.2</generator><language>kr-ko</language><lastBuildDate>Wed, 13 May 2026 13:40:29 +0900</lastBuildDate><atom:link href="https://hugo-blog-static-site.haxlys.workers.dev/mental_models/lossy-compression/index.xml" rel="self" type="application/rss+xml"/><item><title>Useful Memories Become Faulty When Continuously Updated by LLMs</title><link>https://hugo-blog-static-site.haxlys.workers.dev/posts/faulty-memory-summary/</link><pubDate>Wed, 13 May 2026 13:40:29 +0900</pubDate><guid>https://hugo-blog-static-site.haxlys.workers.dev/posts/faulty-memory-summary/</guid><description>LLM 에이전트가 경험을 텍스트 메모리로 반복 재작성할수록 성능이 떨어질 수 있다는 Dylan Zhang의 &amp;#39;Faulty Memory&amp;#39; 핵심 결과를 정리합니다.</description></item></channel></rss>