Deterministic Reproducibility in Financial AI Systems: A Formal Architectural Model
DOI:
https://doi.org/10.15662/IJRAI.2026.0901012Keywords:
Deterministic reproducibility, Model determinism, Financial artificial intelligence, Regulated AI systemsAbstract
The paper is a professional architectural plan of deterministic reproducibility of financial artificial intelligence systems which would be executed under regulatory solutions. According to the regulators of the financial sector, the AI-based actions must be reproduced with the same exact outcome a number of years afterwards, something that cannot be achieved with most of the current AI products that are made nondeterministic. It proposed a system-level design that achieves reproducibility, through data snapshots, which are immutable, has pipelines with versions, is deterministically modeled in execution, and has cryptographically verifiable audit evidence. The problem of reproducibility can be also suggested to be an architectural property as well as not a model characteristic by various financial task-based quantitative experiments. The results suggest that long horizon financial compliance entails deterministic decision rebuilding that is practicable.
References
[1] Franco, R. (2025). LLM Output Drift: Cross-Provider Validation & Mitigation for Financial Workflows. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2511.07585
[2] Kao, L. (2025). Constant-Size Cryptographic Evidence Structures for regulated AI workflows. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2511.17118
[3] Wang, J. J., & Wang, V. X. (2025). Assessing consistency and reproducibility in the outputs of large language models: evidence across diverse finance and accounting tasks. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2503.16974
[4] Meyman, E. (2025). Deterministic Governance as Epistemic Commitment: Why reproducibility Matters for AI accountability. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.5659170
[5] Onyenahazi, O. B. (2025). Integrating artificial intelligence in financial auditing to enhance accuracy, efficiency, and regulatory compliance outcomes. International Journal of Research Publication and Reviews, 6(7), 23–44. https://doi.org/10.55248/gengpi.6.0725.2402
[6] Carter, R. E., Attia, Z. I., Lopez-Jimenez, F., & Friedman, P. A. (2019). Pragmatic considerations for fostering reproducible research in artificial intelligence. Npj Digital Medicine, 2(1), 42. https://doi.org/10.1038/s41746-019-0120-2
[7] Choowan, P., & Daovisan, H. (2025). Artificial Intelligence in Data Governance for Financial Decision-Making: A Systematic Review. Big Data and Cognitive Computing, 10(1), 8. https://doi.org/10.3390/bdcc10010008
[8] Li, D., Yu, G., Wang, X., & Liang, B. (2025). AuditableLLM: a Hash-Chain-Backed, Compliance-Aware auditable framework for large language models. Electronics, 15(1), 56. https://doi.org/10.3390/electronics15010056
[9] Han, H. (2025). Challenges of reproducible AI in biomedical data science. BMC Medical Genomics, 18(S1), 8. https://doi.org/10.1186/s12920-024-02072-6
[10] Rahrooh, A., Garlid, A. O., Bartlett, K., Coons, W., Petousis, P., Hsu, W., & Bui, A. A. (2023). Towards a framework for interoperability and reproducibility of predictive models. Journal of Biomedical Informatics, 149, 104551. https://doi.org/10.1016/j.jbi.2023.104551
