JBDistill Generates Its Own Jailbreaks - 81.8% Attack Rate
Johns Hopkins and Microsoft's JBDistill achieves 81.8% attack success rate across 13 LLMs by auto-generating fresh adversarial prompts on demand.
Microsoft research
Johns Hopkins and Microsoft's JBDistill achieves 81.8% attack success rate across 13 LLMs by auto-generating fresh adversarial prompts on demand.
