IVML  
  about | r&d | publications | courses | people | links
   

C. Troussas, A. Krouska, Ph. Mylonas, C. Sgouropoulou, I. Voyiatzis
Retrieval-Augmented Generation for Adaptive Intelligent Tutoring: A Framework Integrating LLMs with Learner Modeling
11th South-East Europe Design Automation, Computer Engineering, Computer Networks and Social Media Conference (SEEDA-CECNSM 2026), Thessaloniki, Greece, September 18-20, 2026
ABSTRACT
Intelligent tutoring systems have historically focused on providing personalized educational experiences, but often falter due to narrow content coverage and static adaptation. In this paper we propose a new framework combining Retrieval-Augmented Generation (RAG) with adaptive learner modeling to enable intelligent tutoring systems that provide individualized, context-sensitive learning material. Our framework integrates a dense retrieval module that consults a database of curated course materials, a dynamic learner model that updates the learner's knowledge using Bayesian knowledge tracing, and an LLM that integrates the retrieved knowledge and the learner model to generate personalized explanations, hints, and assessments. We experiment with our framework in the context of five undergraduate computer science courses and across 247 students during a 10 week semester. We show that our proposed RAG-Adaptive system generates greater learning gains (average normalized gain of 0.839) than a basic RAG system (average normalized gain of 0.790), a stand-alone LLM tutor (average normalized gain of 0.719) and a static ITS (average normalized gain of 0.648). Student satisfaction responses demonstrate overwhelming support for the adaptive system (and in particular for its relevance and adaptivity) while a controlled ablation study demonstrates the impact each of the components had on the overall learning gain.
18 September, 2026
C. Troussas, A. Krouska, Ph. Mylonas, C. Sgouropoulou, I. Voyiatzis, "Retrieval-Augmented Generation for Adaptive Intelligent Tutoring: A Framework Integrating LLMs with Learner Modeling", 11th South-East Europe Design Automation, Computer Engineering, Computer Networks and Social Media Conference (SEEDA-CECNSM 2026), Thessaloniki, Greece, September 18-20, 2026
[ BibTex] [ Print] [ Back]

© 00 The Image, Video and Multimedia Systems Laboratory - v1.12