
  <rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
    <channel>
      <title>HANS</title>
      <link>https://www.hansking.cn/blog</link>
      <description>A posts created with Next.js and Tailwind.css</description>
      <language>en-us</language>
      <managingEditor>admin@hansking.cn (HANS)</managingEditor>
      <webMaster>admin@hansking.cn (HANS)</webMaster>
      <lastBuildDate>Sun, 22 Jun 2025 10:00:00 GMT</lastBuildDate>
      <atom:link href="https://www.hansking.cn/tags/Computer Use/feed.xml" rel="self" type="application/rss+xml"/>
      
  <item>
    <guid>https://www.hansking.cn/blog/2026/ComputerUse-liu-lan-qi-cao-zuo-de-zhuang-tai-ji-she-ji</guid>
    <title>Computer Use：浏览器操作的状态机设计</title>
    <link>https://www.hansking.cn/blog/2026/ComputerUse-liu-lan-qi-cao-zuo-de-zhuang-tai-ji-she-ji</link>
    <description>[{&quot;value&quot;:&quot;Computer Use：浏览器操作的状态机设计&quot;,&quot;url&quot;:&quot;#computer-use浏览器操作的状态机设计&quot;,&quot;depth&quot;:1},{&quot;value&quot;:&quot;一、把浏览器任务建模成状态机&quot;,&quot;url&quot;:&quot;#一把浏览器任务建模成状态机&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;二、观察结果要包含语义和版本&quot;,&quot;url&quot;:&quot;#二观察结果要包含语义和版本&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;三、动作应该是受限的工具集合&quot;,&quot;url&quot;:&quot;#三动作应该是受限的工具集合&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;四、等待不能靠固定 sleep&quot;,&quot;url&quot;:&quot;#四等待不能靠固定-sleep&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;五、表单填写要分为填充和提交&quot;,&quot;url&quot;:&quot;#五表单填写要分为填充和提交&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;六、动作结果要重新观察&quot;,&quot;url&quot;:&quot;#六动作结果要重新观察&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;七、导航和域名必须受限&quot;,&quot;url&quot;:&quot;#七导航和域名必须受限&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;八、失败恢复要区分可重试和不可重试&quot;,&quot;url&quot;:&quot;#八失败恢复要区分可重试和不可重试&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;九、高风险动作必须人工确认&quot;,&quot;url&quot;:&quot;#九高风险动作必须人工确认&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;十、评测看任务结果和动作过程&quot;,&quot;url&quot;:&quot;#十评测看任务结果和动作过程&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;总结：浏览器操作首先是状态管理&quot;,&quot;url&quot;:&quot;#总结浏览器操作首先是状态管理&quot;,&quot;depth&quot;:2}]</description>
    <pubDate>Sun, 22 Jun 2025 10:00:00 GMT</pubDate>
    <author>admin@hansking.cn (HANS)</author>
    <category>AI</category><category>LLM</category><category>Computer Use</category><category>Agent</category><category>浏览器自动化</category>
  </item>

  <item>
    <guid>https://www.hansking.cn/blog/2026duo-mo-tai-Agent-rang-mo-xing-du-tu-du-biao-du-ping-mu</guid>
    <title>多模态 Agent：让模型读图、读表、读屏幕</title>
    <link>https://www.hansking.cn/blog/2026duo-mo-tai-Agent-rang-mo-xing-du-tu-du-biao-du-ping-mu</link>
    <description>[{&quot;value&quot;:&quot;多模态 Agent：让模型读图、读表、读屏幕&quot;,&quot;url&quot;:&quot;#多模态-agent让模型读图读表读屏幕&quot;,&quot;depth&quot;:1},{&quot;value&quot;:&quot;一、先区分三类多模态任务&quot;,&quot;url&quot;:&quot;#一先区分三类多模态任务&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;1. 视觉问答&quot;,&quot;url&quot;:&quot;#1-视觉问答&quot;,&quot;depth&quot;:3},{&quot;value&quot;:&quot;2. 视觉结构化&quot;,&quot;url&quot;:&quot;#2-视觉结构化&quot;,&quot;depth&quot;:3},{&quot;value&quot;:&quot;3. 视觉行动&quot;,&quot;url&quot;:&quot;#3-视觉行动&quot;,&quot;depth&quot;:3},{&quot;value&quot;:&quot;二、图像预处理决定模型能不能看清&quot;,&quot;url&quot;:&quot;#二图像预处理决定模型能不能看清&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;三、表格不能只当成一张图片&quot;,&quot;url&quot;:&quot;#三表格不能只当成一张图片&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;四、结构化输出要有证据和校验&quot;,&quot;url&quot;:&quot;#四结构化输出要有证据和校验&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;五、屏幕理解的关键是状态，不是坐标&quot;,&quot;url&quot;:&quot;#五屏幕理解的关键是状态不是坐标&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;六、Agent 要知道什么时候自己不确定&quot;,&quot;url&quot;:&quot;#六agent-要知道什么时候自己不确定&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;七、评测多模态 Agent 要测整条链路&quot;,&quot;url&quot;:&quot;#七评测多模态-agent-要测整条链路&quot;,&quot;depth&quot;:2},{&quot;value&quot;:&quot;八、我最后留下的工程原则&quot;,&quot;url&quot;:&quot;#八我最后留下的工程原则&quot;,&quot;depth&quot;:2}]</description>
    <pubDate>Tue, 19 May 2026 10:00:00 GMT</pubDate>
    <author>admin@hansking.cn (HANS)</author>
    <category>AI</category><category>LLM</category><category>多模态</category><category>Agent</category><category>Computer Use</category><category>top</category>
  </item>

    </channel>
  </rss>
