$\Psi_0$: An Open Foundation Model Towards Universal Humanoid Loco-Manipulation

TL;DR

$ ext{Psi}_0$ is an open humanoid foundation model trained through staged learning on high-quality human videos and robot data, achieving superior loco-manipulation performance with less data than previous methods.

Contribution

The paper introduces $ ext{Psi}_0$, a novel staged training paradigm that decouples learning from heterogeneous data sources for humanoid loco-manipulation.

Findings

Achieves over 40 ext% improvement in success rate over baselines.

Uses only 800 hours of human videos and 30 hours of robot data.

Outperforms models trained on ten times more data.

Abstract

We introduce $\Psi_0$ (Psi-Zero), an open foundation model to address challenging humanoid loco-manipulation tasks. While existing approaches often attempt to address this fundamental problem by co-training on large and diverse human and humanoid data, we argue that this strategy is suboptimal due to the fundamental kinematic and motion disparities between humans and humanoid robots. Therefore, data efficiency and model performance remain unsatisfactory despite the considerable data volume. To address this challenge, \ours\;decouples the learning process to maximize the utility of heterogeneous data sources. Specifically, we propose a staged training paradigm with different learning objectives: First, we autoregressively pre-train a VLM backbone on large-scale egocentric human videos to acquire generalizable visual-action representations. Then, we post-train a flow-based action expert on high-quality humanoid robot data to learn precise robot joint control. Our research further identifies a critical yet often overlooked data recipe: in contrast to approaches that scale with noisy Internet clips or heterogeneous cross-embodiment robot datasets, we demonstrate that pre-training on high-quality egocentric human manipulation data followed by post-training on domain-specific real-world humanoid trajectories yields superior performance. Extensive real-world experiments demonstrate that \ours\ achieves the best performance using only about 800 hours of human video data and 30 hours of real-world robot data, outperforming baselines pre-trained on more than 10$\times$ as much data by over 40\% in overall success rate across multiple tasks. We will open-source the entire ecosystem to the community, including a data processing and training pipeline, a humanoid foundation model, and a real-time action inference engine.