Academia Publishing Hub

Retrieval-Augmented Generation Architecture for Improving the Reliability of Generative Information Systems

← Усі матеріали збірника Тези доповіді

Retrieval-Augmented Generation Architecture for Improving the Reliability of Generative Information Systems

Maksym PetrovNational Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Kyiv, Ukraine
КонференціяМіжнародна наукова конференція “Contemporary Science, Technology and Society”
Дата21.09.2026
Сторінки80–82
ОрганізаторAcademia Publishing Hub

Анотація

The paper substantiates the use of retrieval-augmented generation as an architectural approach for generative information systems operating on local, regulatory, or frequently updated data. The workflow is considered as a sequence of corpus preparation, chunk indexing, semantic retrieval, context construction, and answer generation. A practical set of reliability criteria is proposed, including retrieval relevance, evidence coverage, source traceability, and resistance to stale data. Retrieval augmentation does not automatically eliminate erroneous outputs; however, it provides a technical basis for controlling information provenance and updating knowledge without full model retraining.

Ключові слова

retrieval-augmented generation, large language models, information systems, semantic search, reliability, artificial intelligence

How to cite (DSTU 8302:2015)

Petrov M. Retrieval-Augmented Generation Architecture for Improving the Reliability of Generative Information Systems. Contemporary Science, Technology and Society : Proceedings of the International Scientific Conference / Academia Publishing Hub (London, 21 September 2026). London, 2026. P. 80–82.

Сертифікати учасників