Here are some examples of the questions in the benchmark. If these are representative, they seem pretty cut and dry. https://artificialanalysis.ai/evaluations/omniscience#exampl...