Tag
The paper proposes EvalXRL, a benchmark for evaluating Explainable Reinforcement Learning methods by using an LLM coding agent to diagnose and fix bugs in RL agents.