LLMs hallucinate. And they conveniently hallucinate graphs too, which allows for efficient comparisons between them, using simple graph library tools.