Towards Socially and Morally Aware RL agent: Reward Design With LLM

Zhaoyue Wang

When we design and deploy an Reinforcement Learning (RL) agent, reward functions motivates agents to achieve an objective. An incorrect or incomplete specification of the objective…