Reverse Kullback-Leibler (KL) distillation has shown empirical promise for training language models, but its statistical guarantees and limits remain poorly …
机构:Google
来源:arXiv 2610.30844 | AI4Papers 论文推荐平台