在海量网络语料上训练的大语言模型能够记忆并复现敏感或私有数据，从而引发法律和伦理方面的担忧。遗忘（unlearning），即通过调优模型使其忘记训练数据中存在的信息，为我们在训练之后保护私有数据提供了一种途径。尽管目前已存在若干用于此类遗忘的方法，但尚不清楚它们在多大程度上能够使模型等同于那些从一开始就从未学习过待遗忘数据的模型。为应对这一挑战，我们提出了 TOFU，即虚构遗忘任务（Task of Fictitious Unlearning），作为一个基准，旨在帮助加深我们对遗忘问题的理解。我们提供了一个包含 200 个多样化合成作者档案的数据集，每个档案由 20 个问答对组成，并将这些档案的一个子集称为遗忘集（forget set），作为遗忘的目标。我们构建了一套协同工作的评估指标，以提供遗忘效果的整体图景。最后，我们给出了一组来自现有遗忘算法的基线结果。重要的是，我们所考虑的所有基线均未表现出有效的遗忘，这促使人们继续努力开发能够有效调优模型的遗忘方法，使模型真正表现得如同从未在遗忘数据上进行过训练一样。
英文摘要：
Large language models trained on massive corpora of data from the web can memorize and reproduce sensitive or private data raising both legal and ethical concerns. Unlearning, or tuning models to forget information present in their training data, provides us with a way to protect private data after training. Although several methods exist for such unlearning, it is unclear to what extent they result in models equivalent to those where the data to be forgotten was never learned in the first place. To address this challenge, we present TOFU, a Task of Fictitious Unlearning, as a benchmark aimed at helping deepen our understanding of unlearning. We offer a dataset of 200 diverse synthetic author profiles, each consisting of 20 question-answer pairs, and a subset of these profiles called the forget set that serves as the target for unlearning. We compile a suite of metrics that work together to provide a holistic picture of unlearning efficacy. Finally, we provide a set of baseline results from existing unlearning algorithms. Importantly, none of the baselines we consider show effective unlearning motivating continued efforts to develop approaches for unlearning that effectively tune models so that they truly behave as if they were never trained on the forget data at all.
Attention Smoothing Is All You Need For Unlearning
Department of Computer Science, Wayne State University（韦恩州立大学计算机科学系）；Department of Computer Science and Engineering, Michigan State University（密歇根州立大学计算机科学与工程系）；Institute for AI and Data Science, Wayne State University（韦恩州立大学人工智能与数据科学研究所）
Replay What Matters: Off-Policy Replay for Efficient LLM Reinforcement Unlearning
The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州））；Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所）；University of Glasgow（格拉斯哥大学）
From Anchors to Supervision: Memory-Graph Guided Corpus-Free Unlearning for Large Language Models
College of Computer Science and Artificial Intelligence, Fudan University（复旦大学计算机科学与人工智能学院）；Shanghai University of Electric Power（上海电力大学）；School of Information Science and Engineering, East China University of Science and Technology（东华大学信息科学与工程学院）
Soft Token Attacks Cannot Reliably Audit Unlearning in Large Language Models
Ludwig Maximilian University of Munich（慕尼黑路德维希-马克西米利安大学）；Munich Center for Machine Learning (MCML)（慕尼黑机器学习中心）；Aalto University（艾尔沃斯大学）；Intel Labs（英特尔实验室）
UniErase: Towards Balanced and Precise Unlearning in Language Models
University of Science and Technology of China（中国科学技术大学）；University of the Chinese Academy of Sciences（中国科学院大学）；National University of Singapore（新加坡国立大学）；Nanyang Technological University（南洋理工大学）；A*STAR（科技研究局）；Zhejiang University（浙江大学）
GUARD: Generation-time LLM Unlearning via Adaptive Restriction and Detection
The Hong Kong University of Science and Technology (Guangzhou)（香港科学与技术大学（广州））；UC Santa Cruz（加州大学圣克鲁兹分校）；UIUC（伊利诺伊大学香槟分校）；Southeast University（东南大学）；BIAI, ZJUT（脑科学与类脑智能研究院，浙江大学）；D5Data.ai
Storage Is Not Strategy: State-Conditioned Support Control for LLM Unlearning
Southeast University（东南大学）；University of Sydney（悉尼大学）；Northwestern University（西北大学）；Together AI；The Chinese University of Hong Kong Shenzhen（香港中文大学（深圳））
Wisdom is Knowing What not to Say: Hallucination-Free LLMs Unlearning via Attention Shifting
Faculty of Information Technology, Monash University（墨尔本大学信息科技学院）；Key Laboratory of Computing Power Network and Information Security, Ministry of Education, Shandong Computer Science Center, Qilu University of Technology (Shandong Academy of Sciences)（教育部计算能力网络与信息安全重点实验室，山东计算机科学中心，齐鲁大学（山东科学院））；Shandong Provincial Key Laboratory of Computing Power Internet and Service Computing, Shandong Fundamental Research Center for Computer Science（山东省计算能力互联网与服务计算重点实验室，山东省计算机科学基础研究中心）；School of Computer Science and Technology, Anhui University（安徽大学计算机科学与技术学院）
ToFU: Transforming How Federated Learning Systems Forget User Data
School of Computer Science（计算机科学学院）；Statistics, Trinity College Dublin, Ireland（统计学，都柏林信任学院，爱尔兰）；School of Computer Science, University College Dublin, Ireland（计算机科学学院，都柏林大学学院，爱尔兰）
Consistency-Aware Editing for Entity-level Unlearning in Language Models
Shanxi University, China（山西大学）；Cardiff University, UK（卡地夫大学）；Singapore University of Technology and Design, Singapore（新加坡科技设计大学）；ILCC, School of Informatics, University of Edinburgh, Edinburgh, UK（爱丁堡大学信息学院ILCC）
On the Hidden Costs of Counterfactual Knowledge Training in LLM Unlearning
Beijing University of Posts and Telecommunications（北京邮电大学）；Huazhong University of Science and Technology（华中科技大学）；Shandong University（山东大学）；NLPR, MAIS, Institute of Automation, Chinese Academy of Sciences（神经网络计划、人工智能研究所、中国科学院自动化研究所）
Inference-Time Machine Unlearning via Gated Activation Redirection
MALTA, Machine Learning Theory and Applications Lab, PUCRS, Porto Alegre, Brazil（MALTA机器学习理论与应用实验室，PUCRS，波士顿-阿尔格雷，巴西）；Harvard University（哈佛大学）；Kunumi Institute, Brazil（库努米研究所，巴西）
