Institutional Repository of Vilnius University
Didžiųjų kalbos modelių taikymai mokymosi procese /
Abstract
dc:descriptionThe COVID19 pandemic accelerated the development of remote learning, yet existing learning management systems often lack individualized support for students and tools to reduce teachers’ workload. Additionally, the global teacher shortage (with a projected deficit of 44 million primary and secondary teachers by 2030) increases administrative burdens and diminishes teaching quality. The integration of large language models into educational environments enables the creation of personalized chatbots capable of providing realtime contextual assistance to students while simultaneously reducing mentors’ administrative workload, thereby improving overall learning efficiency. The thesis includes a literature review identifying the shortcomings of current virtual learning environments and the potential applications of artificial intelligence in education. An interactive learning management system was designed and implemented for mentors and students, featuring rolebased access control, course and assignment management, and an integrated chatbot. The system was developed using the Next.js framework, React library, TypeScript language, Prisma ORM with a relational database, Tailwind CSS for styling, and NextAuth.js for authentication. The artificial intelligence functionality is based on the OpenAI GPT4omini model, with serverside contextual prompt engineering tailored to the user’s progress and course data. An empirical study compared four language models (GPT4omini, FlanT5base, DistilGPT2, TinyLlama) in terms of response generation latency, pedagogical quality, and ethical compliance (using 16 thematic queries). Additional system testing (20 queries) demonstrated that GPT4omini exhibits the lowest latency (median 2740 ms), the highest pedagogical quality (average 1.94/2), and full ethical compliance (2/2), ensuring personalized responses and resistance to unethical queries. The developed prototype demonstrates the practical application of large language models in remote learning, effectively addressing personalization gaps for students and administrative workload issues for mentors, while creating prerequisites for further advancement of artificial intelligence in educational systems.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Tamošiūnas, Adas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD277690419&prefLang=en_US