A Training-Free Trick Lets LLMs Shift Personality at Inference Time by Splitting the Reward Signal in Two

A Training-Free Trick Lets LLMs Shift Personality at Inference Time by Splitting the Reward Signal in Two | BedrockNews