LLM Unlearning via Loss Adjustment with Only Forget Data

Unlearning in Large Language Models (LLMs) is essential for ensuring ethical and responsible AI use, especially in addressing privacy leak, bias, safety, and evolving regulations. Existing approaches to LLM unlearning often rely on retain data or a reference LLM, yet they struggle to adequately bala...

Ausführliche Beschreibung

Bibliographische Detailangaben

Veröffentlicht in:	arXiv.org 2024-10
Hauptverfasser:	Wang, Yaxuan, Jiaheng Wei, Liu, Chris Yuhao, Pang, Jinlong, Liu, Quan, Ankit Parag Shah, Bao, Yujia, Liu, Yang, Wei, Wei
Format:	Artikel
Sprache:	eng
Schlagworte:	Datasets Large language models Tofu
Online-Zugang:	Volltext
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!