AutoDAN-Reasoning: Enhancing Strategies Exploration based Jailbreak Attacks with Test-Time Scaling

Recent advancements in jailbreaking large language models (LLMs), such as AutoDAN-Turbo, have demonstrated the power of automated strategy discovery. AutoDAN-Turbo employs a lifelong learning agent to build a rich library of attack strategies from scratch. While highly effective, its test-time generation process involves sampling a strategy and generating a single corresponding attack prompt, which may not fully exploit the potential of the learned strategy library. In this paper, we propose to further improve the attack performance of AutoDAN-Turbo through test-time scaling. We introduce two distinct scaling methods: Best-of-N and Beam Search. The Best-of-N method generates N candidate attack prompts from a sampled strategy and selects the most effective one based on a scorer model. The Beam Search method conducts a more exhaustive search by exploring combinations of strategies from the library to discover more potent and synergistic attack vectors. According to the experiments, the proposed methods significantly boost performance, with Beam Search increasing the attack success rate by up to 15.6 percentage points on Llama-3.1-70B-Instruct and achieving a nearly 60% relative improvement against the highly robust GPT-o4-mini compared to the vanilla method.

Key Contributions

• Best-of-N scaling: generates N candidate jailbreak prompts from a sampled strategy and selects the most effective via a scorer model
• Beam Search scaling: exhaustively searches combinations of strategies from the AutoDAN-Turbo library to discover synergistic attack vectors
• Demonstrated up to 15.6 pp attack success rate improvement on Llama-3.1-70B-Instruct and ~60% relative gain against GPT-o4-mini

🛡️ Threat Analysis

Details

Similar Papers