<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Alignment on Text Matrix</title><link>https://txtmix.com/tags/alignment/</link><description>Recent content in Alignment on Text Matrix</description><generator>Hugo</generator><language>zh-cn</language><lastBuildDate>Sat, 23 May 2026 08:55:34 +0800</lastBuildDate><atom:link href="https://txtmix.com/tags/alignment/index.xml" rel="self" type="application/rss+xml"/><item><title>AI 心理学：Anthropic 如何用 171 个情绪向量重新逼近模型内部世界</title><link>https://txtmix.com/posts/tech/ai-psychology-anthropic-emotion-vectors/</link><pubDate>Wed, 15 Apr 2026 18:25:00 +0800</pubDate><guid>https://txtmix.com/posts/tech/ai-psychology-anthropic-emotion-vectors/</guid><description>这不是一篇把模型简单拟人化的文章，而是一篇把 Anthropic 最新研究串成一条逻辑链的技术解读：角色空间、情绪向量、奖励黑客、对齐伪装与 CoT 不忠实，如何共同指向更深的模型内部状态问题。</description></item></channel></rss>