The Policy Cliff: A Theoretical Analysis of Reward-Policy Maps in Large Language Models

Open in new window