UniCross: Unified Cross-Skill Dexterous Manipulation Synthesis

Hui Zhang1,2, Julian Ferchow2, Jie Song3, Mirko Meboldt1

1ETH Zürich, Switzerland 2inspire AG, Zürich, Switzerland 3HKUST (Guangzhou), China

Overview

We present UniCross, a unified dexterous manipulation framework that synthesizes grasping, relocation, in-hand rotation, and in-hand translation using a shared hand-object relational formulation. This cross-skill formulation enables a single policy to perform individual skills and their seamless composition for long-horizon manipulation. The embodiment-agnostic framework also transfers effectively across different hand morphologies.

Abstract

Many dexterous manipulation tasks require the object to remain securely held throughout the interaction. From the perspective of hand-object relational motion, such manipulation comprises four canonical skills: grasping, relocation, in-hand rotation, and in-hand translation.

Human hands flexibly compose these skills to accomplish complex tasks. Existing approaches, however, model these skills separately with skill-specific action constraints, objectives, or even dedicated hand morphologies, which breaks the compatibility and continuity required for long-horizon composition. In this work, we present a unified framework that models all four skills in a single formulation that shares the same state and action spaces and a common objective structure.

This formulation enables the distillation of a single cross-skill policy that performs strongly on every skill, generalizes to unseen objects, stays robust to disturbances, and chains skills seamlessly into long-horizon manipulation. The framework also transfers effectively across different hand morphologies. Overall, our results suggest that different dexterous manipulation skills can be viewed as instantiations of a shared task formulation, revealing the intrinsic consistency across different behaviors.

Long-term Motion

Long-term motions generated with our single (per-hand) unified policy.

Cross-Skill Dexterous Manipulation

A single unified policy can perform grasping, relocation, in-hand rotation, in-hand translation, and seamless skill composition.

Grasping

Relocation

In-hand Rotation

In-hand Translation

Skill Composition

Generalization Across Hand Morphologies

The same cross-skill formulation transfers to different dexterous hand embodiments.

Allegro Skill Composition

Sharpa Skill Composition

Adaptive Motion Across Object Geometries & Scales

The policy automatically adapts motion patterns and contact transitions to object geometry and scale.

Generalization to Unseen Objects

The unified cross-skill policy preserves effectively coherent interaction patterns on unseen object geometries.

Implementation in Different Simulators

Originally developed in IsaacGym, our framework is also available in IsaacLab and mjlab for easy community use.

IsaacLab

IsaacGym

mjlab

BibTeX

@article{zhang2026unicross,
  title={{UniCross}: Unified Cross-Skill Dexterous Manipulation Synthesis},
  author={Zhang, Hui and Ferchow, Julian and Song, Jie and Meboldt, Mirko},
  journal={arXiv preprint arXiv:2607.28198},
  year={2026}
}