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

2026-05-18

AI Summary

Existing vision-language-action (VLA) systems struggle to support high-degree-of-freedom, dexterous bimanual manipulation and lack open-source, end-to-end frameworks. This work proposes the first open-source VLA system tailored for such tasks, leveraging a hybrid teleoperation setup that combines an exoskeleton backpack with the Apple Vision Pro, alongside a MuJoCo-based digital twin environment to collect large-scale real-world and simulated data. The system introduces a discriminator-based, data quality-aware weighting mechanism to optimize training of diffusion Transformer policies. Evaluated on dexterous manipulation tasks, it achieves an average success rate of 66.7%—a 15% improvement over the baseline—and 90% on basic tasks, while demonstrating strong out-of-distribution generalization and cross-embodiment transfer capabilities.

Abstract

Vision-Language-Action (VLA) models have recently become a central direction in embodied AI, but current systems are restricted to either dual-gripper control or single-arm dexterous hand manipulation. While low-dimensional gripper control can often be handled with simpler methods, high-dimensional dexterous hand control benefits greatly from full end-to-end VLA learning. In this work, we introduce Dexora, the first open-source VLA system that natively targets dual-arm, dual-hand high-DoF manipulation. We design a hybrid teleoperation pipeline that decouples gross arm kinematics (captured with a custom exoskeleton backpack) from fine finger motion (markerless hand tracking via Apple Vision Pro), and that drives both a physical dual-arm dual-hand platform and an identical MuJoCo digital twin. Using that interface, we assemble a large training corpus: an embodiment-matched synthetic corpus (100K simulated trajectories, 6.5M frames) and a real-world dataset of 10K teleoperated episodes (2.92M frames). To mitigate noisy teleoperation demonstrations, we propose a data-quality-aware training recipe: an offline discriminator provides clip-level weights for diffusion-transformer policy training, down-weighting low-quality demonstrations. Empirically, Dexora outperforms competitive VLA baselines on both basic and dexterous benchmarks (e.g., average dexterous success 66.7% vs. 51.7%), attains 90% success on basic tasks, and shows robust out-of-distribution and cross-embodiment generalization. Ablations confirm the importance of real data and the discriminator for dexterity.

Problem

• Vision-Language-Action
• dexterous manipulation
• bimanual control
• high-DoF
• embodied AI

Innovation

• Vision-Language-Action (VLA)
• bimanual dexterous manipulation
• hybrid teleoperation
• data-quality-aware training
• digital twin