Towards Meta-Cognitive Knowledge Editing for Multimodal LLMs
On the map: Multimodal LLMs / vision-language models → Image editing and quality assessment
Abstract (arXiv)
Knowledge editing enables multimodal large language models (MLLMs) to efficiently update outdated or incorrect information.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.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Image editing and quality assessment, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 42 of its 71 references that are in the map. First posted on arXiv 6 Sep 2025.
Builds on
42 of 71 references.
Closest ancestors · 6
Same direction, newest first
- 2025Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens20 Apr
- 2024Lifelong Knowledge Editing for Vision Language Models with Low-Rank Mixture-of-Experts23 Nov
- 2024Towards Unified Multimodal Editing with Enhanced Knowledge Collaboration30 Sep
- 2024MC-MKE: A Fine-Grained Multimodal Knowledge Editing Benchmark Emphasizing Modality Consistency19 Jun
- 2024VLKEB: A Large Vision-Language Model Knowledge Editing Benchmark12 Mar
- 2024MIKE: A New Benchmark for Fine-grained Multimodal Entity Knowledge Editing18 Feb
Other references · 36
- 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,297
- 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,585
- 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 4,356
- 2021Zero-Shot Text-to-Image Generationcited by 4,288
- 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,234
- 2021Prefix-Tuning: Optimizing Continuous Prompts for Generationcited by 3,915
- 2023InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuningcited by 2,876
- 2022Locating and Editing Factual Associations in GPTcited by 2,691
- 2023InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Taskscited by 2,565
- 2016Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answeringcited by 2,516
- 2015VQA: Visual Question Answeringcited by 2,506
- 2023Video-LLaMA: An Instruction-tuned Audio-Visual Language Model for Video Understandingcited by 1,471
24 more
- 2015Microsoft COCO Captions: Data Collection and Evaluation Servercited by 1,462
- 2020Transformer Feed-Forward Layers Are Key-Value Memoriescited by 1,336
- 2021Unsupervised Dense Information Retrieval with Contrastive Learningcited by 1,216
- 2019OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledgecited by 1,121
- 2022Mass-Editing Memory in a Transformercited by 1,001
- 2020What Makes for Good Views for Contrastive Learning?cited by 846
- 2021P-Tuning v2: Prompt Tuning Can Be Comparable to Fine-tuning Universally Across Scales and Taskscited by 689
- 2023LLaMA-Adapter V2: Parameter-Efficient Visual Instruction Modelcited by 602
- 2021Editing Factual Knowledge in Language Modelscited by 553
- 2021Fast Model Editing at Scalecited by 498
- 2022Memory-Based Model Editing at Scalecited by 447
- 2023MQuAKE: Assessing Knowledge Editing in Language Models via Multi-Hop Questionscited by 300
- 2023Can We Edit Factual Knowledge by In-Context Learning?cited by 285
- 2022Aging with GRACE: Lifelong Model Editing with Discrete Key-Value Adaptorscited by 271
- 2023Transformer-Patcher: One Mistake worth One Neuroncited by 197
- 2023PMET: Precise Model Editing in a Transformercited by 173
- 2022The Shapley Value in Machine Learningcited by 98
- 2023Fine-tuning Multimodal LLMs to Follow Zero-shot Demonstrative Instructionscited by 77
- 2023Can We Edit Multimodal Large Language Models?cited by 44
- 2024Auto-Encoding Morph-Tokens for Multimodal LLMcited by 30
- 2023Self-supervised Meta-Prompt Learning with Meta-Gradient Regularization for Few-shot Generalizationcited by 15
- 2025FocusDiff: Advancing Fine-Grained Text-Image Alignment for Autoregressive Visual Generation through RLcited by 13
- 2020A Baseline for Shapley Values in MLPs: from Missingness to Neutralitycited by 11
- 2024TeamLoRA: Boosting Low-Rank Adaptation with Expert Collaboration and Competitioncited by 10
Built on by
Citing papers per year: 2026: 1.
1 paper.
Similar, not cited
Close in wording, no citation either way
- 2024ComprehendEdit: A Comprehensive Dataset and Evaluation Framework for Multimodal Knowledge Editing
- 2025Benchmarking and Rethinking Knowledge Editing for Large Language Models
- 2024Navigating the Dual Facets: A Comprehensive Evaluation of Sequential Memory Editing in Large Language Models
- 2025MemEIC: A Step Toward Continual and Compositional Knowledge Editing
- 2024Knowledge Editing through Chain-of-Thought
- 2026Multimodal Knowledge Edit-Scoped Generalization for Online Recursive MLLM Editing
- 2025Believe It or Not: How Deeply do LLMs Believe Implanted Facts?
- 2024See the Unseen: Better Context-Consistent Knowledge-Editing by Noises