$Ψ_0$: An Open Foundation Model Towards Universal Humanoid Loco-Manipulation

• URL: http://arxiv.org/abs/2603.12263v1
• Date: Thu, 12 Mar 2026 17:59:51 GMT
• Title: $Ψ_0$: An Open Foundation Model Towards Universal Humanoid Loco-Manipulation
• Abstract summary: We introduce an open foundation model to address challenging humanoid loco-manipulation tasks.<n>Our research identifies a critical yet often overlooked data recipe.<n>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.
• Score: 39.811210435945924
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: We introduce $Ψ_0$ (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.

Related papers

• HumanScale: Egocentric Human Video Can Outperform Real-Robot Data for Embodied Pretraining [93.5192770515694]
  Embodied foundation models are expected to benefit from data scaling like large language models, but face a much tighter data bottleneck.<n>Teleoperated real-robot trajectories remain the dominant pretraining source due to their precise action supervision and embodiment alignment.<n>Embodied foundation models pretrained on egocentric data achieve a 24% lower validation loss on real-robot action prediction.
  arXiv  Detail & Related papers  (2026-06-18T17:37:34Z)
• HumanoidMimicGen: Data Generation for Loco-Manipulation via Whole-Body Planning [52.022681285103126]
  We present HumanoidMimicGen, a method for generating humanoid legged loco-manipulation data.<n>Our method adapts contact-rich whole-body skills from a handful of source demonstrations to new states.<n>We show that whole-body visuomotor policies co-trained with data generated by HumanoidMimicGen outperform those trained only on real-world data by 20%.
  arXiv  Detail & Related papers  (2026-05-26T21:57:11Z)
• EgoVerse: An Egocentric Human Dataset for Robot Learning from Around the World [63.2070298441355]
  EgoVerse is a collaborative platform for human data-driven robot learning.<n>We release 1,362 hours (80k episodes) of human demonstrations spanning 1,965 tasks, 240 scenes, and 2,087 unique demonstrators.<n>We conduct a large-scale study of human-to-robot transfer with experiments replicated across multiple labs, tasks, and robot embodiments.
  arXiv  Detail & Related papers  (2026-04-08T21:27:06Z)
• EgoHumanoid: Unlocking In-the-Wild Loco-Manipulation with Robot-Free Egocentric Demonstration [67.13034606664333]
  EgoHumanoid is the first framework to co-train a vision-language-action policy using egocentric human demonstrations.<n>A portable system for scalable human data collection is developed.
  arXiv  Detail & Related papers  (2026-02-10T18:59:03Z)
• Sim-and-Human Co-training for Data-Efficient and Generalizable Robotic Manipulation [113.13282853889818]
  SimHum is a framework to simultaneously extract kinematic prior from simulated robot actions and visual prior from real-world human observations.<n>Based on the two complementary priors, we achieve data-efficient and generalizable robotic manipulation in real-world tasks.
  arXiv  Detail & Related papers  (2026-01-27T09:41:28Z)
• Emergence of Human to Robot Transfer in Vision-Language-Action Models [88.76648919814771]
  Vision-language-action (VLA) models can enable broad open world generalization, but require large and diverse datasets.<n>We show that human-to-robot transfer emerges once the VLA is pre-trained on sufficient scenes, tasks, and embodiments.
  arXiv  Detail & Related papers  (2025-12-27T00:13:11Z)
• Humanoid Everyday: A Comprehensive Robotic Dataset for Open-World Humanoid Manipulation [16.701354625940308]
  Humanoid Everyday is a large-scale and diverse humanoid manipulation dataset.<n>It aggregates high-quality multimodal sensory data, including RGB, depth, LiDAR, and tactile inputs, together with natural language annotations.<n>We conduct an analysis of representative policy learning methods on our dataset, providing insights into their strengths and limitations.
  arXiv  Detail & Related papers  (2025-10-09T20:43:27Z)
• CEDex: Cross-Embodiment Dexterous Grasp Generation at Scale from Human-like Contact Representations [53.37721117405022]
  Cross-embodiment dexterous grasp synthesis refers to adaptively generating and optimizing grasps for various robotic hands.<n>We propose CEDex, a novel cross-embodiment dexterous grasp synthesis method at scale.<n>We construct the largest cross-embodiment grasp dataset to date, comprising 500K objects across four types with 20M total grasps.
  arXiv  Detail & Related papers  (2025-09-29T12:08:04Z)
• Humanoid Policy ~ Human Policy [41.34186233320398]
  We train a human-humanoid behavior policy, which we term Human Action Transformer (HAT)<n>The state-action space of HAT is unified for both humans and humanoid robots and can be differentiably retargeted to robot actions.<n>We show that human data improves both generalization and robustness of HAT with significantly better data collection efficiency.
  arXiv  Detail & Related papers  (2025-03-17T17:59:09Z)
• EgoMimic: Scaling Imitation Learning via Egocentric Video [22.902881956495765]
  We present EgoMimic, a full-stack framework which scales manipulation via human embodiment data. EgoMimic achieves this through: (1) a system to capture human embodiment data using the ergonomic Project Aria glasses, (2) a low-cost bimanual manipulator that minimizes the kinematic gap to human data, and (4) an imitation learning architecture that co-trains on human and robot data.
  arXiv  Detail & Related papers  (2024-10-31T17:59:55Z)
• Mitigating the Human-Robot Domain Discrepancy in Visual Pre-training for Robotic Manipulation [16.809190349155525]
  We propose a novel adaptation paradigm that leverages readily available paired human-robot video data to bridge the domain gap.<n>Our method employs a human-robot contrastive alignment loss to align the semantics of human and robot videos, adapting pre-trained models to the robot domain in a parameter-efficient manner.
  arXiv  Detail & Related papers  (2024-06-20T11:57:46Z)
• Learning Human Action Recognition Representations Without Real Humans [66.61527869763819]
  We present a benchmark that leverages real-world videos with humans removed and synthetic data containing virtual humans to pre-train a model. We then evaluate the transferability of the representation learned on this data to a diverse set of downstream action recognition benchmarks. Our approach outperforms previous baselines by up to 5%.
  arXiv  Detail & Related papers  (2023-11-10T18:38:14Z)
• Where is my hand? Deep hand segmentation for visual self-recognition in humanoid robots [129.46920552019247]
  We propose the use of a Convolution Neural Network (CNN) to segment the robot hand from an image in an egocentric view. We fine-tuned the Mask-RCNN network for the specific task of segmenting the hand of the humanoid robot Vizzy.
  arXiv  Detail & Related papers  (2021-02-09T10:34:32Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.