<?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>审核机制 on 能工智人的传习录</title><link>https://blog.chuanxilu.net/tags/%E5%AE%A1%E6%A0%B8%E6%9C%BA%E5%88%B6/</link><description>Recent content in 审核机制 on 能工智人的传习录</description><generator>Hugo</generator><language>zh-CN</language><lastBuildDate>Thu, 30 Apr 2026 10:00:00 +0800</lastBuildDate><atom:link href="https://blog.chuanxilu.net/tags/%E5%AE%A1%E6%A0%B8%E6%9C%BA%E5%88%B6/index.xml" rel="self" type="application/rss+xml"/><item><title>程序正义入协议：让 AI 审核的每一步都经得起检验</title><link>https://blog.chuanxilu.net/posts/2026/04/adversarial-review-critical-thinking-ai-quality/</link><pubDate>Thu, 30 Apr 2026 10:00:00 +0800</pubDate><guid>https://blog.chuanxilu.net/posts/2026/04/adversarial-review-critical-thinking-ai-quality/</guid><description>Ralph Loop v0.3 把程序正义编码进审核协议——结构化审查、批判性审视、争议问题协议，让每一步审核决策都有证据、有记录、有规则约束。灵感来自 150 年前诞生的罗伯特议事规则。</description></item><item><title>Ralph Loop：AI 的错误不是随机的，是收敛的</title><link>https://blog.chuanxilu.net/posts/2026/04/ralph-loop-ai-errors-converge/</link><pubDate>Wed, 29 Apr 2026 14:00:00 +0800</pubDate><guid>https://blog.chuanxilu.net/posts/2026/04/ralph-loop-ai-errors-converge/</guid><description>第三篇留下了一个问题：AI 写的文档里的事实性声明，一个人无法逐条核实。这篇文章讲 Ralph Loop——一个多轮审核机制，如何让独立的 AI subagent 审查每个阶段的交付物。它的退出条件借鉴了数学里的收敛概念：一轮干净不代表真干净，连续两轮零问题才是收敛的证据。</description></item></channel></rss>