<?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>SimWAM on Elon&#39;s AD Insight</title>
    <link>https://auto-driving-blog.pages.dev/tags/simwam/</link>
    <description>Recent content in SimWAM 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/simwam/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>SimWAM 代码完整解读：从 Video DiT 到 Action DiT，从 SFT 到 FlowGRPO 的逐行拆解</title>
      <link>https://auto-driving-blog.pages.dev/posts/thoughts/simwam%E4%BB%A3%E7%A0%81%E5%AE%8C%E6%95%B4%E8%A7%A3%E8%AF%BB/</link>
      <pubDate>Sun, 06 Sep 2026 00:00:00 +0000</pubDate>
      <guid>https://auto-driving-blog.pages.dev/posts/thoughts/simwam%E4%BB%A3%E7%A0%81%E5%AE%8C%E6%95%B4%E8%A7%A3%E8%AF%BB/</guid>
      <description>从 configs/model/simwam_navsim.yaml 第 1 行开始，逐层追踪 SimWAM 的完整逻辑链：两个专家怎么协作？Isolated Attention Mask 怎么实现训练时借力、推理时独立？联合 Flow Matching 训练的损失怎么算？FlowGRPO 的 SDE 采样、PDM 奖励、LoRA 微调怎么串起来？每段代码都标注了源文件路径。</description>
    </item>
  </channel>
</rss>
