捷径与边界：可验证的 RL 失准与上下文安全学习

Reinforcement learning (RL) can improve an agent's task capability while reinforcing behavior that violates contextual execution requirements. We study this reward-driven misalignment, including reward hacking, through verifiable reward--context conflicts, where compliant task completion remains feasible while some exe

阅读原文 · Xinjie Shen