Meta-Behavioral Pattern Framework Enhances Knowledge Tracing

A new framework called MBP-KT, introduced in arXiv:2605.08697, improves knowledge tracing by extracting global collaborative information from learners' meta-behavioral patterns. The method transforms raw interaction sequences into combinations of meta-behavioral patterns to better capture learning behaviors. A parameter-free module then extracts global collaborative representations to enhance prediction of learner knowledge states. The approach addresses limitations of existing methods that rely on raw sequences and tailored modules, offering a generalizable solution.