This was part of New Directions in Reinforcement Learning and Control

Understanding the foundation model pipeline through coverage

Dylan Foster, Microsoft Research