Focal: A Proposed Method of Leveraging LLMs for Automating Assessments

Abstract

In response to the growing need for frequent, high-quality assessments in the expanding field of online learning and the significant time burden their manual creation places on educators, this study proposes Focal, an end-to-end assessment generation pipeline. Focal employs large language models, notably Text-to-Text Transfer Transformers, fine-trained on diverse learning materials, to generate and evaluate pedagogically sound questions and their corresponding answers. The Focal pipeline is designed to integrate with Learning Management Systems, providing educators an automated means of creating assessments that align with their curriculum. This not only eases the task of creating and evaluating assessments but also frees educators to focus on other crucial responsibilities. The system is domain agnostic and its efficacy is continually improved by training and evaluating it using data from multiple subject areas. By automating the traditionally labor-intensive process of assessment production, Focal aims to increase efficiency in online education and enhance the learning experience for students.