One-Policy-Fits-All (OPFA) learns a single manipulation policy across 11 different embodiments, including grippers and dexterous hands, using geometry-aware action latents.
arXiv · March 17, 2026 · 2603.14522
The Takeaway
It allows end-to-end co-training of data from radically different robots without embodiment-specific tuning. This enables 'foundation policies' for robotics where data from one robot type can directly improve the performance of another with as few as 8 demonstrations.
From the abstract
Cross-embodiment manipulation is crucial for enhancing the scalability of robot manipulation and reducing the high cost of data collection. However, the significant differences between embodiments, such as variations in action spaces and structural disparities, pose challenges for joint training across multiple sources of data. To address this, we propose One-Policy-Fits-All (OPFA), a framework that enables learning a single, versatile policy across multiple embodiments. We first learn a Geometr