arXiv 2609.32722 · cs.LG · 2026-09-26
Scaling Properties of Same-Family On-Policy Distillation
Yuntai Bao, Qinfeng Li, Guoqing Jiang, Liwei Chen, Zhiheng Qin, Xuanping Li, Wenqi Zhang, Xuhong Zhang
Zhejiang University
在最新榜 #90 2026-W40 第 8 名 本站收录于 2026-10-21
一句话要点
研究同族模型间在线策略蒸馏的扩展规律：训练早期准确率随 KL 距离近似线性上升。
Studies scaling of same-family on-policy distillation, finding accuracy rises roughly linearly with KL distance early in training.
阅读 arXiv 原文 PDF Hugging Face 讨论 项目页
主题与学科
同主题：高效训练与推理
- The Teacher Is a Direction, Not a Destination: Extrapolating RL-Induced Representation Residuals in On-Policy Distillation ▲511
- Beyond Domain-Routed Supervision: Cross-Teacher Guarding for Multi-Teacher On-Policy Distillation ▲251
- On-Policy or Off-Policy Learning? A Systematic Study of Distillation Dynamics ▲205
- Beyond Teacher Assignment: Domain-Normalized Multi-Teacher On-Policy Distillation ▲181
- TokenRouter: Efficient Serving System for Token-Level LLM Routing ▲143
数据说明
元数据来自 arXiv API（已核实）；点赞来自 Hugging Face 每日论文；引用来自 Semantic Scholar。最近刷新 2026-10-22。
要点由本站根据摘要自写，不转载摘要原文，请阅读原文。