<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Tianao Zhang | Ruixuan Yang</title><link>https://yrxxxxxxxx1007.github.io/authors/tianao-zhang/</link><atom:link href="https://yrxxxxxxxx1007.github.io/authors/tianao-zhang/index.xml" rel="self" type="application/rss+xml"/><description>Tianao Zhang</description><generator>HugoBlox Kit (https://hugoblox.com)</generator><language>en-us</language><lastBuildDate>Sat, 26 Sep 2026 00:00:00 +0000</lastBuildDate><item><title>Fewer Tokens, More Self-Teaching: On-Policy Self-Distillation for Extreme Visual Token Reduction</title><link>https://yrxxxxxxxx1007.github.io/publications/lt-opd/</link><pubDate>Sat, 26 Sep 2026 00:00:00 +0000</pubDate><guid>https://yrxxxxxxxx1007.github.io/publications/lt-opd/</guid><description>&lt;h2 id="overview"&gt;Overview&lt;/h2&gt;
&lt;p&gt;Recovering multimodal capabilities under extreme visual-token reduction through on-policy self-distillation and a progressive token-budget curriculum.&lt;/p&gt;
&lt;p&gt;
&lt;figure &gt;
&lt;div class="flex justify-center "&gt;
&lt;div class="w-full" &gt;
&lt;img alt="LT-OPD framework with a low-token student, a frozen full-token teacher, and a token-budget curriculum"
srcset="https://yrxxxxxxxx1007.github.io/publications/lt-opd/featured_hu_6aa0ccbbac72b2d9.webp 320w, https://yrxxxxxxxx1007.github.io/publications/lt-opd/featured_hu_243610fc64eb846f.webp 480w, https://yrxxxxxxxx1007.github.io/publications/lt-opd/featured_hu_2991ac9026c4f9fc.webp 760w"
sizes="(max-width: 480px) 100vw, (max-width: 768px) 90vw, (max-width: 1024px) 80vw, 760px"
src="https://yrxxxxxxxx1007.github.io/publications/lt-opd/featured_hu_6aa0ccbbac72b2d9.webp"
width="760"
height="382"
loading="lazy" data-zoomable /&gt;&lt;/div&gt;
&lt;/div&gt;&lt;/figure&gt;
&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;arXiv 2026 · Preprint.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;&lt;a href="https://arxiv.org/abs/2609.32353" target="_blank" rel="noopener"&gt;Read the paper&lt;/a&gt; · &lt;a href="https://arxiv.org/pdf/2609.32353" target="_blank" rel="noopener"&gt;Download PDF&lt;/a&gt; · &lt;a href="https://github.com/Yrxxxxxxxx1007/LT-OPD" target="_blank" rel="noopener"&gt;Code&lt;/a&gt;&lt;/p&gt;</description></item></channel></rss>