<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>🦾 移动操作 on Elon&#39;s AD Insight</title>
    <link>https://auto-driving-blog.vercel.app/tags/-%E7%A7%BB%E5%8A%A8%E6%93%8D%E4%BD%9C/</link>
    <description>Recent content in 🦾 移动操作 on Elon&#39;s AD Insight</description>
    <generator>Hugo</generator>
    <language>zh-cn</language>
    <lastBuildDate>Sun, 19 Jul 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://auto-driving-blog.vercel.app/tags/-%E7%A7%BB%E5%8A%A8%E6%93%8D%E4%BD%9C/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>论文精读｜ABot-M0.5：统一的移动操作世界动作模型</title>
      <link>https://auto-driving-blog.vercel.app/posts/paper-reading/%E8%AE%BA%E6%96%87%E7%B2%BE%E8%AF%BB-2607-00678/</link>
      <pubDate>Sun, 19 Jul 2026 00:00:00 +0000</pubDate>
      <guid>https://auto-driving-blog.vercel.app/posts/paper-reading/%E8%AE%BA%E6%96%87%E7%B2%BE%E8%AF%BB-2607-00678/</guid>
      <description>ABot-M0.5 用&amp;rsquo;三层对齐&amp;rsquo;策略重新设计了移动操作的世界动作模型：潜在动作模块在帧级粒度捕捉视觉状态转移，解决时间粒度粗糙导致的精细接触丢失。双层 Mixture-of-Transformers 解耦导航与操作的动作空间，dream-forcing 训练策略打通训练与推理的分布鸿沟。在长程任务和精细控制上同时达到 SOTA。</description>
    </item>
  </channel>
</rss>
