Exploring Multilingual and Cross-lingual Properties of Large Language Models

Student: Bukashkin Anton

Educational Programme: Financial Technology and Data Analysis (Master)

Year of Graduation: 2024

This thesis investigates the multilingual and cross-lingual properties of large language models, specifically their ability to understand and respond to languages that are underrepresented in their training data. We examine three large language models (LLaMA2, Mistral, and Gemma) and analyze their attention weights to identify patterns and mechanisms that enable them to respond to questions in multiple languages. Our results provide insights into the behavior of these models and shed light on the challenges of interpreting their multilingual capabilities. The findings of this study have implications for the development of more accurate and efficient language models, as well as their applications in various domains.

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses