<?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>🚫 无感知标注 on Elon&#39;s AD Insight</title>
    <link>https://auto-driving-blog.vercel.app/tags/-%E6%97%A0%E6%84%9F%E7%9F%A5%E6%A0%87%E6%B3%A8/</link>
    <description>Recent content in 🚫 无感知标注 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/-%E6%97%A0%E6%84%9F%E7%9F%A5%E6%A0%87%E6%B3%A8/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>论文精读｜World4Drive：用意图感知的物理隐世界模型，摆脱感知标注做端到端规划</title>
      <link>https://auto-driving-blog.vercel.app/posts/paper-reading/%E8%AE%BA%E6%96%87%E7%B2%BE%E8%AF%BB-2507-00603/</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-2507-00603/</guid>
      <description>World4Drive 是首个将多模态驾驶意图与物理隐世界模型结合的端到端框架，无需 3D 框/高精地图等昂贵感知标注。它借 Grounded-SAM 的语义先验和 Metric3D 的深度先验构建&amp;rsquo;懂物理&amp;rsquo;的隐空间，让模型在多种驾驶意图下想象未来、用世界模型打分选最优轨迹。L2 误差降 18.1%、碰撞率降 46.7%、收敛快 3.75 倍，被 ICCV 2025 收录。</description>
    </item>
  </channel>
</rss>
