<?xml version="1.0" encoding="UTF-8"?>
<feed xmlns="http://www.w3.org/2005/Atom" xml:lang="zh-CN">
  <title>Embodied AI Reading Station</title>
  <subtitle>156 篇具身 AI 论文，用能读懂的语言重写</subtitle>
  <link href="https://estelledc.github.io/embodied-ai-reading-station/feed.xml" rel="self" type="application/atom+xml"/>
  <link href="https://estelledc.github.io/embodied-ai-reading-station/" rel="alternate" type="text/html"/>
  <id>https://estelledc.github.io/embodied-ai-reading-station/</id>
  <updated>2026-06-02T11:03:35.224Z</updated>
  <author><name>Jason</name></author>
  <entry>
    <title>Issue Nº 01 — How robots begin to think</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/issues/01/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/issues/01/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <summary>第一期 · 13 篇论文 · 让机器人看见、想清、动起来</summary>
    <content type="html">&lt;p&gt;第一期 · 13 篇论文 · 让机器人看见、想清、动起来&lt;/p&gt;&lt;p&gt;Published: 2026 · Spring&lt;/p&gt;&lt;p&gt;&lt;a href=&quot;https://estelledc.github.io/embodied-ai-reading-station/issues/01/&quot;&gt;Read full issue →&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>Issue Nº 02 — From robot brains to robot bodies</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/issues/02/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/issues/02/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <summary>第二期 · 156 篇论文 · 从 VLA 鼻祖到 2025 前沿，11 主题全景</summary>
    <content type="html">&lt;p&gt;第二期 · 156 篇论文 · 从 VLA 鼻祖到 2025 前沿，11 主题全景&lt;/p&gt;&lt;p&gt;Published: 2026 · Late Spring&lt;/p&gt;&lt;p&gt;&lt;a href=&quot;https://estelledc.github.io/embodied-ai-reading-station/issues/02/&quot;&gt;Read full issue →&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>Issue Nº 03 — A field, drawn</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/issues/03/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/issues/03/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <summary>第三期 · 给每个概念一张图 · 给每个段落一段类比</summary>
    <content type="html">&lt;p&gt;第三期 · 给每个概念一张图 · 给每个段落一段类比&lt;/p&gt;&lt;p&gt;Published: 2026 · Early Summer&lt;/p&gt;&lt;p&gt;&lt;a href=&quot;https://estelledc.github.io/embodied-ai-reading-station/issues/03/&quot;&gt;Read full issue →&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>Issue Nº 04 — Tools for the long read</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/issues/04/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/issues/04/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <summary>第四期 · 18 个工具 · 把单向读物变成有记忆的阅读伴侣</summary>
    <content type="html">&lt;p&gt;第四期 · 18 个工具 · 把单向读物变成有记忆的阅读伴侣&lt;/p&gt;&lt;p&gt;Published: 2026 · Early Summer&lt;/p&gt;&lt;p&gt;&lt;a href=&quot;https://estelledc.github.io/embodied-ai-reading-station/issues/04/&quot;&gt;Read full issue →&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>Issue Nº 05 — Routes through the field</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/issues/05/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/issues/05/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <summary>第五期 · 14 个新工具 · 把站点变成入门到深入的完整路径</summary>
    <content type="html">&lt;p&gt;第五期 · 14 个新工具 · 把站点变成入门到深入的完整路径&lt;/p&gt;&lt;p&gt;Published: 2026 · Late Spring&lt;/p&gt;&lt;p&gt;&lt;a href=&quot;https://estelledc.github.io/embodied-ai-reading-station/issues/05/&quot;&gt;Read full issue →&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>Issue Nº 06 — Bodies that learn</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/issues/06/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/issues/06/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <summary>第六期 · 硬件视角看 VLA · humanoid / legged / dexterous / wearable</summary>
    <content type="html">&lt;p&gt;第六期 · 硬件视角看 VLA · humanoid / legged / dexterous / wearable&lt;/p&gt;&lt;p&gt;Published: 2026 · Summer&lt;/p&gt;&lt;p&gt;&lt;a href=&quot;https://estelledc.github.io/embodied-ai-reading-station/issues/06/&quot;&gt;Read full issue →&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>Issue Nº 07 — One hundred small things</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/issues/07/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/issues/07/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <summary>第七期 · 100 件 enrichment 的总账 · 从 13 篇精读到完整阅读环境</summary>
    <content type="html">&lt;p&gt;第七期 · 100 件 enrichment 的总账 · 从 13 篇精读到完整阅读环境&lt;/p&gt;&lt;p&gt;Published: 2026 · Mid Summer&lt;/p&gt;&lt;p&gt;&lt;a href=&quot;https://estelledc.github.io/embodied-ai-reading-station/issues/07/&quot;&gt;Read full issue →&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <title>№ 156 · UniSim</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/papers/unisim/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/papers/unisim/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <category term="World Model &amp; Video Policy"/>
    <summary>看过海量视频后，你给它一个动作（说一句话 / 推一下机械臂 / 挪一下镜头），它就生成接下来世界长什么样的视频——像一台会脑补现实的&quot;游戏机&quot;。</summary>
  </entry>
  <entry>
    <title>№ 155 · Navigation World Models</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/papers/navigation-world-models/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/papers/navigation-world-models/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <category term="World Model &amp; Video Policy"/>
    <summary>让机器人&quot;走&quot;之前先在脑子里放一段未来几秒的画面，看会不会撞墙，再决定真的怎么走。</summary>
  </entry>
  <entry>
    <title>№ 154 · Genie: Generative Interactive Environments</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/papers/genie/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/papers/genie/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <category term="World Model &amp; Video Policy"/>
    <summary>Genie 看一堆游戏录屏，自己猜出每帧之间&quot;按了什么键&quot;，再用这个&quot;按键&quot;画出下一帧——把死视频变成能玩的小游戏。</summary>
  </entry>
  <entry>
    <title>№ 153 · GAIA-1</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/papers/gaia-1/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/papers/gaia-1/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <category term="World Model &amp; Video Policy"/>
    <summary>GAIA-1 是个会做梦的开车模拟器：给它一段街景视频的开头加一句&quot;我现在打方向盘&quot;，它能接着画出后面几秒街上看到的画面。</summary>
  </entry>
  <entry>
    <title>№ 152 · Cosmos World Foundation Model Platform</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/papers/cosmos-world-foundation/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/papers/cosmos-world-foundation/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <category term="World Model &amp; Video Policy"/>
    <summary>NVIDIA 用 2000 万小时真实视频，训了一个能&quot;猜下一秒物理世界长啥样&quot;的大模型，给机器人和无人车当通用底座。</summary>
  </entry>
  <entry>
    <title>№ 151 · 1X World Model Challenge</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/papers/1x-world-model-2025/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/papers/1x-world-model-2025/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <category term="World Model &amp; Video Policy"/>
    <summary>1X 教人形机器人 Neo &quot;脑补下一秒画面&quot;：拿现成视频 AI 当底子，喂自家机器人录像微调，再做成公开赛让大家来卷。</summary>
  </entry>
  <entry>
    <title>№ 150 · TWM: Transformer-based World Models</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/papers/transformer-world-model/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/papers/transformer-world-model/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <category term="World Model &amp; Video Policy"/>
    <summary>agent 在脑子里&quot;做梦&quot;练本事。这篇把梦的引擎从 RNN 换成 Transformer，记得更长，做得更准。</summary>
  </entry>
  <entry>
    <title>№ 149 · Transformers are Sample-Efficient World Models</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/papers/iris-world-model/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/papers/iris-world-model/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <category term="World Model &amp; Video Policy"/>
    <summary>把游戏画面切成一格格&quot;积木&quot;，让 AI 像写句子一样接龙下一帧，然后让它在脑子里&quot;自己跟自己玩&quot;练强化学习——只玩两小时就接近人类水平。</summary>
  </entry>
  <entry>
    <title>№ 148 · Dreamer V3: Mastering Diverse Domains through World Models</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/papers/dreamer-v3/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/papers/dreamer-v3/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <category term="World Model &amp; Video Policy"/>
    <summary>同一套设置，让一个 AI 自己玩 150 多种游戏都不用改参数，还第一次靠自己挖到《我的世界》里的钻石。</summary>
  </entry>
  <entry>
    <title>№ 147 · Mastering Atari with Discrete World Models</title>
    <link href="https://estelledc.github.io/embodied-ai-reading-station/papers/dreamer-v2/"/>
    <id>https://estelledc.github.io/embodied-ai-reading-station/papers/dreamer-v2/</id>
    <updated>2026-06-02T11:03:35.224Z</updated>
    <category term="World Model &amp; Video Policy"/>
    <summary>让 AI 闭眼&quot;做白日梦&quot;练打老游戏，第一次只靠脑子里想象就打到人类水平。</summary>
  </entry>
</feed>
