SciGroveBeta
Robotics

Dexora: Open-source VLA for High-DoF Bimanual Dexterity

Zongzheng Zhang, Jingrui Pang

Featured May 28, 2026

Original Paper

This analysis was generated by SciGrove. Upload your own PDFs or enter a DOI — and get the same AI breakdown on any paper.

Get started

AI-generated analysis — This is SciGrove's AI interpretation of the paper, not peer-reviewed content. Always refer to the original paper.

Simply

A new robot system called Dexora lets robots use two hands with many fingers to do tricky tasks, learning from human examples but ignoring the shaky ones to get really good at complex movements.

In depth
The paper introduces Dexora, the first open-source Vision-Language-Action (VLA) system designed for dual-arm, dual-hand, high-degree-of-freedom (DoF) manipulation. It achieves this by combining a novel hybrid teleoperation pipeline for scalable data collection with a data-quality-aware training recipe that uses an offline discriminator to down-weight noisy demonstrations, leading to robust and generalizable dexterous robot control.

Key Takeaways

  • 1
    Dexora is the first open-source VLA system to natively support dual-arm, dual-hand, high-DoF (36-DoF) manipulation, addressing a critical gap in embodied AI.
  • 2
    A hybrid teleoperation pipeline (exoskeleton for arms, Apple Vision Pro for fingers) and a large, embodiment-matched dataset (100K simulated, 10K real trajectories) enable scalable and accurate data collection for complex skills.
  • 3
    A data-quality-aware training recipe uses an offline discriminator to assign clip-level weights, effectively mitigating the impact of noisy teleoperation data and improving policy learning for dexterity.

Conceptual Flow

HIGH LEVEL
1
Methodology: Learning Dexterous Bimanual Control

The system collects human movements using special gear, then trains a robot brain to copy these movements, making sure to only learn from the good examples.

Human Operator
Robot Arms & Hands
Collect Good Examples
Robot Brain Learns
Smooth Robot Actions
2
Results: High-DoF Performance and Generalization

The robot can now do very complex tasks with two hands better than before, and it can even use its learned skills on simpler robots.

Complex Tasks
Simple Robots
Perform & Adapt
High Success Rate
Works on Many Robots