发表机构
Seoul National University; AIIS, Seoul National University（首尔大学; 首尔大学AIIS）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出多语言机器遗忘基准$\mu^2$-Bench，覆盖多语言并评估训练与保留语言，证明有效遗忘需考虑多语言特性。
AI 中文摘要
有害内容和私人数据等不良信息通过直接训练和间接跨语言传播在多语言大语言模型（LLMs）中扩散。多语言机器遗忘（MMU）旨在移除此类信息，然而其评估仍未被充分探索，导致尚不清楚遗忘是否真正在所有语言中消除了目标知识。为填补这一空白，我们引入了$\mu^2$-Bench，一个模拟跨多种语言的记忆、遗忘和评估全流程的MMU基准。它1）覆盖广泛的语言集合，2）在训练语言和保留语言上均进行评估，3）评估分散在多种语言中的知识。我们证明，成功的MMU需要反映多语言特征的方法，并进行分析以提供对MMU更深入的见解。
英文摘要
Undesired information such as harmful content and private data propagates through Multilingual Large Language Models (LLMs) via direct training and indirect cross-linguistic spread. Multilingual Machine Unlearning (MMU) aims to remove such information, yet its evaluation remains underexplored, leaving unclear whether unlearning truly eliminates target knowledge across all languages. To bridge this gap, we introduce $μ^2$-Bench, an MMU benchmark that simulates the full pipeline of memorization, unlearning, and evaluation across diverse languages. It 1) spans a broad set of languages, 2) evaluates on both training and hold-out languages, and 3) assesses knowledge as dispersed across multiple languages. We show that successful MMU requires methods that reflect multilingual characteristics, and conduct analysis to provide deeper insights into MMU.
