Automatic LLM Red Teaming

Red teaming is critical for identifying vulnerabilities and building trust in current LLMs. However, current automated methods for Large Language Models (LLMs) rely on brittle prompt templates or single-turn attacks, failing to capture the complex, interactive nature of real-world adversarial dialogues. We propose a novel paradigm: training an AI to strategically `break' another AI. By formalizing red teaming as a Markov Decision Process (MDP) and employing a hierarchical Reinforcement Learning (RL) framework, we effectively address the inherent sparse reward and long-horizon challenges. Our generative agent learns coherent, multi-turn attack strategies through a fine-grained, token-level harm reward, enabling it to uncover subtle vulnerabilities missed by existing baselines. This approach sets a new state-of-the-art, fundamentally reframing LLM red teaming as a dynamic, trajectory-based process (rather than a one-step test) essential for robust AI deployment.

Key Contributions

First formalization of multi-turn LLM red teaming as a Markov Decision Process (MDP), enabling value-based sequential adversarial dialogue planning.
Hierarchical RL framework separating utterance-level strategy (high-level policy) from token-level generation (low-level policy) to handle sparse, long-horizon rewards.
Token-level marginal contribution reward via token masking for fine-grained credit assignment, enabling the agent to identify which tokens drive jailbreak success.