Institutional Repository of Vilnius University
Large language models for lithuanian language /
Abstract
dc:descriptionEfforts to develop large language models for the Lithuanian language have been limited, primarily due to data and resource constraints. In this work, we aim to address this issue by training models specifically tailored for Lithuanian. We enhance existing multilingual large language models through additional training and develop a new Lithuanian-specific model with an optimized tokenizer. To evaluate their performance, we test these models across a diverse set of benchmarks. The results highlight both the strengths and weaknesses of Lithuanian LLMs while suggesting areas for improvement, including higher quality data collection, synthetic data generation, advanced training techniques, and more effective model design.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Plesevičius, Dominykas,
Subjects
dc:subject × 1Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- eng
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD229583321&prefLang=en_US