---
id: 20261002-T0-10
title: "在线蒸馏中的离策略教师：一个被忽视的不对称性问题"
title_en: "On the Off-Policy Teacher in On-Policy Distillation"
url: https://ai.daily.yangsir.net/daily/20261002-T0-10
issue_date: 2026-10-02
publish_date: 2026-10-01T04:00:00.000Z
category: research
source_name: "arXiv cs.LG (ML)"
source_url: https://arxiv.org/abs/2609.38360
---

# 在线蒸馏中的离策略教师：一个被忽视的不对称性问题

研究分析了在线策略蒸馏（OPD）中的一个基本不对称性问题：学生模型从自身策略生成的轨迹中学习，但教师模型的监督信号来自不同的分布。论文探讨了这种离策略教师对蒸馏效果的影响，并分析了其中的理论机制。该工作为理解和改进在线蒸馏方法提供了新的视角。

## English Version

**On the Off-Policy Teacher in On-Policy Distillation**

A new paper analyzes a fundamental asymmetry in on-policy distillation (OPD): the student learns from trajectories generated by its own policy, while the teacher's supervision comes from a different distribution. The work examines how this off-policy teacher affects distillation outcomes and analyzes the underlying theoretical mechanisms, offering a new perspective for understanding and improving on-policy distillation.

---

**来源**：[arXiv cs.LG (ML)](https://arxiv.org/abs/2609.38360)

**详情页**：https://ai.daily.yangsir.net/daily/20261002-T0-10

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*