Prompt engineering for identifying patentable technical solutions in scientific publications

Highlights A methodology for identifying patentable solutions in scientific articles, implemented through prompt engineering methods using the large language model DeepSeek-V3.2, has been developed. Standard matrices for drafting patent applications for three types of objects (method, device, substance) have been created, integrating the requirements of Russian patent legislation and systematized applicant errors. Six scientific principles for using AI to transform scientific results into intellectual property objects have been formulated. Relevance. A significant portion of scientific experimental publications contains patentable technical solutions that have not been identified by the authors. Most articles were published more than 12 months ago, which leads to missing the novelty grace period and creates obstacles for patenting. Purpose of the study is to develop a methodology for identifying patentable solutions in scientific articles using AI, enabling the transformation of published results into patent applications. Materials and Methods. The study was based on an analysis of FIPS (Federal Institute of Industrial Property) inquiries regarding the examination of invention applications. The DeepSeek-V3.2 language model was used for processing the articles. A primary assessment checklist (10 criteria) and standard matrices for drafting applications for methods, devices, and substances were developed. Approbation was carried out on six articles of various topics, including an English-language publication. Results. Of the six articles, five (83.3%) contained patentable solutions. Based on these, 9 patent applications with 15 independent claims were drafted. Publications with an expired novelty grace period (up to 7 years) were successfully circumvented by introducing new essential features. The English-language case confirmed the language independence of the methodology. Six scientific principles for using AI to identify inventions were formulated. Conclusion. The developed methodology enables the identification of inventions in scientific publications. The methodology is conveyed through prompts and can be reproduced in a new chat session. The key success factor is the presence of quantitative experimental data in the publication. The proposed strategy for circumventing prior publications allows for patenting solutions years after publication, provided the article states the problem without disclosing a specific solution. Practical significance of the work. The developed prompt engineering system, checklist, and standard matrices can be used by patent services and inventors of the Armed Forces of the Russian Federation to identify and secure rights to domestic developments, thereby contributing to Russia's technological sovereignty.

1. Supotnitskiy MV. Typical Mistakes in Claims and Specifications of the Inventions in the NBC Protection Corps. Journal of NBC Protection Corps. 2023;7(1):73–81. (In Russ.). EDN: untpoj. https://doi.org/10.35825/2587-5728-2023-7-1-73-81

2. Guo D, Yang D, Zhang H, Song J, Wang P, Zhu Q, et al. DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning. Nature. 2025;645:633–638. https://doi.org/10.1038/s41586-025-09422-z

3. Hoyt RE, Bajwa M. Measuring the Accuracy and Reproducibility of DeepSeek R1, Claude 3.5 Sonnet, and GPT-4.1 on Complex Clinical Scenarios. Appl Clin Inform. 2026;17(1):64–72. https://doi.org/10.1055/a-2807-4256

4. Xu P, Wu Y, Jin K, Chen X, He M, Shi D. DeepSeek-R1 outperforms Gemini 2.0 Pro, OpenAI o1, and o3-mini in bilingual complex ophthalmology reasoning. Adv Ophthalmol Pract Res. 2025;5(3):189–95. https://doi.org/10.1016/j.aopr.2025.05.001

5. Debnath T, Siddiky MNA, Rahman ME, Das P, Guha AK, Rahman RH, et al. A Comprehensive Survey of Prompt Engineering Techniques in Large Language Models. TechRxiv. 2025. https://doi.org/10.36227/techrxiv.174140719.96375390/v2

6. Wang H, Kim J, Lee S, Park C. A Comprehensive Survey of Automatic Prompt Optimization: Taxonomies, Frameworks, and Future Directions. arXiv [preprint]. 2025. arXiv:2506.15147

7. Kim D, Lee S, Park J, Choi Y. Theoretical Foundations of Prompt Engineering: From Heuristics to Expressivity. arXiv [preprint]. 2026 Jan 13. arXiv:2512.12688v2 https://doi.org/10.48550/arXiv.2512.12688

8. Ari U. 5C Prompt Contracts: A Minimalist, Creative-Friendly, Token-Efficient Design Framework for Individual and SME LLM Usage. arXiv [preprint]. 2025 Aug 5. arXiv:2507.07045 https://doi.org/10.48550/arXiv.2507.07045

9. Freeman B, Kicklighter A, Erdman M, Gordon Z. Toward Epistemic Stability: Engineering Consistent Procedures for Industrial LLM Hallucination Reduction. arXiv [preprint]. 2026 Mar 8. arXiv:2603.10047v1

10. Huang K, Chen J, Li M, Zhang Y. Prompt Engineering for Requirements Engineering: A Literature Review and Roadmap. arXiv [preprint]. 2025 Jul 10. arXiv:2507.07682 https://doi.org/10.48550/arXiv.2507.07682

11. Morishige M, Koshihara R. Ensuring Reproducibility in Generative AI Systems for General Use Cases: A Framework for Regression Testing and Open Datasets. arXiv [preprint]. 2025. arXiv:2505.02854 https://doi.org/10.48550/arXiv.2505.02854

12. Zeng Q, Jin C, Wang X, Zheng Y, Li Q. AIRepr: An Analyst-Inspector Framework for Evaluating Reproducibility of LLMs in Data Science. In: Findings of the Association for Computational Linguistics: EMNLP 2025. Suzhou, China: Association for Computational Linguistics; 2025. P. 10170–201.

13. Tao X, Tula A, Chen X. From prompt design to iterative generation: Leveraging LLMs in PSE applications. Computers & Chemical Engineering. 2025. https://doi.org/10.1016/j.compchemeng.2025.108284

14. Zhou Y, Muresanu AI, Han Z, et al. Large Language Models Are Human-Level Prompt Engineers. arXiv [preprint]. 2022. arXiv:2211.01910 https://doi.org/10.48550/arXiv.2211.01910

15. Vilakati S. Prompt engineering for accurate statistical reasoning with large language models in medical research. Frontiers in Artificial Intelligence. 2025;8:1658316. https://doi.org/10.3389/frai.2025.1658316

16. Li Z, Wang X, Yang Y, Yao Z, Xiong H, Du M. Adaptive Prompting in the Metaverse: An Iterative Prompt Optimization Framework for Enhancing LLM Performance Across Diverse Tasks. In: 2025 IEEE International Conference on Metaverse Computing, Networking and Applications (MetaCom); 2025. P. 245–52. https://doi.org/10.1109/MetaCom63298.2025.00045

17. Kargaran AH, Modarressi A, Nikeghbal N, Diesner J, Yvon F, Schütze H, et al. MEXA: Multilingual Evaluation of English-Centric LLMs via Cross-Lingual Alignment. In: Findings of the Association for Computational Linguistics: ACL 2025. Vienna, Austria: Association for Computational Linguistics; 2025. P. 27001–23. https://doi.org/10.48550/arXiv.2410.05873

18. Doddapaneni S, Khan MSUR, Venkatesh D, Dabre R, Kunchukuttan A, Khapra MM, et al. Cross-Lingual Auto Evaluation for Assessing Multilingual LLMs. In: Proceedings of the 63rd Annual Meeting of the Association forComputational Linguistics (Volume 1: Long Papers). Vienna, Austria: Association for Computational Linguistics; 2025. P. 29297–329. https://doi.org/10.48550/arXiv.2410.13394

19. Dzhermakyan VYu. Patent law under the Civil Code of the Russian Federation: article-by-article commentary, application practice, reflections. 3rd ed., revised and enlarged. Moscow; 2014. (In Russ.).

20. Panda P, Sycheva A. ChatGPT. Master podskazok, ili kak sozdavat' sil'nye prompty dlya neyroseti. Saint Petersburg: Piter; 2024. 224 p. (In Russ.).