Automatically creating evaluation criteria tailored to individual questions or tasks rather than using fixed, pre-written rubrics.