Google: Towards 3D Foundational Robotics Models
Abstract
Our results show that hybrid linear attention enables efficient scaling, strong reasoning, and robust long-context capability, providing a practical path toward next-generation trillion-scale foundation models.
Successful Page Load