Fetching the paper…

On Targeted Manipulation and Deception when Optimizing LLMs for User Feedback · Around