Hybrid Reasoning with Large Language Models for CFO-Oriented Financial Decision Support

Rong Nie

doi:10.54691/28mjw680

Authors

Rong Nie Hohhot No.2 High School, Hohhot, China

DOI:

https://doi.org/10.54691/28mjw680

Keywords:

Large language models, hybrid reasoning, decision support systems, financial intelligence.

Abstract

The Chief Financial Officer (CFO) has gone beyond the scope of financial reporting and compliance to strategic decisions in more complex and data-saturated business settings. Although the recent development of large language models (LLMs) demonstrates a potential to succeed in financial text analysis and decision support, the current solutions are more generic and text-focused, which reduces their credibility in quantitative reasoning and financial work that is sensitive to regulations. In the present paper, the vertical large language model framework that we recommend to be used in the CFO-oriented decision support is namely domestic retail and export-oriented manufacturing industries. The suggested structure assumes a modular hybrid reasoning system that incorporates structured financial information processing, explicit number processing, regulatory attitude processing, and language-based decision generation. The separation between deterministic financial calculation and natural language explanation of the framework increases numerical accuracy, consistency of decisions, and interpretability and reduces the hallucination risks often linked to language-based models that are entirely driven by language.

Downloads

Download data is not yet available.

References

[1] Appelbaum, D., Kogan, A., & Vasarhelyi, M. (2017). The impact of big data on accounting and auditing. Accounting Horizons, 31(3), 1–27.

[2] Araci, D. (2019). Finbert: Financial sentiment analysis with pre-trained language models. In Proceedings of the 2019 ACL Workshop on Economics and Natural Language Processing (pp. 70–79).

[3] Brown, T. B., et al. (2020). Language models are few-shot learners. Advances in Neural Information Processing Systems, 33, 1877–1901.

[4] Bustamante, M. C., et al. (2021). Machine learning in accounting: A systematic literature review. Accounting Perspectives, 20(4), 547–582.

[5] Hochreiter, R. (2017). Regtech: The application of modern information technology in regulatory monitoring and compliance. Journal of Risk Management in Financial Institutions, 10(4), 362–374.

[6] OpenAI. (2023). Gpt-4 technical report. arXiv preprint arXiv:2303.08774.

[7] Yang, Y., et al. (2020). Finbert: A pretrained language model for financial communications. arXiv preprint arXiv:2006.08097.

[8] Zoni, L., & Merchant, K. A. (2016). How cfos use management accounting systems to support strategy formulation and implementation. Management Accounting Research, 30, 4–20.