Исследование прикладного использования языковых моделей на основе метода генерации с дополненной выборкой

Искусственный интеллект и принятие решений. 2025. № 2. С. 73–89.

Loginova I., Grozovskiy F.

The article presents a qualitative analysis of Russian and global cases of development and implementation of Retrieval-Augmented Generation models (RAG models) to address applied analytical and business tasks. RAG models outperform traditional large language models in accuracy, relevance, and contextual appropriateness of generated responses by utilizing external knowledge sources. This makes Retrieval-Augmented Generation an important area of research and development in artificial intelligence. The analysis covered 21 cases of RAG model development and use by companies and government organizations in Russia and abroad. The results of the analysis indicate that the goals of practical application of RAG models are mainly cost reduction, as well as improvement of customer and user experience.

Keywords: искусственный интеллект, Artificial intelligence (AI), Большие языковые модели (LLMs), Large language models (LLM), Retrieval-Augmented Generation (RAG), development and implementation of RAG models, Retrieval-Augmented Generation (RAG), разработка и внедрение RAG-моделей