<?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>환각 on WithinTrend — Trending Topics &amp; Guides</title><link>https://withintrend.org/tags/%ED%99%98%EA%B0%81/</link><description>Recent content in 환각 on WithinTrend — Trending Topics &amp; Guides</description><generator>Hugo -- 0.146.0</generator><language>ko-kr</language><lastBuildDate>Sun, 21 Jun 2026 05:55:00 +0900</lastBuildDate><atom:link href="https://withintrend.org/tags/%ED%99%98%EA%B0%81/index.xml" rel="self" type="application/rss+xml"/><item><title>GPT-5.5는 모를 때 86% 확률로 지어낸다 — 큰 모델일수록 환각이 늘어나는 역설</title><link>https://withintrend.org/posts/2026-06-21-gpt-5-5-hallucination-86-percent/</link><pubDate>Sun, 21 Jun 2026 05:55:00 +0900</pubDate><guid>https://withintrend.org/posts/2026-06-21-gpt-5-5-hallucination-86-percent/</guid><description>AA-Omniscience 벤치마크에서 GPT-5.5는 정확도 1위(57%)와 환각률 86%를 동시에 기록했다. 큰 모델은 더 많이 알지만 더 자신 있게 거짓말한다는 신호다.</description></item></channel></rss>