AdvKT: An Adversarial Multi-Step Training Framework for Knowledge Tracing

2025-04-07

AI Summary

Existing knowledge tracing (KT) models employ single-step training paradigms, which misalign with the inherently multi-step inference process, leading to error accumulation; compounded by data sparsity, this severely limits recommendation performance. To address these challenges, we propose an end-to-end adversarial multi-step training framework. First, we introduce adversarial learning into KT—designing a generator-discriminator co-training mechanism to suppress error propagation in multi-step predictions. Second, we develop a generative data augmentation strategy tailored for sparse educational sequences. Our framework jointly integrates multi-step rolling prediction, sequential modeling, and adversarial optimization. Extensive experiments on four real-world educational datasets demonstrate significant improvements over state-of-the-art KT models, effectively mitigating both error accumulation and data sparsity.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Multiagent Systems: Adversarial Agents
• Game Theory and Economic Paradigms: Adversarial Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Knowledge Tracing (KT) monitors students' knowledge states and simulates their responses to question sequences. Existing KT models typically follow a single-step training paradigm, which leads to discrepancies with the multi-step inference process required in real-world simulations, resulting in significant error accumulation. This accumulation of error, coupled with the issue of data sparsity, can substantially degrade the performance of recommendation models in the intelligent tutoring systems. To address these challenges, we propose a novel Adversarial Multi-Step Training Framework for Knowledge Tracing (AdvKT), which, for the first time, focuses on the multi-step KT task. More specifically, AdvKT leverages adversarial learning paradigm involving a generator and a discriminator. The generator mimics high-reward responses, effectively reducing error accumulation across multiple steps, while the discriminator provides feedback to generate synthetic data. Additionally, we design specialized data augmentation techniques to enrich the training data with realistic variations, ensuring that the model generalizes well even in scenarios with sparse data. Experiments conducted on four real-world datasets demonstrate the superiority of AdvKT over existing KT models, showcasing its ability to address both error accumulation and data sparsity issues effectively.

Problem

• Reduces error accumulation in multi-step knowledge tracing
• Addresses data sparsity in intelligent tutoring systems
• Enhances generalization with adversarial learning and data augmentation

Innovation

• Adversarial learning with generator-discriminator paradigm
• Multi-step training reduces error accumulation
• Specialized data augmentation for sparse data