Tag
This paper presents a framework for training reinforcement learning agents that can be coached in real-time to adopt different styles while performing core tasks, demonstrated in Horizon Forbidden West, Gran Turismo, and a humanoid walking domain.