Mostly just questions where you compare responses from two AI models to a single prompt and choose which one is better based on accuracy, relevance and ethical guidelines. A couple pseudocode questions where you prove understanding of basic concepts and a single coding question that I don't remember too well, but was pretty simple.