1 paper · 1 filter
Lars Malmqvist
This study reveals how frontier Large Language Models LLMs can "game the system" when faced with impossible situations, a critical security and alignment concern. Using a novel tex…