—Suitably complex exercises motivate students and facilitates a deeper understanding. Manually constructing suchproblems consumes time that teachers can otherwise use to mentor students. Many software tools and services for automaticgeneration of mathematics problems are available on the web, but they provide only materials up to high school level. In addition, no standardized methods are provided to evaluate and control thecomputational complexity of generated problems. In this paper, we proposed a framework for evaluating computational complex-ity from the learners’ perspective, aiming to apply our framework to the automatic generation of college-level mathematics problems with controlled computational complexity. Our framework helpsteachers prepare learning materials and thereby save time for mentoring students.