Large language models often acquire specialized capabilities through successive pretraining stages, but these gains can come at the cost of forgetting earlie…
机构:清华
来源:arXiv 2610.32141 | AI4Papers 论文推荐平台