<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Teacher Humanoid: latest updates</title>
    <link>https://teacherhumanoid.com/updates</link>
    <description>Education robotics now draws on open-source software, shared datasets and large contests. These figures and dated updates show what students and labs can use today and what is still maturing.</description>
    <language>en</language>
    <lastBuildDate>Sun, 04 Oct 2026 12:00:00 GMT</lastBuildDate>
    <atom:link href="https://teacherhumanoid.com/feed.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>World Robot Conference contests draw 11,418 teams from 26 countries in Beijing</title>
      <link>https://teacherhumanoid.com/updates#2026-08-25-world-robot-conference-contests-draw-11-418-teams-from-26-countries-i</link>
      <guid isPermaLink="false">teacherhumanoid.com:2026-08-25-world-robot-conference-contests-draw-11-418-teams-from-26-countries-i</guid>
      <pubDate>Tue, 25 Aug 2026 12:00:00 GMT</pubDate>
      <category>Milestone</category>
      <description>Beijing&apos;s 2026 World Robot Conference, held August 19 to 23, reported 11,418 teams and 17,802 competitors from 26 countries in its on-site competitions. The event also drew 557,000 visitors and showed more than 3,000 robot-related products, according to the city government&apos;s closing report. Why it matters: Large contests show how widely robot building has spread. Teachers entering teams should check each event&apos;s rules, kit limits and travel costs early, and plan practice robots well before the season. Source: Beijing Municipal Government (https://english.beijing.gov.cn/beijinginfo/sci/latesttrends/202608/t20260825_4836401.html)</description>
    </item>
    <item>
      <title>Survey maps how far robot foundation models are from running on a specific robot</title>
      <link>https://teacherhumanoid.com/updates#2026-08-19-survey-maps-how-far-robot-foundation-models-are-from-running-on-a-spe</link>
      <guid isPermaLink="false">teacherhumanoid.com:2026-08-19-survey-maps-how-far-robot-foundation-models-are-from-running-on-a-spe</guid>
      <pubDate>Wed, 19 Aug 2026 12:00:00 GMT</pubDate>
      <category>Research</category>
      <description>Researchers surveyed 21 groups of robot-learning methods, including RT-X, Octo, OpenVLA and π0, and mapped how much each shares across robot bodies against how much work remains to run it on one robot. Three researchers coded the methods independently, and none was placed in the safety and recovery row. Why it matters: Success rates alone hide how much new data or tuning a model needed. When a lab or class adopts an open model, budget time to adapt it to the exact robot and ask what was collected to reach published results. Source: arXiv (https://arxiv.org/pdf/2608.18433)</description>
    </item>
    <item>
      <title>China publishes a 2026 standard system for humanoid robots and embodied AI</title>
      <link>https://teacherhumanoid.com/updates#2026-08-04-china-publishes-a-2026-standard-system-for-humanoid-robots-and-embodi</link>
      <guid isPermaLink="false">teacherhumanoid.com:2026-08-04-china-publishes-a-2026-standard-system-for-humanoid-robots-and-embodi</guid>
      <pubDate>Tue, 04 Aug 2026 12:00:00 GMT</pubDate>
      <category>Policy</category>
      <description>A preview of Beijing&apos;s 2026 World Robot Conference reports that China has named embodied intelligence a core direction for future industries and that its industry ministry has released the Humanoid Robot and Embodied Intelligence Standard System (2026 Edition). It describes the sector moving from technical testing toward large-scale factory deployment. Why it matters: Standards shape how humanoids are tested, connected and certified. A university building a humanoid lab should ask each maker which standards its platform follows, and keep the answers with the purchase record. Source: Tiger Brokers news feed (https://www.itiger.com/news/1108805531)</description>
    </item>
    <item>
      <title>Hugging Face&apos;s LeRobot 0.6 adds world-model policies and six simulation benchmarks</title>
      <link>https://teacherhumanoid.com/updates#2026-07-07-hugging-face-s-lerobot-0-6-adds-world-model-policies-and-six-simulati</link>
      <guid isPermaLink="false">teacherhumanoid.com:2026-07-07-hugging-face-s-lerobot-0-6-adds-world-model-policies-and-six-simulati</guid>
      <pubDate>Tue, 07 Jul 2026 12:00:00 GMT</pubDate>
      <category>Launch</category>
      <description>Hugging Face released LeRobot 0.6.0, its open-source robot-learning toolkit. The update adds world-model policies such as VLA-JEPA, more vision-language-action models, reward-model support, human-in-the-loop correction tools, cloud training through Hugging Face Jobs and six simulation benchmarks in its evaluation command. Why it matters: Shared evaluation code lets student teams compare methods the same way and reproduce published work. Check which robots each policy supports and the computing power needed to train before planning a unit around it. Source: Let&apos;s Data Science (https://letsdatascience.com/news/hugging-face-releases-lerobot-06-robotics-toolkit-908e3ba5)</description>
    </item>
    <item>
      <title>Ai2 releases MolmoAct 2 with open weights, code and a 720-hour robot dataset</title>
      <link>https://teacherhumanoid.com/updates#2026-05-05-ai2-releases-molmoact-2-with-open-weights-code-and-a-720-hour-robot-d</link>
      <guid isPermaLink="false">teacherhumanoid.com:2026-05-05-ai2-releases-molmoact-2-with-open-weights-code-and-a-720-hour-robot-d</guid>
      <pubDate>Tue, 05 May 2026 12:00:00 GMT</pubDate>
      <category>Research</category>
      <description>The Allen Institute for AI released MolmoAct 2, an open robot model that reasons about 3D space before acting. It shares model weights, training code and data, including more than 720 hours of two-arm tabletop demonstrations that Ai2 calls the largest open dataset of its kind. Why it matters: Open weights, code and data let a university lab reproduce and adapt the work on its own hardware. Check the license and the robots in the data, then compare your arm against them using Spec Cards on The Humanoid Group. Source: Allen Institute for AI (https://allenai.org/blog/molmoact2)</description>
    </item>
    <item>
      <title>NVIDIA opens early access to Isaac Lab 3.0 and GR00T N1.7 robot tools</title>
      <link>https://teacherhumanoid.com/updates#2026-03-16-nvidia-opens-early-access-to-isaac-lab-3-0-and-gr00t-n1-7-robot-tools</link>
      <guid isPermaLink="false">teacherhumanoid.com:2026-03-16-nvidia-opens-early-access-to-isaac-lab-3-0-and-gr00t-n1-7-robot-tools</guid>
      <pubDate>Mon, 16 Mar 2026 12:00:00 GMT</pubDate>
      <category>Launch</category>
      <description>At GTC 2026, NVIDIA announced Isaac Lab 3.0 in early access with the Newton 1.0 physics engine, GR00T N1.7 in early access with commercial licensing, and the Cosmos 3 world model. It also previewed GR00T N2, which it says will be released by the end of 2026. Why it matters: Simulators and ready-made robot models are what university labs build on, but early access comes with limits. Confirm what is free for teaching, which graphics cards you need, and which robots each tool supports. Source: NVIDIA Newsroom (https://nvidianews.nvidia.com/news/nvidia-and-global-robotics-leaders-take-physical-ai-to-the-real-world)</description>
    </item>
  </channel>
</rss>
