Towards Meta-Cognitive Knowledge Editing for Multimodal LLMs

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Fan, Zhaoyu, Pan, Kaihang, Zhou, Mingze, Qin, Bosheng, Li, Juncheng, Zhang, Shengyu, Zhang, Wenqiao, Tang, Siliang, Wu, Fei, Zhuang, Yueting
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866916937534013440
author Fan, Zhaoyu
Pan, Kaihang
Zhou, Mingze
Qin, Bosheng
Li, Juncheng
Zhang, Shengyu
Zhang, Wenqiao
Tang, Siliang
Wu, Fei
Zhuang, Yueting
author_facet Fan, Zhaoyu
Pan, Kaihang
Zhou, Mingze
Qin, Bosheng
Li, Juncheng
Zhang, Shengyu
Zhang, Wenqiao
Tang, Siliang
Wu, Fei
Zhuang, Yueting
contents Knowledge editing enables multimodal large language models (MLLMs) to efficiently update outdated or incorrect information. However, existing benchmarks primarily emphasize cognitive-level modifications while lacking a focus on deeper meta-cognitive processes. To bridge this gap, we introduce CogEdit, a novel benchmark designed to evaluate MLLMs' meta-cognitive knowledge editing abilities across three levels: (1) Counterfactual-Driven Editing, assessing self-awareness of knowledge correctness changes; (2) Boundary Constraint Editing, ensuring appropriate generalization without unintended interference; and (3) Noise-Robust Editing, promoting reflective evaluation of uncertain information. To advance meta-cognitive editing, we propose MIND (Meta-cognitive INtegrated Dynamic Knowledge Editing), a framework that constructs a meta-knowledge memory for self-awareness, employs game-theoretic interactions to monitor knowledge activation, and incorporates label refinement for noise-robust updates. Extensive experiments show that MIND significantly outperforms existing cognitive editing approaches, achieving strong performance on both traditional and meta-cognitive knowledge editing benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2509_05714
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Towards Meta-Cognitive Knowledge Editing for Multimodal LLMs
Fan, Zhaoyu
Pan, Kaihang
Zhou, Mingze
Qin, Bosheng
Li, Juncheng
Zhang, Shengyu
Zhang, Wenqiao
Tang, Siliang
Wu, Fei
Zhuang, Yueting
Artificial Intelligence
Computer Vision and Pattern Recognition
Knowledge editing enables multimodal large language models (MLLMs) to efficiently update outdated or incorrect information. However, existing benchmarks primarily emphasize cognitive-level modifications while lacking a focus on deeper meta-cognitive processes. To bridge this gap, we introduce CogEdit, a novel benchmark designed to evaluate MLLMs' meta-cognitive knowledge editing abilities across three levels: (1) Counterfactual-Driven Editing, assessing self-awareness of knowledge correctness changes; (2) Boundary Constraint Editing, ensuring appropriate generalization without unintended interference; and (3) Noise-Robust Editing, promoting reflective evaluation of uncertain information. To advance meta-cognitive editing, we propose MIND (Meta-cognitive INtegrated Dynamic Knowledge Editing), a framework that constructs a meta-knowledge memory for self-awareness, employs game-theoretic interactions to monitor knowledge activation, and incorporates label refinement for noise-robust updates. Extensive experiments show that MIND significantly outperforms existing cognitive editing approaches, achieving strong performance on both traditional and meta-cognitive knowledge editing benchmarks.
title Towards Meta-Cognitive Knowledge Editing for Multimodal LLMs
topic Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.05714