Cross-Dataset Bloom Question Classification: Supervised Models and Prompted LLMs

The study evaluates the effectiveness of machine learning (ML) and deep learning (DL) models, as well as large language models (LLMs), for classifying assessment questions according to Bloom's taxonomy across different datasets. While traditional supervised ML/DL methods showed significant performance degradation on unseen datasets, LLMs demonstrated greater stability, particularly when using a prompting strategy that included in-context examples and course-specific action verbs. This research highlights the potential of LLMs as a robust alternative for automatic question classification in diverse educational settings, complemented by a user-friendly interface designed to assist instructors.