arXiv cs.CL· Fengli Wu, Vaidehi Patil, Jaehong Yoon, Yue Zhang, Mohit Bansal·· 12 小时前AI 评分31
MedForget：面向医疗 AI 的层级感知多模态遗忘基准与 CHIP 方法
Hierarchy-Aware Multimodal Unlearning for Medical AI
AI 导读
研究提出 MedForget——一个将医疗机构数据建模为嵌套结构的层级感知多模态遗忘基准，可对 retain/forget 划分做细粒度评估。实验显示现有遗忘方法难以在不损失保留效用的前提下实现层级感知遗忘。为此提出免训练的 CHIP 方法，通过选择性移除目标专属权重子空间、保留同级共享信息，在所有层级上取得最高 forget-retain 性能差距，并保持有竞争力的下游效用。
正文 · 原文
Abstract:Multimodal large language models (MLLMs) are increasingly used in sensitive domains such as medical AI, where privacy regulations, including HIPAA and GDPR, require the removal of specific individuals' or institutions' data. This motivates machine unlearning, which aims to remove the influence of target data from a trained model. However, existing unlearning benchmarks fail to reflect the hierarchical and multimodal structure of real-world medical data, limiting their ability to properly evaluate unlearning in practice. Therefore, we introduce MedForget, a hierarchy-aware multimodal unlearning benchmark that models data from medical institutions as a nested structure, enabling fine-grained evaluation across retain and forget splits. Experiments show that current unlearning methods struggle to achieve effective hierarchy-aware forgetting without degrading retained utility, measured by performance on clinically relevant prediction tasks. To address this limitation, we propose Cross-modal Hierarchy-Informed Projection for unlearning (CHIP), a training-free, hierarchy-aware multimodal unlearning method that deletes information by selectively removing target-specific weight subspaces while preserving sibling-shared information. Our results show that CHIP achieves the highest forget-retain performance gap across all hierarchy levels while maintaining competitive downstream utility compared to existing methods. Overall, MedForget provides a practical benchmark for evaluating hierarchy-aware multimodal unlearning for medical data, while CHIP offers an effective solution that generalizes across architectures and balances deletion with utility.
| Comments: | Published at TMLR 2026, Dataset and Code: this https URL, First two authors contributed equally |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2512.09867 [cs.CV] |
| (or arXiv:2512.09867v4 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2512.09867 arXiv-issued DOI via DataCite |
Submission history
From: Vaidehi Patil [view email]
[v1]
Wed, 10 Dec 2025 17:55:06 UTC (2,247 KB)
[v2]
Tue, 20 Jan 2026 03:41:34 UTC (4,224 KB)
[v3]
Fri, 23 Jan 2026 02:43:59 UTC (4,223 KB)
[v4]
Sat, 31 Oct 2026 00:35:42 UTC (4,318 KB)
来源：arXiv cs.CL · arxiv.org