DexMimic: A Generalist Dexterous Hand-Object Tracking Policy via Value-Guided Reset Allocation

What happened

arXiv:2610.19089v1 Announce Type: new Abstract: Achieving human-like dexterous manipulation remains a central challenge in robotics. Experiments show improved tracking success and training efficiency, faster fine-tuning, and the potential for zero-shot generalization to unseen trajectories and objects; larger training sets further improve efficiency and generalization.

Learning from human hand-object interaction (HOI) datasets offers a promising pathway to diverse dexterous skills, yet existing physics-based approaches train specialist policies for individual trajectories, thereby limiting skill reuse at scale. Efficient joint training requires allocating each rollout to both an undertrained phase and a starting state that provides support for continued tracking. DexMimic realizes this insight through Value-Guided Reset Allocation, which adapts the initialization distribution online to prioritize difficult phases and select reset states with higher expected tracking returns.

Key facts

• Efficient joint training — requires: allocating each rollout to both an undertrained phase and a starting state that supports continued tracking