EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-SA 4.0

Adaptive Learning Path Optimization with LLMs and Knowledge Graphs in Intelligent Tutoring Systems

Christos Papakostas, Christos Troussas, Akrivi Krouska, Cleo Sgouropoulou

Technical Annals · 28.01.2026 · 2 (3)

DOI: 10.12681/ta.45979 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Technical Annals, online erschienen am 28.01.2026. Verfasst von Christos Papakostas und 3 weiteren Autor:innen. Die Arbeit nennt 40 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Integrating large language models (LLMs) and knowledge graphs (KGs) in intelligent tutoring systems (ITSs) represents a revolutionary development in the field of personalized learning. In this study, we present a systematic literature review and an empirical analysis of frameworks for learning path optimization based on the synergies of LLMs and KGs. We conducted a thorough survey of 127 studies in this research area, covering papers published from 2019 to 2025. The surveyed methods have been categorized according to four main paradigms: rule-based adaptation, collaborative filtering, reinforcement learning, and adaptive systems with augmented capabilities due to LLMs. Based on the results obtained from the survey, we propose a new methodology of learning path generation referred to as the KG-LLM Adaptive Pathfinder (KLAP). The proposed approach leverages a domain-specific knowledge graph together with a fine-tuned large language model, allowing KLAP to develop learning paths in real time while taking into account the user's progress, prerequisite dependencies, and cognitive load factors. Using KLAP, we conducted an empirical evaluation on a sample consisting of 312 computer science students from three courses at the undergraduate level. Compared with traditional adaptive learning path systems, we show KLAP to produce substantial gains in terms of learning effectiveness (Cohen’s d = 0.74, 95% CI [0.51, 0.97]), efficiency (reduction of time-to-mastery by 23.6%), and user satisfaction (mean = 4.31/5.00, standard deviation = 0.52). Additionally, the findings support the hypothesis that KG and LLM components play complementarily super-additive roles.

Zitieren

APA

Papakostas, C., Troussas, C., Krouska, A., & Sgouropoulou, C. (2026). Adaptive Learning Path Optimization with LLMs and Knowledge Graphs in Intelligent Tutoring Systems. Technical Annals, 2(3). https://doi.org/10.12681/ta.45979

Vancouver

Papakostas C, Troussas C, Krouska A, Sgouropoulou C. Adaptive Learning Path Optimization with LLMs and Knowledge Graphs in Intelligent Tutoring Systems. Technical Annals. 2026;2(3). doi:10.12681/ta.45979

BibTeX
@article{Papakostas2026,
  author = {Papakostas, Christos and Troussas, Christos and Krouska, Akrivi and Sgouropoulou, Cleo},
  title = {Adaptive Learning Path Optimization with LLMs and Knowledge Graphs in Intelligent Tutoring Systems},
  journal = {Technical Annals},
  year = {2026},
  volume = {2},
  number = {3},
  doi = {10.12681/ta.45979}
}
Kostenlos starten →