3/3 📊 KT² beats all baselines across six simulated low-resource classroom settings on two public datasets (XES3G5M and MOOCRadar). - Consistently top-1 in AUC, ACC, and F1 - Outperforms traditional deep learning-based KT methods (e.g., AKT, qDKT) as well as prompting-based LLMs

2/3 We incorporate the information from knowledge concept trees. We use a Bayesian probabilistic model that jointly models how well a student understands each knowledge concept, and how likely they are to answer related questions correctly. It builds an initial profile with a few

1/3 How can we reliably trace a student’s understanding with just a few exercise responses? In classrooms, it’s important for teachers to track students’ understanding and predict their future performance on exercises. However, existing knowledge tracing (KT) algorithms often