Physical Intelligence has introduced its latest VLA model, π*0.6, which employs an innovative training technique known as "Recap." This method integrates human demonstrations with autonomous reinforcement learning, significantly enhancing the model's capabilities. The company reports that this advancement has resulted in a doubling of performance throughput for intricate tasks, such as laundry folding and espresso preparation. The announcement marks a notable milestone in the development of intelligent automation technologies.
Leave a comment