Previously, it was established that LLMs such as GPT models are able to generate simple SVG graphics. But can we push the boundaries of LLMs in generating vector graphics by training them on a custom dataset of vector graphics? This research aims at creation of such a dataset for enhancing the performance of LLMs.

Тимофеенко Б.А. (науч. рук. Фильченков А.А.) Creation of a dataset of vector digits for training LLMs for vector graphics generation // Сборник тезисов докладов конгресса молодых ученых. Электронное издание. – СПб: Университет ИТМО, [2024]. URL: https://kmu.itmo.ru/digests/article/13560