paper-with-me

Papers

Calibrating Teacher--Student Discrepancy for On-Policy Distillation

2026-09-18 · Qiangqiang He, Jin Li, MingCai Chen hf

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 by standard OPD during training. This issue is further exacerbated by privileged OPD, where privileged information induces larger teacher-side likelihood shifts, thereby encouraging the student to learn more of the teacher's own deviation. We introduce Calibrated On-Policy Distillation (Cal-OPD), which estimates the teacher's self-deviation region through positive and negative privileged interventions and calibrates the original teacher--student discrepancy by retaining only the component that lies beyond this region. Experiments on mathematical reasoning benchmarks show that, while retaining only about 52--65\% of the original teacher--student discrepancy as the optimization signal, Cal-OPD consistently outperforms standard OPD and its variants across model scales.

📄 PDF Abstract BibTeX arXiv:2609.21619

Code (3)

Tavish9/awesome-daily-AI-arxiv ★ 121
Valiant-Cat/hfpaper
arxivsub/arXivSub_daily_arxiv ★ 4

Tasks

Mathematical Reasoning

Similar Papers 제목 키워드 기반

Fixing the Teacher-Student Knowledge Discrepancy in Distillation

2021-03-31 · Jiangfan Han, Mengya Gao, Yujie Wang, Quanquan Li 외

Training a small student network with the guidance of a larger teacher network is an effective way to promote the performance of the student. Despite the different types, the guided knowledge used to distill is always ke…

image-classificationImage ClassificationKnowledge Distillationobject-detection+1

DASH-OPD: Discrepancy-Aware Switching with Hysteresis for On-Policy Distillation

2026-07-31 · Yuchen Xia, Qianguo Sun, Chao Song, Junlong Wu 외 arxiv

On-policy distillation (OPD) trains student models on their own rollouts to reduce exposure bias. However, in multi-turn agent scenarios, early student errors can lead a trajectory away from the teacher's familiar domain…

On-Policy Adversarial Flow Distillation for Autoregressive Video Generation

2026-05-25 · Yang Luo, Shengju Qian, Xiaohang Tang, Zirui Zhu 외 arxiv

Autoregressive video generators are attractive for streaming, long-horizon, and interactive applications, but distilling strong black-box teachers into causal students remains difficult. The student must learn under its …

Reinforcement LearningVideo Generation

RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement Learning

2026-09-17 · Yan Yu, Zhengxi Lu, Yizhou Liu, Yichen Pan 외 hf

Multi-turn agents trained with reinforcement learning (RL) receive a single scalar reward per trajectory, which motivates self on-policy distillation (OPD) to supply dense token-level supervision from a self-teacher with…

Reinforcement Learning

Does Knowledge Distillation Really Work?

2021-06-10 · NeurIPS 2021 12 · Samuel Stanton, Pavel Izmailov, Polina Kirichenko, Alexander A. Alemi 외

Knowledge distillation is a popular technique for training a small student network to emulate a larger teacher model, such as an ensemble of networks. We show that while knowledge distillation can improve student general…

Knowledge Distillation