Cross-Hand Latent Representation for Vision-Language-Action Models

TL;DR

This paper introduces XL-VLA, a scalable framework for vision-language-action learning in dexterous robots that uses a shared latent action space to enable cross-embodiment training and improve manipulation performance.

Contribution

The paper proposes a novel embodiment-invariant latent action space integrated into VLA models, facilitating scalable cross-embodiment learning for dexterous manipulation.

Findings

XL-VLA outperforms baseline models in manipulation tasks.

The shared latent space enables efficient transfer across different robotic hands.

Experimental results validate the effectiveness of the proposed framework.

Abstract

Dexterous manipulation is essential for real-world robot autonomy, mirroring the central role of human hand coordination in daily activity. Humans rely on rich multimodal perception--vision, sound, and language-guided intent--to perform dexterous actions, motivating vision-based, language-conditioned manipulation systems for robots. However, training reliable vision-language-action (VLA) models for dexterous manipulation requires large-scale demonstrations across many robotic hands. In addition, as new dexterous embodiments appear rapidly, collecting data for each becomes costly and impractical, creating a need for scalable cross-embodiment learning. We introduce XL-VLA, a vision-language-action framework integrated with a unified latent action space shared across diverse dexterous hands. This embodiment-invariant latent space is directly pluggable into standard VLA architectures, enabling seamless cross-embodiment training and efficient reuse of both existing and newly collected data. Experimental results demonstrate that XL-VLA consistently outperforms baseline VLA models operating in raw joint spaces, establishing it as an effective solution for scalable cross-embodiment dexterous manipulation.