Semantic Pareto-DQN: Multi-Objective RL Framework for Financial Anomaly Detection

Researchers propose Semantic Pareto-DQN, a multi-objective reinforcement learning framework designed to address extreme class imbalance and "fraud collapse" in financial anomaly detection. The system synthesizes transaction features into natural-language narratives encoded by large language models to create robust state representations.

• Optimizes a vectorial reward that decouples financial efficacy, operational friction, and semantic discovery.
• Maps the continuous Pareto frontier to dynamically navigate costs of missed anomalies versus false positives.
• Achieves superior minority-class recall compared to scalarized baselines on E-Commerce fraud and UCI Credit datasets.

The framework provides an alternative to trade bounded operational friction for financial anomaly discovery without using distortive data resampling.