Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Liu, Kangwei, Wang, Mengru, Luo, Yujie, Yuan, Lin, Sun, Mengshu, Liang, Lei, Zhang, Zhiqiang, Zhou, Jun, Hooi, Bryan, Deng, Shumin
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2503.19041
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915685137907712
author Liu, Kangwei
Wang, Mengru
Luo, Yujie
Yuan, Lin
Sun, Mengshu
Liang, Lei
Zhang, Zhiqiang
Zhou, Jun
Hooi, Bryan
Deng, Shumin
author_facet Liu, Kangwei
Wang, Mengru
Luo, Yujie
Yuan, Lin
Sun, Mengshu
Liang, Lei
Zhang, Zhiqiang
Zhou, Jun
Hooi, Bryan
Deng, Shumin
contents Fine-tuning enables large language models (LLMs) to adapt to specific domains, but often compromises their previously established safety alignment. To mitigate the degradation of model safety during fine-tuning, we introduce LookAhead Tuning, a lightweight and effective data-driven approach that preserves safety during fine-tuning. The method introduces two simple strategies that modify training data by previewing partial answer prefixes, thereby minimizing perturbations to the model's initial token distributions and maintaining its built-in safety mechanisms. Comprehensive experiments demonstrate that LookAhead Tuning effectively maintains model safety without sacrificing robust performance on downstream tasks. Our findings position LookAhead Tuning as a reliable and efficient solution for the safe and effective adaptation of LLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2503_19041
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LookAhead Tuning: Safer Language Models via Partial Answer Previews
Liu, Kangwei
Wang, Mengru
Luo, Yujie
Yuan, Lin
Sun, Mengshu
Liang, Lei
Zhang, Zhiqiang
Zhou, Jun
Hooi, Bryan
Deng, Shumin
Computation and Language
Artificial Intelligence
Computer Vision and Pattern Recognition
Machine Learning
Multimedia
Fine-tuning enables large language models (LLMs) to adapt to specific domains, but often compromises their previously established safety alignment. To mitigate the degradation of model safety during fine-tuning, we introduce LookAhead Tuning, a lightweight and effective data-driven approach that preserves safety during fine-tuning. The method introduces two simple strategies that modify training data by previewing partial answer prefixes, thereby minimizing perturbations to the model's initial token distributions and maintaining its built-in safety mechanisms. Comprehensive experiments demonstrate that LookAhead Tuning effectively maintains model safety without sacrificing robust performance on downstream tasks. Our findings position LookAhead Tuning as a reliable and efficient solution for the safe and effective adaptation of LLMs.
title LookAhead Tuning: Safer Language Models via Partial Answer Previews
topic Computation and Language
Artificial Intelligence
Computer Vision and Pattern Recognition
Machine Learning
Multimedia
url https://arxiv.org/abs/2503.19041