Linked Presentation: Making Them Ask and Answer: Jailbreaking Large Language Models in Few Queries via Disguise and Reconstruction
