Diversity Helps Jailbreak Large Language Models
This exposes a critical flaw in current LLM safety training for AI security, suggesting incremental improvements may mask vulnerabilities rather than eliminate them.
The researchers tackled the problem of jailbreaking large language models by developing a technique that leverages LLMs' ability to diverge from prior context, achieving up to a 62.83% higher success rate in compromising ten leading chatbots while using only 12.9% of the queries.
We have uncovered a powerful jailbreak technique that leverages large language models' ability to diverge from prior context, enabling them to bypass safety constraints and generate harmful outputs. By simply instructing the LLM to deviate and obfuscate previous attacks, our method dramatically outperforms existing approaches, achieving up to a 62.83% higher success rate in compromising ten leading chatbots, including GPT-4, Gemini, and Llama, while using only 12.9% of the queries. This revelation exposes a critical flaw in current LLM safety training, suggesting that existing methods may merely mask vulnerabilities rather than eliminate them. Our findings sound an urgent alarm for the need to revolutionize testing methodologies to ensure robust and reliable LLM security.