Evaluating hallucination in large language models is a complex task.