<?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>⚡ 推演式决策 on Elon&#39;s AD Insight</title>
    <link>https://auto-driving-blog.pages.dev/tags/-%E6%8E%A8%E6%BC%94%E5%BC%8F%E5%86%B3%E7%AD%96/</link>
    <description>Recent content in ⚡ 推演式决策 on Elon&#39;s AD Insight</description>
    <image>
      <title>Elon&#39;s AD Insight</title>
      <url>https://auto-driving-blog.pages.dev/images/share.png</url>
      <link>https://auto-driving-blog.pages.dev/images/share.png</link>
    </image>
    <generator>Hugo</generator>
    <language>zh-cn</language>
    <lastBuildDate>Fri, 24 Jul 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://auto-driving-blog.pages.dev/tags/-%E6%8E%A8%E6%BC%94%E5%BC%8F%E5%86%B3%E7%AD%96/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>个人思考｜JEPA-DRIVE：用联合嵌入预测架构做自监督驾驶决策</title>
      <link>https://auto-driving-blog.pages.dev/posts/paper-reading/jepa-drive%E8%87%AA%E7%9B%91%E7%9D%A3%E9%A9%BE%E9%A9%B6%E5%86%B3%E7%AD%96%E7%B2%BE%E8%AF%BB/</link>
      <pubDate>Fri, 24 Jul 2026 00:00:00 +0000</pubDate>
      <guid>https://auto-driving-blog.pages.dev/posts/paper-reading/jepa-drive%E8%87%AA%E7%9B%91%E7%9D%A3%E9%A9%BE%E9%A9%B6%E5%86%B3%E7%AD%96%E7%B2%BE%E8%AF%BB/</guid>
      <description>JEPA-DRIVE 是一个基于 Joint Embedding Predictive Architecture 的自监督驾驶决策系统。它用结构化隐空间世界模型编码驾驶场景，从 65,536 个离散行为词汇中组合生成候选轨迹，再用 cycle energy 重建误差作为自监督评分信号。V17 在 NAVSIM 上达到 0.8839 selected_pdm，V21 正在用纯自监督 cycle energy 突破 PDM 代理标签的 64 值瓶颈（8×A800 分布式训练中）。全模型仅 3.88M 参数。</description>
    </item>
  </channel>
</rss>
