<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Prefix Tuning on Nalar</title>
    <link>https://nalar.dev/tags/prefix-tuning/</link>
    <description>Recent content in Prefix Tuning on Nalar</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Fri, 11 Sep 2026 00:00:00 +0700</lastBuildDate>
    <atom:link href="https://nalar.dev/tags/prefix-tuning/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Adapt Language Models with Prefix Tuning</title>
      <link>https://nalar.dev/adapt-language-models-with-prefix-tuning/</link>
      <pubDate>Fri, 11 Sep 2026 00:00:00 +0700</pubDate>
      <guid>https://nalar.dev/adapt-language-models-with-prefix-tuning/</guid>
      <description>&lt;h1 id=&#34;adapt-language-models-with-prefix-tuning&#34;&gt;Adapt Language Models with Prefix Tuning&lt;/h1&gt;&#xA;&lt;p&gt;Full fine-tuning changes a model&amp;rsquo;s weights for each task. That can be effective, but storing and serving a separate full checkpoint for every task becomes expensive as model size and task count grow. Prefix tuning offers a different arrangement: keep the pretrained model frozen and train a small set of task-specific states that participate in attention.&lt;/p&gt;&#xA;&lt;p&gt;This article builds a practical mental model for prefix tuning, shows how it differs from text prompts and low-rank weight adapters, and explains the trade-offs that matter when training or serving several task variants.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
