Lucca Machado da Silva

The paper proposes Semantic Pareto-DQN, a multi-objective reinforcement learning framework for financial anomaly detection using large language models and natural-language narratives, achieving superior minority-class recall.