Knowledge Tracing (KT) is critical in Intelligent Tutoring Systems (ITS) for monitoring learner progress, yet many existing models fail to provide clarity on the actual knowledge state. AlignKT addresses this gap by aligning the initial knowledge state with a pedagogically-informed ideal, yielding improved performance in real-world datasets.
This model not only integrates a robust frontend-to-backend architecture but also employs a contrastive learning module to strengthen the alignment process. Experiments reveal that AlignKT surpasses seven KT baseline models, achieving state-of-the-art results, which signifies its potential to enhance the efficacy of instructional support provided by ITS. As educational technology evolves, effective knowledge tracing could ultimately transform personalized learning experiences.
👉 Pročitaj original: arXiv AI Papers
