Noise Contrastive Alignment of Language Models with Explicit Rewards | Read Paper on Bytez