Mamba4KT:An Efficient and Effective Mamba-based Knowledge Tracing Model

TL;DR

Knowledge tracing on increasingly large educational datasets faces time and space constraints in attention-based and RNN approaches. This work introduces Mamba4KT, a knowledge tracing model built on a selective state-space (Mamba) framework with Rasch-model embeddings and FFN blocks to achieve competitive predictive accuracy while greatly reducing training and inference time and memory usage, all while enabling sequence- and exercise-level interpretability. Empirical results across ASSISTments and Eedi demonstrate strong efficiency gains with comparable accuracy to state-of-the-art KT models; ablations underscore the value of Rasch-based embeddings in improving prediction. Overall, Mamba4KT presents a scalable, interpretable KT solution that addresses practical deployment challenges in large-scale educational settings.

Abstract

Knowledge tracing (KT) enhances student learning by leveraging past performance to predict future performance. Current research utilizes models based on attention mechanisms and recurrent neural network structures to capture long-term dependencies and correlations between exercises, aiming to improve model accuracy. Due to the growing amount of data in smart education scenarios, this poses a challenge in terms of time and space consumption for knowledge tracing models. However, existing research often overlooks the efficiency of model training and inference and the constraints of training resources. Recognizing the significance of prioritizing model efficiency and resource usage in knowledge tracing, we introduce Mamba4KT. This novel model is the first to explore enhanced efficiency and resource utilization in knowledge tracing. We also examine the interpretability of the Mamba structure both sequence-level and exercise-level to enhance model interpretability. Experimental findings across three public datasets demonstrate that Mamba4KT achieves comparable prediction accuracy to state-of-the-art models while significantly improving training and inference efficiency and resource utilization. As educational data continues to grow, our work suggests a promising research direction for knowledge tracing that improves model prediction accuracy, model efficiency, resource utilization, and interpretability simultaneously.

Figures (6)

• Figure 1: Trends in data volume for education scenarios.
• Figure 2: Model performance vs. resource consumption, where x-axis lists various KT models, left y-axis shows training resource usage, right y-axis displays batch training and inference time, bar color indicates total training time.
• Figure 3: The overview of the proposed framework.
• Figure 4: Comparison of model time and space consumption when dealing with sequences of different lengths.
• Figure 5: Sequence-level model interpretability.