Read the paper · Research page

Official title: Precise Manipulation with Efficient Online RL

Physical Intelligence’s RL-token approach for post-training VLAs from small amounts of real-world online experience.