机器遗忘在视觉-语言模型中的鲁棒性研究
On the Robustness of Machine Unlearning for Vision-Language Models
AI总结：
本文首次系统调查了视觉-语言模型机器遗忘的鲁棒性，通过提出三种攻击范式揭示现有方法往往隐藏而非彻底移除目标知识。
摘要：
视觉-语言模型（VLM）可能会记忆训练数据中的不良信息，这激发了人们对机器遗忘的兴趣。在这项工作中，我们首次对VLM遗忘进行了系统调查和鲁棒性分析。我们提供了现有VLM遗忘方法的全面分类和回顾，以及在多种提示设置下的统一评估。然后，我们提出了三种攻击范式，以检验被遗忘的多模态知识是否可以通过上下文提示或下游微调重新激活。大量实验表明，许多现有方法在这些攻击下仍然脆弱，这表明当前方法往往隐藏而非完全移除目标知识。我们的研究为当前VLM遗忘方法的鲁棒性和局限性提供了新见解，并强调了需要更可靠的多模态遗忘策略。代码可在https://github.com/XMUDeepLIT/VLM-UnL-Attack获取。
英文摘要：
Vision-language models (VLMs) may memorize undesirable information from training data, motivating growing interest in machine unlearning. In this work, we present the first systematic survey and robustness analysis of VLM unlearning. We provide a comprehensive taxonomy and review of existing VLM unlearning methods, together with unified evaluations under multiple prompt settings. We then propose three attack paradigms to examine whether forgotten multimodal knowledge can be reactivated through contextual prompting or downstream retraining. Extensive experiments show that many existing methods remain vulnerable under these attacks, indicating that current approaches often hide rather than fully remove target knowledge. Our study provides new insights into the robustness and limitations of current VLM unlearning methods and highlights the need for more reliable multimodal unlearning strategies. Code is available at https://github.com/XMUDeepLIT/VLM-UnL-Attack.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
