<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title>KL - 标签 - lilfry's library</title><link>https://lilfry09.github.io/tags/kl/</link><description>KL - 标签 - lilfry's library</description><generator>Hugo -- gohugo.io</generator><language>zh-CN</language><managingEditor>lilfry@sjtu.edu.cn (lilfry)</managingEditor><webMaster>lilfry@sjtu.edu.cn (lilfry)</webMaster><copyright>All rights reserved.</copyright><lastBuildDate>Sun, 09 Aug 2026 21:00:00 +0800</lastBuildDate><atom:link href="https://lilfry09.github.io/tags/kl/" rel="self" type="application/rss+xml"/><item><title>读 SFT、RL 与 On-Policy Distillation：后训练真正的变量是 on-policy data</title><link>https://lilfry09.github.io/ai/sft-rl-opd-on-policy-data/</link><pubDate>Sun, 09 Aug 2026 21:00:00 +0800</pubDate><author>fry</author><guid>https://lilfry09.github.io/ai/sft-rl-opd-on-policy-data/</guid><description><![CDATA[<p>这篇原文最值得带走的一句话，我会改写成：</p>
<p><strong>后训练不是在同一条损失曲线上做微调，而是在改模型会访问到的分布形状。</strong></p>
<p>SFT 把模型往外部示范拉，RL 让模型在自己会走到的状态上追逐奖励，OPD 则让 student 在自己的 prefixes 上对齐 teacher。只要盯住“数据来自哪里”，很多看起来玄学的现象就会变得很朴素：为什么 RL 和 OPD 常常比 SFT 更不容易忘，为什么 student 甚至能超过 teacher。</p>
<p>原文在这里：<a href="https://nrehiew.github.io/blog/sft_rl_opd/" target="_blank" rel="noopener noreffer ">SFT, RL, and On-Policy Distillation Through a Distributional Lens</a>。</p>
<p></p>]]></description></item></channel></rss>