Допомога великій мовній моделі при класифікації на багато класів

Вадим Тульчинський; Дмитро Жарков

doi:10.71942/b2gm-xt31

Preprint / Version 1

Helping a large language model with multi-class classification

Authors

Vadim Tulchinsky Glushkov Institute of Cybernetics, National Academy of Sciences of Ukraine https://orcid.org/0000-0002-0280-223X
Dmytro Zharkov V.M. Glushkov Institute of Cybernetics https://orcid.org/0009-0006-2700-5313

DOI:

https://doi.org/10.71942/b2gm-xt31

Keywords:

machine learning, artificial intelligence, large language models, multiclass classification, bag of words, pairwise comparison method, analytic hierarchy process

Abstract

Several approaches to improving the accuracy of product classification based on their textual descriptions are presented. Both the bag-of-words method based on abbreviated descriptions and the pairwise comparison method demonstrate the ability to significantly improve the performance of large language models in solving such a complex task. It seems that decision support methods developed for humans can be useful for generative artificial intelligence models in many cases.