Humanoid World Models: Open World Foundation Models for Humanoid Robotics

TL;DR

Humanoid World Models (HWM) are lightweight, open-source predictive models trained on humanoid demonstrations, enabling robots to reason and plan in complex open-world environments with reduced computational resources.

Contribution

We introduce HWM, a family of efficient, open-source models for humanoid robots that predict future egocentric video, with architectural variants and parameter-sharing techniques to reduce size.

Findings

Parameter-sharing reduces model size by up to 53%.

Models trained on 100 hours of data perform well in open-world scenarios.

HWMs are practical for small-lab settings with limited computational resources.

Abstract

Humanoid robots, with their human-like form, are uniquely suited for interacting in environments built for people. However, enabling humanoids to reason, plan, and act in complex open-world settings remains a challenge. World models, models that predict the future outcome of a given action, can support these capabilities by serving as a dynamics model in long-horizon planning and generating synthetic data for policy learning. We introduce Humanoid World Models (HWM), a family of lightweight, open-source models that forecast future egocentric video conditioned on humanoid control tokens. We train two types of generative models, Masked Transformers and Flow-Matching, on 100 hours of humanoid demonstrations. Additionally, we explore architectural variants with different attention mechanisms and parameter-sharing strategies. Our parameter-sharing techniques reduce model size by 33-53% with minimal impact on performance or visual fidelity. HWMs are designed to be trained and deployed in practical academic and small-lab settings, such as 1-2 GPUs.