Being-H0.5: Scaling Human-Centric Robot Learning for Cross-Embodiment Generalization

TL;DR

Being-H0.5 is a novel vision-language-action model that enables robots with different morphologies to learn from human data and generalize across diverse platforms using a unified action space and advanced modeling techniques.

Contribution

The paper introduces a human-centric learning paradigm, a large-scale multimodal dataset, and a Mixture-of-Transformers architecture with novel gating mechanisms for cross-embodiment robot generalization.

Findings

Achieves state-of-the-art on LIBERO and RoboCasa benchmarks.

Demonstrates strong cross-embodiment generalization on five robotic platforms.

Introduces a unified action space and novel gating techniques for robustness.

Abstract

We introduce Being-H0.5, a foundational Vision-Language-Action (VLA) model designed for robust cross-embodiment generalization across diverse robotic platforms. While existing VLAs often struggle with morphological heterogeneity and data scarcity, we propose a human-centric learning paradigm that treats human interaction traces as a universal "mother tongue" for physical interaction. To support this, we present UniHand-2.0, the largest embodied pre-training recipe to date, comprising over 35,000 hours of multimodal data across 30 distinct robotic embodiments. Our approach introduces a Unified Action Space that maps heterogeneous robot controls into semantically aligned slots, enabling low-resource robots to bootstrap skills from human data and high-resource platforms. Built upon this human-centric foundation, we design a unified sequential modeling and multi-task pre-training paradigm to bridge human demonstrations and robotic execution. Architecturally, Being-H0.5 utilizes a Mixture-of-Transformers design featuring a novel Mixture-of-Flow (MoF) framework to decouple shared motor primitives from specialized embodiment-specific experts. Finally, to make cross-embodiment policies stable in the real world, we introduce Manifold-Preserving Gating for robustness under sensory shift and Universal Async Chunking to universalize chunked control across embodiments with different latency and control profiles. We empirically demonstrate that Being-H0.5 achieves state-of-the-art results on simulated benchmarks, such as LIBERO (98.9%) and RoboCasa (53.9%), while also exhibiting strong cross-embodiment capabilities on five robotic platforms.