<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>LLM内省 on AI内参</title>
    <link>https://www.neican.ai/tags/llm%E5%86%85%E7%9C%81/</link>
    <description>Recent content in LLM内省 on AI内参</description>
    <generator>Hugo</generator>
    <language>zh-cn</language>
    <lastBuildDate>Thu, 30 Oct 2025 18:40:07 +0800</lastBuildDate>
    <atom:link href="https://www.neican.ai/tags/llm%E5%86%85%E7%9C%81/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>AI元认知萌芽：Anthropic「概念注入」实验揭示心智黑箱与信任鸿沟</title>
      <link>https://www.neican.ai/insights/aianthropic-20251030184007815-0/</link>
      <pubDate>Thu, 30 Oct 2025 18:40:07 +0800</pubDate>
      <guid>https://www.neican.ai/insights/aianthropic-20251030184007815-0/</guid>
      <description>Anthropic的“概念注入”实验首次证实，大型语言模型（LLM）已具备初步的内部自我审查和认知控制能力，颠覆了传统认知。这项技术突破为理解AI“心智”黑箱提供了前所未有的显微镜，揭示模型可在输出前察觉内部“思想”并区分“意图”。然而，模型内省能力的不稳定性与自我报告的不忠实性，正加速AI可解释性成为构建可信AI系统、应对潜在“隐藏意图”及重塑人机信任关系的关键产业与伦理挑战。</description>
    </item>
  </channel>
</rss>
