The Institute of Informatics of the University of Szeged organized the twenty-second Hungarian Computational Linguistics Conference on January 29 and 30, 2026. At the conference, Zsolt Csibi, Bence György Gortka, Natabara Gyöngyössy, Kornél Nagy, Dávid Márk Nemeskey, Martin Sallai, András Simonyi, András Márk Szekeres and Gábor Palkó gave a presentation entitled Efficient Hungarian LLM Adaptation on Academic Infrastructure.

In the article submitted to the conference, the authors presented a “lightweight” large language model called Racka, which was designed to effectively support the adaptation of large language models to Hungarian even on academic-level high-performance computing (HPC) infrastructure. The aim of the development was to develop a methodology that can resource-efficiently improve the Hungarian language capabilities of available language models, while ensuring that the capabilities in high-resource languages ​​(especially English) are not damaged.

The authors received Best Paper recognition for the study and presentation.

The full conference program is available at: https://rgai.inf.u-szeged.hu/mszny2026/program

The conference proceedings are available at: https://rgai.inf.u-szeged.hu/sites/rgai.inf.u-szeged.hu/files/mszny2026.pdf