<?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>压缩 on 硅基 AGI · 智能体学习与测评</title><link>https://guijiagi.com/tags/%E5%8E%8B%E7%BC%A9/</link><description>Recent content in 压缩 on 硅基 AGI · 智能体学习与测评</description><generator>Hugo</generator><language>zh-cn</language><copyright>本站内容采用 CC BY-NC-SA 4.0 国际许可协议授权</copyright><lastBuildDate>Thu, 16 Jul 2026 10:32:00 +0800</lastBuildDate><atom:link href="https://guijiagi.com/tags/%E5%8E%8B%E7%BC%A9/index.xml" rel="self" type="application/rss+xml"/><item><title>大模型上下文窗口的工程优化：从朴素截断到结构化压缩</title><link>https://guijiagi.com/posts/b2-5052092c/</link><pubDate>Thu, 16 Jul 2026 10:32:00 +0800</pubDate><guid>https://guijiagi.com/posts/b2-5052092c/</guid><description>系统梳理大模型上下文窗口管理的工程方案，涵盖滑动窗口、层次化摘要、选择性记忆与结构化压缩技术</description></item><item><title>LLM 知识蒸馏：从大模型到小模型的能力迁移</title><link>https://guijiagi.com/posts/llm-knowledge-distillation/</link><pubDate>Thu, 25 Jun 2026 12:00:00 +0800</pubDate><guid>https://guijiagi.com/posts/llm-knowledge-distillation/</guid><description>深入讲解 LLM 知识蒸馏的完整体系：黑盒蒸馏、白盒蒸馏、在线/离线蒸馏，涵盖 Logit 蒸馏、中间层蒸馏、指令蒸馏、响应蒸馏等方法，附 PyTorch 代码实现与效果对比。</description></item></channel></rss>