<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Adapter on BOLIN的个人博客</title>
    <link>https://example.org/tags/adapter/</link>
    <description>Recent content in Adapter on BOLIN的个人博客</description>
    <generator>Hugo -- gohugo.io</generator>
    <language>zh-cn</language>
    <copyright>© 2026 BOLIN</copyright>
    <lastBuildDate>Thu, 30 Jul 2026 02:30:00 +0800</lastBuildDate><atom:link href="https://example.org/tags/adapter/index.xml" rel="self" type="application/rss+xml" />
    
    <item>
      <title>Acquire and then Adapt：轻量的是适配器，不是 Flux</title>
      <link>https://example.org/papers/acquire-and-adapt/</link>
      <pubDate>Thu, 30 Jul 2026 02:30:00 +0800</pubDate>
      
      <guid>https://example.org/papers/acquire-and-adapt/</guid>
      <description>FluxIR 并没有把 Flux 变成小模型：12B 主干仍参与每次前向计算，只是不更新权重。真正节省的是可训练参数、梯度与优化器状态。</description>
      
    </item>
    
  </channel>
</rss>
