MC-MKE: A Fine-Grained Multimodal Knowledge Editing Benchmark Emphasizing Modality Consistency

TL;DR

This paper introduces MC-MKE, a benchmark for fine-grained multimodal knowledge editing that emphasizes modality consistency, revealing limitations of current methods in correcting visual and textual errors in multimodal models.

Contribution

The paper presents MC-MKE, a new benchmark for evaluating multimodal knowledge editing, focusing on independent correction of visual and textual errors and highlighting challenges in modality consistency.

Findings

Current editing methods struggle with modality consistency.

MC-MKE enables targeted correction of visual and textual knowledge.

Evaluation reveals limitations in existing multimodal knowledge editing techniques.

Abstract

Multimodal large language models (MLLMs) are prone to non-factual or outdated knowledge issues, which can manifest as misreading and misrecognition errors due to the complexity of multimodal knowledge. Previous benchmarks have not systematically analyzed the performance of editing methods in correcting these two error types. To better represent and correct these errors, we decompose multimodal knowledge into its visual and textual components. Different error types correspond to different editing formats, which edit distinct parts of the multimodal knowledge. We present MC-MKE, a fine-grained Multimodal Knowledge Editing benchmark emphasizing Modality Consistency. Our benchmark facilitates independent correction of misreading and misrecognition errors by editing the corresponding knowledge component. We evaluate four multimodal knowledge editing methods on MC-MKE, revealing their limitations, particularly in terms of modality consistency. Our work highlights the challenges posed by multimodal knowledge editing and motivates further research in developing effective techniques for this task.