Topic

Reinforcement Learning Security

How reward functions, agents, environments, and evaluators can be gamed, exploited, or misaligned under optimization pressure.