On the Robustness of Reward Models for Language Model Alignment | Read Paper on Bytez