arXiv 2605.088002026-05-09cs.CVcs.AI

PPU-Bench: 用于视觉语言模型个性化部分遗忘的现实世界基准

PPU-Bench:Real World Benchmark for Personalized Partial Unlearning in Vision Language Models

Harbin Institute of Technology（哈尔滨工业大学）；Pengcheng Laboratory（鹏城实验室）；The Hong Kong Polytechnic University（香港理工大学）；Sichuan University（四川大学）；Zhejiang Normal University（浙江师范大学）

Jiahui Guang, Zexun Zhan, Zhenlin Xu, Cuiyun Gao, Haiyan Wang, Jing Li, Zhaoquan Gu, Yanchun Zhang

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出PPU-Bench，一个无需微调的现实世界基准，用于评估视觉语言模型中个性化部分遗忘的效果，通过24K多模态和单模态样本测试遗忘与保留的平衡及跨模态一致性。

摘要：

多模态大语言模型（MLLMs）在预训练过程中可能记忆敏感的跨模态信息。然而，现有MLLM遗忘基准依赖于合成知识注入或完整主体级删除，无法捕捉现实中的个性化删除请求，需要细粒度的事实控制。本文引入PPU-Bench，一个用于视觉语言模型个性化部分遗忘的现实世界和无需微调的基准。PPU-Bench包含24K多模态和单模态样本，源自500位公众人物的现有知识，在三个渐进挑战性设置下生成：完全、选择性和个性化遗忘。该基准评估方法是否能移除目标知识同时保持非目标事实、模型效用和跨模态一致性。广泛实验表明，完全遗忘往往抑制视觉身份而非事实知识，而选择性和个性化遗忘暴露了显著的遗忘-保留权衡和跨主体事实边界挑战。在跨图像和提示基攻击下的鲁棒性分析揭示了不同遗忘设置下的不同脆弱性。受这些发现启发，我们提出边界感知优化（BAO），明确建模跨主体遗忘-保留边界。在两种代表性方法上的实验结果表明，BAO能有效强制跨主体事实边界。

英文摘要：

Multimodal Large Language Models (MLLMs) may memorize sensitive cross-modal information during pretraining. However, existing MLLM unlearning benchmarks rely on synthetic knowledge injection or complete subject-level deletion, which fail to capture realistic, personalized deletion requests that require fine-grained factual control. In this paper, we introduce PPU-Bench, a real-world and fine-tuning-free benchmark for personalized partial unlearning in MLLMs. PPU-Bench contains 24K multimodal and unimodal samples derived from pre-existing knowledge of 500 public figures under three progressively challenging settings: Complete, Selective, and Personalized unlearning. The benchmark evaluates whether methods can remove target knowledge while preserving non-target facts, model utility, and cross-modal consistency. Extensive experiments show that Complete Unlearning often suppresses visual identity rather than factual knowledge, while Selective and Personalized Unlearning expose significant forget--retain trade-offs and challenges in intra-subject factual boundaries. Robustness analysis under cross-image and prompt-based attacks reveals distinct vulnerabilities across different unlearning settings. Motivated by these findings, we propose Boundary-Aware Optimization (BAO), which explicitly models intra-subject forget-retain boundaries. Experimental results on two representative methods demonstrate that BAO can effectively enforce intra-subject factual boundaries.

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。