The fact that this is so obvious to humans is why there's no training data that LLMs can use to know the answer.