AAKT: Enhancing Knowledge Tracing with Alternate Autoregressive Modeling

TL;DR

The paper reframes Knowledge Tracing as a generative, autoregressive process and introduces Alternate Autoregressive Knowledge Tracing (AAKT). By interleaving question and response information into alternate sequences and applying a sliding window, AAKT enables end-to-end autoregressive modeling of knowledge states using a GPT-J–style transformer, while an auxiliary task enforces skill-aware embeddings and a time-feature enriches input representations. Across four real-world datasets, AAKT achieves superior AUC, ACC, and RMSE compared with strong baselines, with ablations confirming the importance of alternating sequences, the auxiliary task, and information such as response time. This work demonstrates a viable, scalable autoregressive framework for KT that effectively integrates questions, responses, skills, and timing to improve predictive accuracy and interpretability in educational settings.

Abstract

Knowledge Tracing (KT) aims to predict students' future performances based on their former exercises and additional information in educational settings. KT has received significant attention since it facilitates personalized experiences in educational situations. Simultaneously, the autoregressive modeling on the sequence of former exercises has been proven effective for this task. One of the primary challenges in autoregressive modeling for Knowledge Tracing is effectively representing the anterior (pre-response) and posterior (post-response) states of learners across exercises. Existing methods often employ complex model architectures to update learner states using question and response records. In this study, we propose a novel perspective on knowledge tracing task by treating it as a generative process, consistent with the principles of autoregressive models. We demonstrate that knowledge states can be directly represented through autoregressive encodings on a question-response alternate sequence, where model generate the most probable representation in hidden state space by analyzing history interactions. This approach underpins our framework, termed Alternate Autoregressive Knowledge Tracing (AAKT). Additionally, we incorporate supplementary educational information, such as question-related skills, into our framework through an auxiliary task, and include extra exercise details, like response time, as additional inputs. Our proposed framework is implemented using advanced autoregressive technologies from Natural Language Generation (NLG) for both training and prediction. Empirical evaluations on four real-world KT datasets indicate that AAKT consistently outperforms all baseline models in terms of AUC, ACC, and RMSE. Furthermore, extensive ablation studies and visualized analysis validate the effectiveness of key components in AAKT.

Figures (11)

• Figure 1: An overview of AAKT. Every element in students' history interactions is split into anterior information and posterior information. Subsequently, they are organized in an alternate manner, facilitating autoregressive modeling.
• Figure 2: The overall architecture of the proposed AAKT framework. It comprises three integral components.
• Figure 3: Different modeling strategy between previous KT models and AAKT.
• Figure 4: Overall illustration of sliding window technique in our model. There are slight variations in the approach for the training dataset compared to the testing dataset.
• Figure 5: Four categories of interaction-related information and examples.