<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Self-Distillation | Bingbing Wen</title><link>https://bbwen.github.io/tags/self-distillation/</link><atom:link href="https://bbwen.github.io/tags/self-distillation/index.xml" rel="self" type="application/rss+xml"/><description>Self-Distillation</description><generator>Hugo Blox Builder (https://hugoblox.com)</generator><language>en-us</language><lastBuildDate>Tue, 01 Sep 2026 00:00:00 +0000</lastBuildDate><image><url>https://bbwen.github.io/media/icon_hu_982c5d63a71b2961.png</url><title>Self-Distillation</title><link>https://bbwen.github.io/tags/self-distillation/</link></image><item><title>STOP: Structured On-Policy Pruning of Long-Form Reasoning in Low-Data Regimes</title><link>https://bbwen.github.io/publications/stop-structured-on-policy-pruning/</link><pubDate>Tue, 01 Sep 2026 00:00:00 +0000</pubDate><guid>https://bbwen.github.io/publications/stop-structured-on-policy-pruning/</guid><description>&lt;p&gt;STOP uses a model&amp;rsquo;s own reasoning traces to create more efficient fine-tuning examples. It segments and annotates each trace, builds a reasoning tree, and retains the prefix ending at the earliest correct answering conclusion. This removes unnecessary reasoning after a solution while preserving a coherent path to the answer.&lt;/p&gt;</description></item></channel></rss>