Home » Authors » Jesson Wang, Zhanhao Hu, Minghao Lin, Yuchen Wu, Aylin Caliskan, Mengyuan Li

Jesson Wang, Zhanhao Hu, Minghao Lin, Yuchen Wu, Aylin Caliskan, Mengyuan Li

Public reference bibliography and cataloged PDF materials authored or published by Jesson Wang, Zhanhao Hu, Minghao Lin, Yuchen Wu, Aylin Caliskan, Mengyuan Li.

Found 1 document(s) Page 1 of 1

Safety Offset Learning: When Benign Fine-Tuning Misaligns LLMs

Year: 2026 • Source: arxiv.org • Listed PDF

Safety risks from fine-tuning are commonly characterized by the harmfulness of the training data:harmful fine-tuning is considered dangerous, while benign fine-tuning is often assumed to preservealignment. We argue that ...