CrossBFM distills a shared latent behavior space across humanoid robots in under one GPU-hour
Jan_R_Peters · x · 2026-10-01
- New work CrossBFM (Tan-Dzung Do, Nico Bohlinger, Jan Peters, et al.) treats the behavior latent space itself as the transferable asset across humanoid embodiments.
- Existing Behavior Foundation Models use Forward–Backward representations to give one robot a promptable policy (a latent vector specifying motion imitation, goal reaching, or reward maximization), but cost hundreds of GPU-hours per robot and yield incompatible latent spaces across robots.
- CrossBFM exploits retargeting-induced frame-level correspondence, turning latent transfer into supervised regression with no simulator or training on the target side. A unified encoder with no robot-specific parameters serves all embodiments in under 1 GPU-hour, and latent-conditioned trackers produce whole-body control via standard PPO in 10 more GPU-hours.
- Validated on three humanoids: motion tracking within 0.025 rad of joint-conditioned policies, smooth fall-free goal reaching, and transferable reward optimization. Code coming soon.
Related event: CrossBFM distills shared behavior latent space across humanoid robots(2 posts)→
More from Embodied
- 8th NeurIPS Robot Learning Workshop to debate: Is Physical AI truly going zero-shot? — m_wulfmeier · 2026-11-17
- VR human demos fix persistent failures in point-cloud grasping policies — hcltech-robotics · 2026-11-17
- Ultrasound plus a smartwatch and a lightweight AI model enables full-body tracking — rschu · 2026-11-17
- DeepMind's Raia Hadsell on transferring skills between robots at RAAIS — nathanbenaich · 2026-11-17
- Ropedia-Xperience-2 dataset: 1,500+ 3D scene scans and 131 hours of egocentric video — liuziwei7 · 2026-11-17
- Niantic's Object Interaction Makes Gaussian Splat Scenes Editable and Physics-Ready — lukas_m_ziegler · 2026-11-17