The paper compares the predictive performance of three knowledge-tracing models—Deep Knowledge Tracing (DKT), Dynamic Key-Value Memory Network (DKVMN), and Deep Item Response Theory (Deep-IRT)—on the benchmark datasets. As summarized in the accompanying results table, the three models achieve relatively similar prediction accuracy, indicating that Deep-IRT matches the performance of DKT and DKVMN while additionally providing interpretable estimates of student ability and item difficulty.