MIRA: Automated Generation of Measurable Non-Functional Requirements and Test Scenarios Using Multi-Agent Systems and RAG
Resumo
The specification of Non-Functional Requirements (NFRs) is critical to software quality, but suffers from human subjectivity and the hallucination tendency of generic LLMs. This work addresses the gap in the automated generation of accurate and testable NFRs from high-level user stories. We propose MIRA, a multi-agent architecture based on Retrieval-Augmented Generation (RAG) that orchestrates the contextualized production of requirements and test scenarios. From 11 representative user stories, MIRA automatically generated 25 NFRs and 49 test scenarios across 10 functional domains. Validation with 15 software professionals, totaling 165 evaluations per dimension, demonstrated high clarity (µ = 4.45), measurability (µ = 4.39) and test adherence (µ = 4.47) on a 5-point Likert scale, with strong internal consistency (Cronbach’s α > 0.84). It is concluded that the approach mitigates generative inconsistencies, acting as an effective accelerator in the integration between requirements engineering and validation.
Referências
Doran, G. T. (1981). There’s a s.m.a.r.t. way to write management’s goals and objectives. Management Review, 70(11):35–36.
Gao, Y., Xiong, Y., Gao, X., Jia, K., Pan, J., Bi, Y., Dai, Y., Sun, J., Wang, H., and Wang, H. (2023). Retrieval-augmented generation for large language models: A survey. arXiv preprint arXiv:2312.10997, 2(1).
Hair, J. F., Black, W. C., Babin, B. J., and Anderson, R. E. (2019). Multivariate Data Analysis. Cengage Learning, 8 edition.
Hevner, A. R., March, S. T., Park, J., and Ram, S. (2004). Design science in information systems research. MIS Quarterly, 28(1):75–105.
ISO/IEC/IEEE (2018). 29148:2018 – Systems and Software Engineering: Life Cycle Processes – Requirements Engineering. International Organization for Standardization.
Khan, J. A., Qayyum, S., and Dar, H. S. (2025). Large language model for requirements engineering: A systematic literature review.
Kula, E., Greuter, E., van Deursen, A., and Gousios, G. (2022). Factors affecting on-time delivery in large-scale agile software development. IEEE Transactions on Software Engineering, 48:3573–3592.
Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., et al. (2020). Retrieval-augmented generation for knowledge-intensive nlp tasks. In Proceedings of the 34th Conference on Neural Information Processing Systems (NeurIPS), pages 9459–9474.
Li, Y., Liu, P., Wang, H., Chu, J., and Wong, W. E. (2025). Evaluating large language models for software testing. Computer Standards & Interfaces, 93:103942.
Seaman, C. (1999). Qualitative methods in empirical studies of software engineering. IEEE Transactions on Software Engineering, 25:557–572.
Sommerville, I. (2011). Engenharia de Software. Pearson, São Paulo, Brasil, 9 edition.
Vasconcelos, A. K. M., Moreira, E. R. P., Sousa, R. A. S., Monteiro, G. C. V., Bandeira, A. P., de Souza, J. T., Maia, P. H. M., and Santos, I. S. (2026). Replication package for semish 2026. [link]. Public repository for replication package.
Viviani, L., Guerra, E., Melegati, J., and Wang, X. (2023). An empirical study about the instability and uncertainty of non-functional requirements. In International Conference on Agile Software Development, pages 77–93. Springer Nature Switzerland Cham.
Wang, J., Huang, Y., Chen, C., Liu, Z., Wang, S., and Wang, Q. (2024). Software testing with large language models: Survey, landscape, and vision. IEEE Transactions on Software Engineering, 50(4):911–936.
Yao, S. et al. (2023). React: Synergizing reasoning and acting in language models. In Proceedings of the International Conference on Learning Representations (ICLR).
