Search papers, labs, and topics across Lattice.
The paper introduces Localized and Disentangled Knowledge Editing (LDKE) to address the limitations of existing Multimodal Knowledge Editing (MKE) methods, which struggle with generalization and unintended side effects. LDKE employs a Fast Localization module to identify and update critical model layers and a Disentanglement Classifier to route inputs, preserving unrelated knowledge. Experiments demonstrate that LDKE achieves superior performance in propagating edits to related contexts while maintaining locality across various benchmarks and MLLMs.
MLLM knowledge editing can be surgically precise: LDKE propagates edits to related contexts while preventing unintended alterations to visually or semantically linked information.
Existing methods in Multimodal Knowledge Editing (MKE) have advanced the ability to correct outdated or inaccurate knowledge in Multimodal Large Language Models (MLLMs). However, they exhibit a critical limitation: while effectively modifying target factual pairs, they fail to generalize edits to logically related queries and often cause unintended alterations to unrelated but visually or semantically linked information. We identify and formalize two underlying failure modes causing this issue: Causal Misalignment, which confines edits to the specific sample, and Feature Entanglement, which causes unintended alterations to coupled but irrelevant information. To address these issues, we propose Localized and Disentangled Knowledge Editing (LDKE), a new framework that achieves precise and generalized editing by localizing fact-specific model layers and disentangling target-relevant inputs from irrelevant ones. Our approach introduces a Fast Localization module to identify and update critical layers efficiently, along with a Disentanglement Classifier that routes inputs appropriately to preserve unrelated knowledge. Extensive experiments across various benchmarks and MLLMs demonstrate that LDKE achieves superior performance in propagating edits to related contexts while maintaining high locality.
