EVINCE: Optimizing Adversarial LLM Dialogues via Conditional Statistics and Information Theory

Open in new window