Understanding Your Agent: Leveraging Large Language Models for Behavior Explanation