CLST: Cold-Start Mitigation in Knowledge Tracing by Aligning a Generative Language Model as a Students' Knowledge Tracer

Generative large language models (LLMs) are widely utilized in education to assist students to learn and instructors to teach. In addition, generative LLMs are used to support personalized learning by recommending learning content in intelligent tutoring systems (ITSs). Nonetheless, there are few studies utilizing generative LLMs in the field of knowledge tracing (KT), which is a key component of ITSs. KT, which uses students' problem-solving histories to estimate their current levels of knowledge, is regarded as a key technology for personalized learning. Nevertheless, most existing KT models are characterized by their development with an ID-based paradigm, which results in a low performance in cold-start scenarios. These limitations can be mitigated by leveraging the vast quantities of external knowledge possessed by generative LLMs. In this study, we propose cold-start mitigation in knowledge tracing by aligning a generative language model as a students’ knowledge tracer (CLST) as a framework that utilizes a generative LLM as a knowledge tracer. Upon collecting data from math, social studies, and science subjects, we framed the KT task as a natural language processing task, wherein problem-solving data are expressed in natural language, and fine-tuned the generative LLM using the formatted KT dataset. Subsequently, we evaluated the performance of the CLST in situations of data scarcity using various baseline models for comparison. The results indicate that the CLST significantly enhanced performance with a dataset of fewer than 100 students in terms of prediction, reliability, and cross-domain generalization.