Zero-WAM: In-Context World-Action Modeling for Robotic Task Generalization

Robbyant-Research · hf · 2026-08-28

Zero-WAM enables robotic manipulation of unseen tasks by conditioning a causal video-action model on in-context human video guidance. It is supported by an automatically generated dataset and a future-chunk prediction objective.

Original post →

More from Embodied

Embodied channel →