<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>模型对齐 :: 标签 :: x7peeps</title><link>https://x7peeps.com/tags/%E6%A8%A1%E5%9E%8B%E5%AF%B9%E9%BD%90/index.html</link><description/><generator>Hugo</generator><language>zh-CN</language><lastBuildDate>Mon, 20 Jul 2026 16:00:00 +0800</lastBuildDate><atom:link href="https://x7peeps.com/tags/%E6%A8%A1%E5%9E%8B%E5%AF%B9%E9%BD%90/index.xml" rel="self" type="application/rss+xml"/><item><title>大模型微调与适配器安全取证深度分析</title><link>https://x7peeps.com/%E5%AE%89%E5%85%A8/%E5%BA%94%E6%80%A5%E5%93%8D%E5%BA%94/0x03%E5%8F%96%E8%AF%81%E5%88%86%E6%9E%90/%E5%A4%A7%E6%A8%A1%E5%9E%8B%E5%BE%AE%E8%B0%83%E4%B8%8E%E9%80%82%E9%85%8D%E5%99%A8%E5%AE%89%E5%85%A8%E5%8F%96%E8%AF%81%E6%B7%B1%E5%BA%A6%E5%88%86%E6%9E%90/index.html</link><pubDate>Mon, 20 Jul 2026 16:00:00 +0800</pubDate><guid>https://x7peeps.com/%E5%AE%89%E5%85%A8/%E5%BA%94%E6%80%A5%E5%93%8D%E5%BA%94/0x03%E5%8F%96%E8%AF%81%E5%88%86%E6%9E%90/%E5%A4%A7%E6%A8%A1%E5%9E%8B%E5%BE%AE%E8%B0%83%E4%B8%8E%E9%80%82%E9%85%8D%E5%99%A8%E5%AE%89%E5%85%A8%E5%8F%96%E8%AF%81%E6%B7%B1%E5%BA%A6%E5%88%86%E6%9E%90/index.html</guid><description>系统剖析大模型微调与适配器生态的安全威胁与取证分析方法论，涵盖LoRA/QLoRA微调后门植入检测、SFT训练数据投毒取证、RLHF奖励劫持与对齐逃逸分析、适配器权重篡改与合并攻击检测、微调流水线供应链安全审计，结合开源社区与企业LLM落地场景的真实攻击案例，构建从训练日志到模型行为的全链路取证能力</description></item></channel></rss>