Implementation of Large Language Models in the Educational Process

Le résumé fourni par la source

The aim of the article is to present a methodology for using LLMs in higher education, specifically in the context of teaching the course Applications of Database Systems. The paper describes the architectural principles of LLMs and distinguishes between commercial and open-source solutions, with an emphasis on the possibilities for local deployment. It highlights the pedagogical, ethical, and technological barriers that accompany their integration into teaching. In the experimental part, the LLaMA 3.2 1B and Qwen 2.5 7B models were tested in the LM Studio environment on eight types of tasks, ranging from SQL query generation and code explanation to automated assessment of student solutions. The results indicate that LLMs can significantly reduce the time needed to prepare didactic materials, provide consistent feedback, and support personalized learning. At the same time, limitations were observed in the language quality for Slovak texts, the risk of inaccurate or fabricated outputs, and the need for human oversight in the evaluation of complex tasks. The discussion also addresses issues of data privacy and provides recommendations for the responsible use of these tools in university settings. The article offers practical recommendations for educators and institutions planning to integrate LLMs into teaching, and establishes a foundation for future research focused on model localization and the development of AI literacy.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.