Psi R0

Overview

Psi R0 is the first end-to-end VLA embodied model. Leveraging reinforcement learning and a bidirectional training framework, it achieves high generalization and robust multi-skill chaining for complex task execution.

Key features

• ▸First end-to-end VLA embodied model
• ▸Reinforcement learning provides high generalization and robustness
• ▸Bidirectional training framework connects multiple skills
• ▸High-quality real-world data alignment improves long-horizon task success rate
• ▸End-to-end technical architecture with VLM and manipulation model
• ▸Over 99% successful grasping rate with only 20 real-world data points
• ▸Strong adaptability to flexible objects