<?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>测试时优化 on Elon&#39;s AD Insight</title>
    <link>https://auto-driving-blog.vercel.app/tags/%E6%B5%8B%E8%AF%95%E6%97%B6%E4%BC%98%E5%8C%96/</link>
    <description>Recent content in 测试时优化 on Elon&#39;s AD Insight</description>
    <generator>Hugo</generator>
    <language>zh-cn</language>
    <lastBuildDate>Sat, 18 Jul 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://auto-driving-blog.vercel.app/tags/%E6%B5%8B%E8%AF%95%E6%97%B6%E4%BC%98%E5%8C%96/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>论文精读：TOAD — Test-Time Trajectory Optimization for Autonomous Driving</title>
      <link>https://auto-driving-blog.vercel.app/posts/paper-reading/%E8%AE%BA%E6%96%87%E7%B2%BE%E8%AF%BB-2606-07170/</link>
      <pubDate>Sat, 18 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-2606-07170/</guid>
      <description>端到端自动驾驶规划器通常先生成候选轨迹再评分，但评分器无法影响候选集本身，弱候选集会限制性能。TOAD将评分器重新定义为学习到的轨迹级奖励函数，在测试时使用交叉熵方法（CEM）搜索最大化该奖励的轨迹，无需重训练即可即插即用。在NAVSIM-v1（94.7 PDMS）、NAVSIM-v2（56.3 EPDMS）和闭环HUGSIM基准上，TOAD一致提升了六种基规划器的表现，并深入分析了评分器泛化能力对测试时搜索的关键影响。</description>
    </item>
  </channel>
</rss>
