# Psi0: An Open Foundation Model Towards Universal Humanoid Loco-Manipulation

arXiv: https://arxiv.org/abs/2603.12263 · HTML: https://mlsanity.com/paper/2603.12263/

## Key facts

- Cited by 27 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 12 Mar 2026.

## Built on by (most cited first, 27 of 27)

- [StarVLA: A Lego-like Codebase for Vision-Language-Action Model Developing](https://mlsanity.com/paper/2604.05014/) (2026)
- [MotionWAM: Towards Foundation World Action Models for Real-Time Humanoid Loco-Manipulation](https://mlsanity.com/paper/2606.09215/) (2026)
- [ActiveMimic: Egocentric Video Pretraining with Active Perception](https://mlsanity.com/paper/2606.06194/) (2026)
- [ω-0: A Latent Predictive World Action Model for Concurrent Humanoid Loco-Manipulation](https://mlsanity.com/paper/2608.06375/) (2026)
- [OpenHLM: An Empirical Recipe for Whole-Body Humanoid Loco-Manipulation](https://mlsanity.com/paper/2606.22174/) (2026)
- [Fast SAM 3D Body: Accelerating SAM 3D Body for Real-Time Full-Body Human Mesh Recovery](https://mlsanity.com/paper/2603.15603/) (2026)
- [SceneBot: Contact-Prompted General Humanoid Whole Body Tracking with Scene-Interaction](https://mlsanity.com/paper/2606.27581/) (2026)
- [SUGAR: A Scalable Human-Video-Driven Generalizable Humanoid Loco-Manipulation Learning Framework](https://mlsanity.com/paper/2605.20373/) (2026)
- [VLK: Learning Humanoid Loco-Manipulation from Synthetic Interactions in Reconstructed Scenes](https://mlsanity.com/paper/2606.30645/) (2026)
- [LEGS: Fine-Tuning Teleop-Free VLAs for Humanoid Loco-manipulation in an Embodied Gaussian Splatting World](https://mlsanity.com/paper/2606.01458/) (2026)
- [HANDOFF: Humanoid Agentic Task-Space Whole-Body Control via Distilled Complementary Teachers](https://mlsanity.com/paper/2606.06493/) (2026)
- [VLALeaks: Membership Inference Attacks against Vision-Language-Action Models](https://mlsanity.com/paper/2606.15165/) (2026)
- [CLIFT: Turning Gemini Robotics On-Device into Humanoid Specialists via Non-Invasive Closed-Loop Iterative Fine-Tuning](https://mlsanity.com/paper/2607.29172/) (2026)
- [HAF: Adapting Generalist VLAs to Humanoid Whole-Body Loco-manipulation via Hierarchical Action Flow and Spectral Latent RL](https://mlsanity.com/paper/2608.16837/) (2026)
- [TANGO: Humanoid Navigation in Cluttered Environments with a Whole-Body Vision-Language-Action Model](https://mlsanity.com/paper/2609.09158/) (2026)
- [Weave: Learning Whole-Body Dexterous Loco-Manipulation from Human-Object Interactions](https://mlsanity.com/paper/2609.16683/) (2026)
- [Rolling-WAM: World Action Models with Rolling Imagination](https://mlsanity.com/paper/2609.30247/) (2026)
- [UniMem: Unifying Multimodal Memory and Control for Vision-Language-Action Models](https://mlsanity.com/paper/2608.22869/) (2026)
- [CoHuB: A Simulation Benchmark for Multi-Humanoid Collaboration](https://mlsanity.com/paper/2609.34782/) (2026)
- [Humanoid Loco-Manipulation With Discrete VLA Model](https://mlsanity.com/paper/2609.35709/) (2026)
- [EgoHumanoid-V2: Human-to-Humanoid Transfer of Coordinated Whole-Body Skills for Loco-Manipulation](https://mlsanity.com/paper/2609.37181/) (2026)
- [HumanoidToolBench: Benchmarking Humanoid Tool Use from Selection to Mobile Execution](https://mlsanity.com/paper/2610.02089/) (2026)
- [EgoLAP: Learning from Egocentric Human Data through Language-Action Reasoning](https://mlsanity.com/paper/2610.08726/) (2026)
- [Being-M0.7: A Latent World-Action Model for Humanoid Robots](https://mlsanity.com/paper/2610.11283/) (2026)
- [LIVIN: Benchmarking Spatial and Embodied Intelligence in Digital Twins of Lived-In Homes](https://mlsanity.com/paper/2610.12069/) (2026)
- [VioLA: Learning Generalist Humanoid Control Policies from Human Data](https://mlsanity.com/paper/2610.12435/) (2026)
- [UniLAP: A Unified Latent Action Policy for Cross-Embodiment Manipulation](https://mlsanity.com/paper/2610.16969/) (2026)

## Abstract (arXiv)

We introduce Psi0 (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× 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.