KT4EQG: Personalized Exercise Question Generation via Knowledge Tracing

KT4EQG is a new educational framework that combines knowledge tracing with AI-powered question generation to create personalized exercise questions for students. The system uses machine learning to model each student's knowledge state and generates customized questions designed to maximize learning outcomes, demonstrating superior effectiveness compared to non-personalized approaches.

KT4EQG represents a meaningful advance in adaptive learning technology by addressing a fundamental gap in educational question generation. Traditional EQG systems generate questions without considering individual student knowledge gaps, resulting in exercises that may be either too easy or too difficult. This research bridges that gap by integrating knowledge tracing—a well-established technique for modeling student learning—with large language models for question synthesis.

The approach reflects growing recognition that personalization at scale requires sophisticated modeling of learner states. Educational technology has increasingly adopted knowledge tracing to predict performance, but few systems leverage these predictions to actively shape content generation. KT4EQG closes this loop by using KT models to identify which knowledge concepts would benefit a student most, then generating contextually appropriate questions around those concepts.

For the EdTech sector, this work has significant implications. Platforms serving millions of students face the challenge of maintaining engagement while accommodating vastly different learning levels. A system that generates truly personalized exercises could reduce student dropout rates and improve learning outcomes—key metrics that drive retention and institutional adoption. The experimental validation on real datasets (XES3G5M and MOOCRadar) suggests the approach transfers to actual educational contexts.

Looking forward, the integration of knowledge tracing with generative AI will likely become standard in adaptive learning systems. The primary challenges involve scaling the computational requirements for real-time personalization and ensuring question quality remains consistent across diverse concepts and difficulty levels. Research extending this work to multi-modal questions or real-time adaptive difficulty adjustment could further enhance educational effectiveness.

• →KT4EQG combines knowledge tracing with LLM-based question generation to create personalized student exercises.
• →The system identifies each student's optimal knowledge concepts to practice, maximizing learning improvement.
• →Experimental results show KT4EQG outperforms non-personalized and limited-personalization approaches.
• →Knowledge tracing enables adaptive learning systems to generate contextually relevant content at scale.
• →The technology addresses a critical gap in educational AI by connecting learner modeling with content generation.