SR-Fraud: An Outcome-Supervised Reflective LLM Agent Framework for Non-Stationary Payment Fraud Detection
SR-Fraud uses a reflective LLM agent to detect emerging payment-fraud bursts before labels mature.
Researchers present SR-Fraud, an outcome-supervised reflective LLM framework for non-stationary payment-fraud detection. A frozen stateless agent scores each transaction from a Hybrid Episodic Window, while an offline reflection agent proposes hypotheses from matured errors that a deterministic verifier must accept. On a production benchmark, SR-Fraud improved every detection metric over its frozen agent and posted higher point estimates than static and periodically retrained CatBoost, including detection of an emerging fraud burst.
- Separates real-time transaction scoring from offline adaptation
- A verifier admits only supported hypotheses into executable knowledge
- Beats its frozen agent and retrained CatBoost on a production benchmark
- Detects an emerging fraud burst before periodic retraining
Full article133 words · extracted from arxiv.org · click to collapse
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.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.27287