**Calibrating Teacher--Student Discrepancy for On-Policy Distillation**

> On-policy distillation (OPD) improves reasoning models by learning the token-level discrepancy between a stronger teacher and an on-policy student. However, this discrepancy does not purely reflect the capability gap between the teacher and the student: it also contains deviations arising from the teacher itself, which are consequently mixed into the observed teacher--student discrepancy and indiscriminately learned...

**来源信息**
- **来源**：HuggingFace Daily Papers（社区热门论文）
- **分类**：行业
- **发布时间**：2026-09-21 08:00（北京时间）
- **原文**：[打开原文](https://huggingface.co/papers/2609.21619)