aiqbench analysis
DexMimic advances robotics by introducing a generalist policy that learns diverse hand-object manipulation skills from human interaction data, overcoming the limitations of specialist trajectory-based training. The core innovation lies in its Value-Guided Reset Allocation mechanism, which dynamically optimizes initialization states to improve tracking efficiency and enable zero-shot generalization to unseen objects. This work represents a significant step toward scalable, adaptable dexterous robots capable of handling complex real-world tasks.
DexMimic: A Generalist Dexterous Hand-Object Tracking Policy via Value-Guided Reset Allocation
arXiv:2610.19089v1 Announce Type: new Abstract: Achieving human-like dexterous manipulation remains a central challenge in robotics. 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. We introduce DexMimic, a generalist dexterous hand-object tracking policy jointly trained across diverse HOI trajectories. Efficient joint training requires allocating each rollout to both an undertrained phase and a starting state that supports 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. We further introduce a benchmark spanning training, fine-tuning, and generalization. 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. Finally, we validate DexMimic on a real-world dexterous manipulation platform.
Abstract:Achieving human-like dexterous manipulation remains a central challenge in robotics. 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. We introduce DexMimic, a generalist dexterous hand-object tracking policy jointly trained across diverse HOI trajectories. Efficient joint training requires allocating each rollout to both an undertrained phase and a starting state that supports 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. We further introduce a benchmark spanning training, fine-tuning, and generalization. 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. Finally, we validate DexMimic on a real-world dexterous manipulation platform.
13 pages, including 4 pages of supplementary material. Project page: this https URL
Robotics (cs.RO); Artificial Intelligence (cs.AI)
arXiv:2610.19089 [cs.RO]
(or arXiv:2610.19089v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2610.19089 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 259 words
Related coverage
InterMimicGen: Scaling Humanoid Loco-Manipulation through Self-Evolving Motion Imitation
arXiv · Oct 6, 2026
X-Reset: Scaling Object-Centric Reinforcement Learning via Cross-Embodiment Resets
arXiv · Sep 30, 2026
DexWeave: Learning Dexterous Humanoid Loco-Manipulation from Human Demonstrations
arXiv · Sep 29, 2026
TopoRetarget: Interaction-Preserving Retargeting for Dexterous Manipulation
arXiv · Jun 16, 2026