This is why I can't really trust LLMs for data tasks like this. I can't really be certain that it won't be hallucinating results