Train decision-making systems through feedback from actions and outcomes within an interactive environment.
Train specialized small models through an agent-driven workflow