Search papers, labs, and topics across Lattice.
This paper introduces Zero-WAM, a causal video-action model that enables robots to generalize across unseen manipulation tasks by leveraging human video guidance as task specifications. By creating the HumanGen dataset, which pairs 74.2K human-robot ICL pairs across 8.6K tasks, the authors address the challenge of data scarcity in training models for zero-shot cross-task generalization. Zero-WAM demonstrates significant improvements in task execution, achieving a 47.0% success rate on seven unseen tasks, outperforming existing video-action baselines by 29.5 percentage points.
Robots can now learn to execute unseen tasks by simply watching human videos, achieving unprecedented generalization in manipulation.
Zero-shot cross-task generalization, where a policy must execute manipulation tasks never seen during training, remains a central challenge in robot learning. In large language models, a novel task can be performed simply by specifying it in the context, without any parameter update. This form of in-context learning (ICL) turns generalization into a problem of task specification. To achieve cross-task generalization, we bring this paradigm to robotic manipulation, and argue that the natural task specification for manipulation is a human video: unlike language, it provides rich visual cues about the intended task evolution. We present Zero-WAM, a causal video-action model that executes unseen tasks by following in-context human video guidance. To address the scarcity of task-rich paired human-robot data, we propose an automatic pipeline that converts task-sampled robot trajectories into semantically matched human videos, yielding HumanGen, a dataset of 74.2K human-robot ICL pairs across 8.6K tasks. For model training, we further introduce an in-context future chunk prediction (IFP) objective that suppresses shortcuts learned from seen tasks and forces the policy to draw task information from the video prompt. On seven unseen tasks in RoboTwin 2.0 simulation, Zero-WAM achieves a 47.0% average success rate, an absolute improvement of 29.5 percentage points over the strongest video-action baseline. In real-world evaluations, it follows human video guidance to generalize to unseen task configurations involving multi-object scenes, long-horizon manipulation, and fine-grained insertion.
