Questioning Internal Knowledge Structure of Large Language Models Through the Lens of the Olympic Games