<?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>模型架构 on 硅基 AGI · 智能体学习与测评</title><link>https://guijiagi.com/tags/%E6%A8%A1%E5%9E%8B%E6%9E%B6%E6%9E%84/</link><description>Recent content in 模型架构 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 10:15:00 +0800</lastBuildDate><atom:link href="https://guijiagi.com/tags/%E6%A8%A1%E5%9E%8B%E6%9E%B6%E6%9E%84/index.xml" rel="self" type="application/rss+xml"/><item><title>MoE架构深度解析：混合专家模型的训练与推理优化</title><link>https://guijiagi.com/posts/b2-4eded1bd/</link><pubDate>Thu, 16 Jul 2026 10:15:00 +0800</pubDate><guid>https://guijiagi.com/posts/b2-4eded1bd/</guid><description>从数学原理到工程实现，全面剖析MoE模型的路由机制、训练挑战与推理优化策略</description></item><item><title>深度解析注意力机制的变体与演进</title><link>https://guijiagi.com/posts/article-55/</link><pubDate>Mon, 13 Jul 2026 02:00:00 +0800</pubDate><guid>https://guijiagi.com/posts/article-55/</guid><description>从标准Self-Attention到Linear Attention、Flash Attention、Sparse Attention，全面梳理注意力机制的演进脉络</description></item><item><title>深度解析MoE架构：混合专家模型如何工作</title><link>https://guijiagi.com/posts/article-33/</link><pubDate>Sun, 12 Jul 2026 22:20:00 +0800</pubDate><guid>https://guijiagi.com/posts/article-33/</guid><description>从路由机制到负载均衡，全面拆解Mixture of Experts架构的设计原理</description></item><item><title>深度解析MoE架构：混合专家模型如何工作</title><link>https://guijiagi.com/posts/b2-0e8204a0/</link><pubDate>Sun, 12 Jul 2026 22:20:00 +0800</pubDate><guid>https://guijiagi.com/posts/b2-0e8204a0/</guid><description>从路由机制到负载均衡，全面拆解Mixture of Experts架构的设计原理</description></item><item><title>混合深度(MoD)：动态计算深度</title><link>https://guijiagi.com/posts/mixture-of-depths-explained/</link><pubDate>Thu, 02 Jul 2026 10:56:00 +0800</pubDate><guid>https://guijiagi.com/posts/mixture-of-depths-explained/</guid><description>解析Mixture-of-Depth如何让模型按需调整计算深度，实现效率与性能的平衡</description></item><item><title>MoE混合专家架构2026详解</title><link>https://guijiagi.com/posts/moe-architecture-2026-detail/</link><pubDate>Thu, 02 Jul 2026 10:52:00 +0800</pubDate><guid>https://guijiagi.com/posts/moe-architecture-2026-detail/</guid><description>从GShard到DeepSeek-V3，全面解析MoE架构的设计演进与工程实践</description></item><item><title>DeepSeek 技术解析：开源大模型的性价比之王</title><link>https://guijiagi.com/posts/deepseek-technical-analysis/</link><pubDate>Wed, 24 Jun 2026 11:40:00 +0800</pubDate><guid>https://guijiagi.com/posts/deepseek-technical-analysis/</guid><description>DeepSeek 系列模型的技术架构、训练方法和工程优化深度分析</description></item><item><title>MoE 混合专家模型选型指南：从 Mixtral 到 DeepSeek</title><link>https://guijiagi.com/posts/mixture-of-experts-guide/</link><pubDate>Tue, 23 Jun 2026 16:10:00 +0800</pubDate><guid>https://guijiagi.com/posts/mixture-of-experts-guide/</guid><description>MoE架构原理、主流模型对比和生产环境选型建议</description></item></channel></rss>