<?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>📐 BEV感知 on Elon&#39;s AD Insight</title>
    <link>https://auto-driving-blog.pages.dev/tags/-bev%E6%84%9F%E7%9F%A5/</link>
    <description>Recent content in 📐 BEV感知 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>Sun, 06 Sep 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://auto-driving-blog.pages.dev/tags/-bev%E6%84%9F%E7%9F%A5/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>论文精读｜Qwen-Drive-1.0：阿里通义首个驾驶视觉-语言基础模型</title>
      <link>https://auto-driving-blog.pages.dev/posts/paper-reading/qwen-drive-1/</link>
      <pubDate>Sun, 06 Sep 2026 00:00:00 +0000</pubDate>
      <guid>https://auto-driving-blog.pages.dev/posts/paper-reading/qwen-drive-1/</guid>
      <description>Qwen-Drive-1.0 基于 Qwen3.5-4B VLM 外挂两个模块：BEV感知头（3D检测+占用预测+地图分割）和 Planning Expert（~1.1B 参数 Flow Matching 轨迹生成器），四阶段渐进训练。关键设计：&lt;strong&gt;不设推理时打分器&lt;/strong&gt;，直接用 Flow Matching 单次采样输出轨迹，RL 后训练用 GRPO 式分组优势优化。NAVSIM v1.1 榜上 90.7 PDMS（RL），WOD-E2E 7.91 RFS，AlpaSim 闭环 0.37 分。</description>
    </item>
  </channel>
</rss>
