1
Trajectory-Refined Distillation
提出轨迹精炼蒸馏新方法，提升大模型知识迁移效果
arXiv:2606.08432v1 Announce Type: new Abstract: On-policy distillation (OPD) has become a central post-training tool for large language models (LLMs),…
提出轨迹精炼蒸馏新方法，提升大模型知识迁移效果
arXiv:2606.08432v1 Announce Type: new Abstract: On-policy distillation (OPD) has become a central post-training tool for large language models (LLMs),…