Payment fraud detection improves with reflective large language model agent

SR-Fraud: An Outcome-Supervised Reflective LLM Agent Framework for Non-Stationary Payment Fraud Detection

Machine LearningCryptography and Security

Summary

Detecting payment fraud quickly is hard because fraudsters keep changing their tricks, and systems often wait too long to update their rules. The authors propose SR-Fraud, which uses a special kind of AI agent that learns from past mistakes and reflects on patterns without slowing down transactions. This agent can spot sudden bursts of new fraud faster than traditional methods that retrain periodically. Their tests show better fraud detection and faster adaptation on real payment data.

What this means in practice

Authors

Xuwei Tan, Yao Ma, Xueru Zhang

Abstract

Real-time payment fraud detection is a non-stationary streaming prediction problem: adversaries adapt before supervised labels mature, and localized burst attacks can cause losses before retraining. Production systems typically rely on tabular classifiers and rules, which can struggle to capture these emerging sequential patterns before periodic retraining occurs. We present SR-Fraud, an outcome-supervised reflective LLM framework that decouples request-time decisions from offline adaptation. A frozen, stateless agent scores each transaction from a Hybrid Episodic Window to track behavioral shifts, while an offline reflection agent proposes boundary hypotheses from matured errors. A deterministic verifier then admits only supported hypotheses into an executable knowledge state. On a production payment-fraud benchmark, SR-Fraud improves all detection metrics over its frozen decision agent, obtains higher point estimates than static and periodically retrained CatBoost, and detects an emerging fraud burst.