Jailbreak-Zero: A Path to Pareto Optimal Red Teaming for Large Language Models

Kai Hu, Abhinav Aggarwal, Mehran Khodabandeh +6

This paper introduces Jailbreak-Zero, a novel red teaming methodology that shifts the paradigm of Large Language Model (LLM) safety evaluation from a constrained example-based appr…