Skoči na glavni sadržaj

Izvorni znanstveni članak

https://doi.org/10.19279/TVZ.PD.2025-13-4-14

EVALUATION OF LARGE LANGUAGE MODELS IN A PROFESSIONAL ENVIRONMENT: IMPLEMENTATION, DOCUMENTATION, REFACTORING AND CODE ANALYSIS

Mario Seider ; Tehničko veleučilište u Zagrebu, Vrbik 8, Zagreb, Hrvatska *
Dominik Forjan ; Tehničko veleučilište u Zagrebu, Vrbik 8, Zagreb, Hrvatska *
Renato Belošević ; Tehničko veleučilište u Zagrebu, Vrbik 8, Zagreb, Hrvatska *
Ivan Cesar ; Tehničko veleučilište u Zagrebu, Vrbik 8, Zagreb, Hrvatska, Student *

* Dopisni autor.


Puni tekst: engleski pdf 566 Kb

str. 266-275

preuzimanja: 0

citiraj


Sažetak

This paper presents a comparative analysis of large language models used in professional working environments to support everyday knowledge-based tasks. The research focuses on practical usage scenarios, including the creation of technical documentation, software development support, natural language processing, and assistance in decision-making processes. The evaluation is conducted through a set of defined work tasks that reflect real working conditions, with models assessed according to response accuracy, consistency, clarity of generated content, and adaptability to task context. The results demonstrate noticeable differences among the analyzed models in handling complex professional queries and in the quality of generated outputs, indicating that performance strongly depends on the nature of the task. The analysis shows that certain models exhibit a clear advantage across most evaluation criteria, particularly in scenarios requiring advanced contextual understanding and technical reasoning. These findings emphasize the importance of selecting an appropriate model based on specific professional requirements and work scenarios. This study contributes to a clearer understanding of the practical applicability and limitations of large language models in everyday professional use.

Ključne riječi

large language models; professional working environment; artificial intelligence; application analysis; natural language processing

Hrčak ID:

352279

URI

https://hrcak.srce.hr/352279

Datum izdavanja:

30.1.2026.

Podaci na drugim jezicima: hrvatski

Posjeta: 0 *