SPACE: 面向多模态大语言模型的无源代理锚点概念擦除
SPACE: Source-free Proxy Anchor Concept Erasure for MLLMs
AI总结：
提出SPACE框架，首个针对多模态大语言模型的无源机器遗忘方法，通过文本引导的代理锚点选择和双约束语义隔离，在不访问目标数据的情况下擦除概念，并保持模型性能。
摘要：
随着多模态大语言模型（MLLMs）面临日益增长的隐私风险和监管约束，机器遗忘（MU）已成为在保持模型性能的同时移除敏感数据的关键解决方案。然而，现有的MU方法通常依赖于目标概念的视觉数据，由于严格的数据保留政策，这些数据往往不可用，因此需要一种无需访问目标数据的无源遗忘方法。在这项工作中，我们提出了无源代理锚点概念擦除（SPACE），这是首个专门针对MLLMs的无源遗忘框架。SPACE包含两个阶段：（1）文本引导的代理锚点选择（TPAS），从共享特征空间中检索语义对齐的代理锚点；（2）双约束语义隔离（DCSI），优化这些锚点以间接擦除目标概念。DCSI将更新限制在保留知识的零空间中，确保结构完整性。我们从理论上证明，SPACE严格限制了保留知识的扰动，并最大化特征谱熵，从而保持模型性能。此外，在六个数据集上的大量实验表明，SPACE实现了与最先进的数据依赖方法相当的性能，验证了其在无源MU场景中的有效性。源代码将发布。
英文摘要：
As Multimodal Large Language Models (MLLMs) face growing privacy risks and regulatory constraints, machine unlearning (MU) has emerged as a crucial solution for removing sensitive data while preserving model performance. However, existing MU methods typically rely on visual data of the target concepts, which is often unavailable due to strict data retention policies, thus creating a demand for source-free unlearning approaches that operate without access to the target data. In this work, we propose Source-free Proxy Anchor Concept Erasure (SPACE), the first source-free unlearning framework specialized for MLLMs. SPACE consists of two stages: (1) Text-Guided Proxy Anchor Selection (TPAS), which retrieves semantically aligned proxy anchors from the shared feature space. (2) Dual-Constraint Semantic Isolation (DCSI), which optimizes these anchors to indirectly erase target concepts. DCSI confines updates to the null space of retained knowledge, ensuring structural integrity. We theoretically prove that SPACE strictly bounds the perturbation on retained knowledge and maximizes feature spectral entropy, thereby maintaining the model's performance. Furthermore, extensive experiments across six datasets show that SPACE achieves performance comparable to that of state-of-the-art data-dependent methods, validating its effectiveness in source-free MU scenarios. The source code will be released.
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。
