How well do LLMs reason over tabular data, really?

Open in new window