aiqbench analysis
This research addresses the critical privacy challenge in multimodal large language models by introducing attribute-level unlearning, allowing specific sensitive details to be removed without retraining or deleting entire identities. The proposed CLRP framework achieves this through activation patching and retain-aware projection, effectively mitigating residual leakage while preserving general model utility across diverse architectures.
Toward Fine-Grained Forgetting:Attribute Unlearning for Multimodal Large Language Models
arXiv:2608.01008v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) exhibit strong vision--language capabilities but may also memorize and disclose sensitive information. Machine unlearning seeks to remove designated knowledge without retraining from scratch while preserving general utility. Existing privacy-oriented benchmarks primarily adopt profile-level deletion, whereas practical requests are often finer grained: a model should forget a specified attribute while retaining non-sensitive information about the same identity. We therefore introduce attribute-level MLLM unlearning as a finer-grained task and construct a benchmark spanning long-text, numeric, and short-text targets, multiple forget ratios, and diverse question types. Our evaluation reveals that target and retained attributes share identity-specific and visual evidence, making selective forgetting susceptible to residual leakage or collateral degradation; accordingly, existing methods exhibit unstable forgetting--retention trade-offs in this setting. To address this challenge, we propose Causal Localization and Retain-Aware Projection (CLRP), a lightweight training-free framework. CLRP uses activation patching to identify the layer that causally mediates target-attribute disclosure, then applies a retain-aware projection that removes the target-attribute subspace while preserving same-identity evidence. Experiments across multiple widely used MLLMs with distinct architectures and parameter scales demonstrate the effectiveness of CLRP.
Abstract:Multimodal large language models (MLLMs) exhibit strong vision--language capabilities but may also memorize and disclose sensitive information. Machine unlearning seeks to remove designated knowledge without retraining from scratch while preserving general utility. Existing privacy-oriented benchmarks primarily adopt profile-level deletion, whereas practical requests are often finer grained: a model should forget a specified attribute while retaining non-sensitive information about the same identity. We therefore introduce attribute-level MLLM unlearning as a finer-grained task and construct a benchmark spanning long-text, numeric, and short-text targets, multiple forget ratios, and diverse question types. Our evaluation reveals that target and retained attributes share identity-specific and visual evidence, making selective forgetting susceptible to residual leakage or collateral degradation; accordingly, existing methods exhibit unstable forgetting--retention trade-offs in this setting. To address this challenge, we propose Causal Localization and Retain-Aware Projection (CLRP), a lightweight training-free framework. CLRP uses activation patching to identify the layer that causally mediates target-attribute disclosure, then applies a retain-aware projection that removes the target-attribute subspace while preserving same-identity evidence. Experiments across multiple widely used MLLMs with distinct architectures and parameter scales demonstrate the effectiveness of CLRP.
Artificial Intelligence (cs.AI)
arXiv:2608.01008 [cs.AI]
(or arXiv:2608.01008v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2608.01008 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 275 words
Related coverage
Reflections on unlearning and inoculation
LessWrong · Sep 20, 2026
[AINews] High Taste Testers Ep. 1
Latent Space · Oct 17, 2026
Mixture of Layers: Dynamic Layer Routing for Visual Reasoning
arXiv · Oct 8, 2026
Exploring and Bridging Knowledge Holes in Unlearned Multimodal Large Language Models
arXiv · Aug 5, 2026