What happens when you let LLMs play guessing games about professions?

Our multi-agent experiment shows how differently language models think as a team.