<?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>数据构造 on MessageDaily</title><link>https://inkeast.github.io/MessageDaily/tags/%E6%95%B0%E6%8D%AE%E6%9E%84%E9%80%A0/</link><description>Recent content in 数据构造 on MessageDaily</description><generator>Hugo</generator><language>zh-cn</language><lastBuildDate>Wed, 09 Sep 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://inkeast.github.io/MessageDaily/tags/%E6%95%B0%E6%8D%AE%E6%9E%84%E9%80%A0/index.xml" rel="self" type="application/rss+xml"/><item><title>Safety for Whom：把安全边界从话题级细化到边界级的数据配方 精读</title><link>https://inkeast.github.io/MessageDaily/posts/2026-09-09-boundary-aware-safety-refusal-paper-reading/</link><pubDate>Wed, 09 Sep 2026 00:00:00 +0000</pubDate><guid>https://inkeast.github.io/MessageDaily/posts/2026-09-09-boundary-aware-safety-refusal-paper-reading/</guid><description>Multiverse Computing 揭示话题级安全对齐的粒度错配：部署需要窄边界（拒操纵、答选举事实）而非话题级一刀切。提出窄边界形式化+离线自生成数据框架（受控话题生成/覆盖修复/补偿数据/边界对），Qwen3-8B 目标域拒绝 9.47%→84.75%、不安全率 26.26%→0.14%，并量化数据成分对安全-可用权衡的决定作用（过度拒绝 74%→5.2%）。</description></item></channel></rss>