HEX: Humanoid-Aligned Experts for Cross-Embodiment Whole-Body Manipulation

TL;DR

HEX is a novel framework enabling coordinated whole-body manipulation in humanoid robots by integrating a universal state representation, a mixture-of-experts predictor, and visual-language cues, leading to superior real-world performance.

Contribution

The paper introduces HEX, a state-centric, scalable approach for humanoid manipulation that models whole-body coordination and integrates visual-language cues effectively.

Findings

HEX achieves state-of-the-art success rates in real-world humanoid tasks.

The framework generalizes well to new tasks and fast-reaction scenarios.

Lightweight history tokens improve temporal context understanding without heavy computation.

Abstract

Humans achieve complex manipulation through coordinated whole-body control, whereas most Vision-Language-Action (VLA) models treat robot body parts largely independently, making high-DoF humanoid control challenging and often unstable. We present HEX, a state-centric framework for coordinated manipulation on full-sized bipedal humanoid robots. HEX introduces a humanoid-aligned universal state representation for scalable learning across heterogeneous embodiments, and incorporates a Mixture-of-Experts Unified Proprioceptive Predictor to model whole-body coordination and temporal motion dynamics from large-scale multi-embodiment trajectory data. To efficiently capture temporal visual context, HEX uses lightweight history tokens to summarize past observations, avoiding repeated encoding of historical images during inference. It further employs a residual-gated fusion mechanism with a flow-matching action head to adaptively integrate visual-language cues with proprioceptive dynamics for action generation. Experiments on real-world humanoid manipulation tasks show that HEX achieves state-of-the-art performance in task success rate and generalization, particularly in fast-reaction and long-horizon scenarios.