arXiv 2609.29233 · cs.CL · 2026-09-24
Post-Training Leaves Behavioral Shadows on Unrelated Decisions
Ziyang Zhang, Yubin Jing, Yuanhao Zeng, Yuyao Li, Haofan Wang, Yichen Gong
Peking University
在最新榜 #35 2026-W40 第 10 名 本站收录于 2026-10-21
一句话要点
发现后训练会在无关输出上留下“行为影子”，每个提示只需教师一个词就能把能力迁移给学生。
Shows post-training leaves behavioural shadows in unrelated outputs; a single teacher word per prompt can transfer capabilities to a student.
阅读 arXiv 原文 PDF Hugging Face 讨论 代码仓库
主题与学科
同主题：大语言模型
- Jeva: System One Models from First Principles ▲371
- OmniEdu: Open Foundation Models for Learning and Teaching ▲240
- PAR: Progress-Aligned Recurrence for Length Generalization in State Tracking ▲148
- MemT-1.0: Towards Memory-Native Evolution ▲119
- MiMo-V2.6: Scaling Reinforcement Learning Towards Self-Improvement ▲92
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。