LLM agents boost explainability in financial fraud detection systems

Researchers combine gradient boosting, graph features, SHAP explanations, and bounded LLM agents to build auditable fraud detection that scales with transaction volume while remaining reviewable fo...