Adaptive learning and financial stability: The role of network topology reshaping

In technology-driven financial markets, traditional static financial network analysis frameworks cannot capture the dynamic adaptation of agents' behaviors and their impact on systemic stability. To address this limitation, this study develops a dynamic multi-agent financial network model, embedded with a behavioral strategy learning mechanism. Simulation results show that decision-making based on learning mechanisms reduces loan default rates for both banks and firms, significantly improving financial system stability. Moreover, behavioral strategy learning inherently involves a trade-off between risk and return, which drives the network topology to evolve from a uniformly dense structure to one that is globally sparse yet locally clustered. The reshaping of the financial network helps contain risk propagation. However, excessive risk aversion may lead to the "financial contraction" and "strong-get-stronger" effect, which increase credit concentration, tighten financing constraints for small and medium-sized firms, and push smaller banks toward more aggressive strategies. These findings provide quantitative evidence supporting RegTech frameworks, demonstrating that regulators should establish early warning signals for financial risks based on loan success rates and network density measurements.