On-policy distillation (OPD) supervises student-generated responses using sampled-token student-teacher log-ratios. A positive log-ratio penalizes a token th…
机构:南大
来源:arXiv 2610.28194 | AI4Papers 论文推荐平台