# SkeLAM: A Robot World Model Using Skeleton-Supervised Latent Actions for Zero-Shot Cross-Embodiment Transfer

- Source: 机器之心 · 微信公众号
- Published: 2026-10-27T00:34:41.000Z
- RoboHot: https://robohot.news/en/items/wknfxo08euwoqvp1xjsbs4gh2
- Original: https://mp.weixin.qq.com/s?__biz=MzA3MzI4MjgzMw==&mid=2651070362&idx=3&sn=9c21d8c66fbf4a058e437690509a24b1&chksm=85984a8bc3c27a61583e32f450f9b1dae822f567cb095c2d6848eb91db91680902c954fd485c&scene=126&sessionid=1793109700#rd

## Summary

Harbin Institute of Technology (Shenzhen), the National University of Singapore, and other institutions propose SkeLAM, built on Cosmos-Predict2.5-2B and trained on about 10 hours of robot interaction data, using robot skeletons to supervise latent actions in video.