<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>MoE on LoopRouter Blog</title><link>https://blog.looprouter.com/tags/moe/</link><description>Recent content in MoE on LoopRouter Blog</description><generator>Hugo</generator><language>zh-cn</language><lastBuildDate>Mon, 21 Sep 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://blog.looprouter.com/tags/moe/index.xml" rel="self" type="application/rss+xml"/><item><title>把稠密模型改造成 MoE 来省钱:量化、剪枝、蒸馏、MoE 化怎么选</title><link>https://blog.looprouter.com/docs/dense-to-moe-inference-cost-tradeoff/</link><pubDate>Mon, 21 Sep 2026 00:00:00 +0000</pubDate><guid>https://blog.looprouter.com/docs/dense-to-moe-inference-cost-tradeoff/</guid><description>&lt;blockquote&gt;&#10;&lt;p&gt;本文基于 arXiv 论文《Accelerating Dense LLMs via L0-regularized Mixture-of-Experts》(arXiv:2609.21672v1,原发表于 ACL 2025)公开内容的整理与二次分析,加入 LoopRouter 在成本治理与接入路由上的视角;这是一篇技术方法论文,本文作为「降本技术选型」的科普与延伸,数据引自原文、未经独立复现。原文见文末链接。&lt;/p&gt;</description></item></channel></rss>