Structured models improve finance predictions across changing conditions
Toward Structured Financial Foundation Models
Artificial IntelligenceComputational Engineering, Finance, and Science
Summary
Financial systems are complicated because they involve many interacting parts, events that happen at different times, and rare but important situations. The authors suggest building specialized computer models that store financial information over time and can predict what might happen depending on different actions. These models should work well even with limited data, explain their decisions clearly, and adapt to changing market conditions. Their approach promotes linking these specialized models so they can be reused and combined for different financial tasks.
What this means in practice
- •For financial technology teams: Build modular, reusable models that capture financial state and predict outcomes under changing market conditions for better risk and strategy analysis.
- •For insurance actuaries: Create simulations of rare but impactful financial events using modular models that adapt to asynchronous and interacting factors in financial data.
A position paper. It proposes an approach and reports no results.
Authors
Natraj Raman, Saumitra Mishra, Vamsi Potluru, Sumitra Ganesh
Abstract
Foundation models promise reusable capabilities and step-change performance gains across tasks. In finance, large-scale pretraining of domain-specialized foundation models can advance both goals. However, prevailing architectures and objectives do not by themselves address the combined demands of financial systems: partial observability, interacting entities, asynchronous events, changing regimes, action-dependent outcomes, and rare states with disproportionate consequences. Moreover, independently developed specialists may learn incompatible representations, while multi-domain monolithic alternatives couple data access, model updates, and governance. We therefore propose a structured-data-first research agenda centered on modular, domain-specialized financial foundation models connected by a representation fabric for reuse and composition across domains and tasks. These models should maintain persistent financial state across aggregation levels and time horizons, capture its dynamics, and simulate action-conditioned futures, while remaining calibrated across regimes and for tail outcomes. They must also learn efficiently under data and access constraints and provide faithful, traceable explanations. We identify research challenges, directions, and capability-oriented evaluation criteria.