A new framework called VR-FraudNet forces large language models to propose structured fraud explanations that a fixed ...