1 paper · 1 filter
Niklas Pfister, Václav Volhejn, Manuel Knott +23
Current evaluations of defenses against prompt attacks in large language model (LLM) applications often overlook two critical factors: the dynamic nature of adversarial behavior an…