论文领读202408[Paper Reading] Less is more for Alignment[Paper Reading] Less is more for Alignment通常情况下,LLM 会经过两个阶段: 在大规模无监督的数据上进行预训练,