Institutional Repository of Vilnius University
LLM visuomenė: RAG grįstas personų kūrimas ir jų poveikis daugiaagentėje diskusijų sistemoje /
Abstract
dc:descriptionThis thesis presents an experimental agentbased discussion system using large language models (LLMs), in which agents are assigned personas generated by a RAGbased (RetrievalAugmented Generation) module that draws on information collected from the web about a given discussion topic. This RAG based method is compared with a baseline method, which relies only on information encoded in the model weights. The study was conducted using two topics relevant to the Lithuanian context and two LLMs differing in architecture and parameter count. The results show that the RAG based method creates more diverse personas; however, this difference does not produce a measurable effect on agents’ voting behavior – the dominant factor shaping the discussion dynamics remains the combination of the model and the topic.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Urbutis, Simonas Jaunius,
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:ELABAETD308338292&prefLang=en_US