2409-1634Research result. Economic Research2409-163410.18413/2409-1634-2026-12-2-0-74237ECONOMICS,MANAGEMENT AND ACCOUNTING IN A FIRMTHE POSSIBILITIES OF USING LANGUAGE MODELS TO SOLVE THE PROBLEMS OF INNOVATION MANAGEMENTTHE POSSIBILITIES OF USING LANGUAGE MODELS TO SOLVE THE PROBLEMS OF INNOVATION MANAGEMENTVodolazhenkoRoman A.VodolazhenkoRoman A.TsoiValentin V.TsoiValentin V.KikotArtem V.KikotArtem V.KopanevArtem A.KopanevArtem A.202612200

The article solves the scientific problem of experimental confirmation of the evaluation and verification of RAG models for use in the innovation management process for cases similar to those implemented in T-bank, namely, the study of the possibilities of implementing patent search using RAG models. The object of the study is the effectiveness of the systems under consideration in comparison with the effectiveness of other types of language models due to the use of an external database within the subject area, in this case, patent analytics. To ensure the practical value of the research results, a comprehensive methodological approach was used, including empirical analysis, formulation and processing of experimental results, and quantitative processing of the data obtained. To test RAG, an experiment was developed and implemented to evaluate the quality of responses provided by various language models. The results show that RAG improves model accuracy by 50% compared to GPS and GigaChat. The scientific novelty of the article lies in the development and experimental verification of an integrated approach to the study of the possibility of solving problems of patent analysis.

The article solves the scientific problem of experimental confirmation of the evaluation and verification of RAG models for use in the innovation management process for cases similar to those implemented in T-bank, namely, the study of the possibilities of implementing patent search using RAG models. The object of the study is the effectiveness of the systems under consideration in comparison with the effectiveness of other types of language models due to the use of an external database within the subject area, in this case, patent analytics. To ensure the practical value of the research results, a comprehensive methodological approach was used, including empirical analysis, formulation and processing of experimental results, and quantitative processing of the data obtained. To test RAG, an experiment was developed and implemented to evaluate the quality of responses provided by various language models. The results show that RAG improves model accuracy by 50% compared to GPS and GigaChat. The scientific novelty of the article lies in the development and experimental verification of an integrated approach to the study of the possibility of solving problems of patent analysis.

language modelsRAGartificial intelligencepatent analyticsmetricslanguage modelsRAGartificial intelligencepatent analyticsmetricsСписок литературыWang, Suyuan & Yin, Xueqian & Wang, Menghao & Guo, Ruofeng & Nan, Kai. (2024) “EvoPat: A Multi-LLM-based Patents Summarization and Analysis Agent”, 10.48550/arXiv.2412.18100.Ding, Yao & Wu, Yuqing & Ding, Ziyang. (2025) “An automatic patent literature retrieval system based on LLM-RAG”, Journal of Technology Innovation and Engineering, 1. 10.63887/jtie.2025.1.3.3.Santos, Fabiano & Peres, Mariana & Junqueira Braga, Edimilson. (2025) COB-2025-2478 Optimization of patent search strategies for mechanical engineering: methodologies, tools, and applications.Khan, A.A. et al. (2024) “Developing Retrieval Augmented Generation (RAG) based LLM Systems from PDFs: An Experience Report”, arXiv preprint arXiv:2410.15944v1, available at: https: // arxiv.org/abs/2410.15944v1Bommasani, R. et al. “On the opportunities and risks of foundation models”, arXiv preprint arXiv:2108.07258, DOI: https: // doi.org/10.48550/arXiv.2108.07258Vaswani, A., Shazeer, N., Parmar, N. et al. (2017) “Attention is all you need”, Advances in Neural Information Processing Systems, 30.Koreisha, Z.A. Parshina, V.S. (2017) “Research of patent and innovation activity as a factor of economic development of Russia”, Issues of innovative economics, 7, 1, 31-39.Dobrokhodov, Ya.I., Surina, A.V. (2024) “Large language models: forecasting development using the Gartner method”, SAEC, 2, available at: https://cyberleninka.ru/article/n/bolshie-yazykovye-modeli-prognozirovanie-razvitiya-po-metodu-gartneraChungulova, G.K., Orazalieva, E.N. (2024) “The possibilities and problems of large language models in education using the example of ChatGPT”, Research and Development/S&R, 4(20), available at: https://cyberleninka.ru/article/n/vozmozhnosti-i-problemy-bolshih-yazykovyh-modeley-v-obrazovanii-na-primere-chatgptKashirina, I.L., Osipov, I.R., Yakovlev, V.A. (2025) “Development and evaluation of a RAG system for semantic relations analysis”, Bulletin of Voronezh State University. Series: System Analysis and Information Technology, 2, 114-126.Lyagoshina, T.V. (2024) “Big language models: impact on public discourse and society as a whole”, Bulletin of Tomsk State University. Philosophy. Sociology. Political science, 79, available at: https://cyberleninka.ru/article/n/bolshie-yazykovye-modeli-vliyanie-na-publichnyy-diskurs-i-obschestvo-v-tselomObolensky, D.M., Shevchenko, V.I. (2024) “The use of the RAG method and large language models in intelligent educational ecosystems”, Economy. Computer science. 3, available at: https://cyberleninka.ru/article/n/ispolzovanie-metoda-rag-i-bolshih-yazykovyh-modeley-v-intellektualnyh-obrazovatelnyh-ekosistemahNaumenko, A.O. (2025) “RAG technology (retrieval-augmented generation) as an innovative approach in LLM”, Bulletin of Science, 8 (89), available at: https://cyberleninka.ru/article/n/tehnologiya-rag-retrieval-augmented-generation-kak-innovatsionnyy-podhod-v-llm