mlsanityData to 14 Oct 2026

Uncovering and Mitigating Transient Blindness in Multimodal Model Editing

arXiv 2511.13243 ↗First posted 17 Nov 2025cs.LG cs.AI cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Multimodal Model Editing (MMED) aims to correct erroneous knowledge in multimodal models. Existing evaluation methods, adapted from textual model editing, overstate success by relying on low-similarity or random inputs, obscure overfitting. We propose a comprehensive locality evaluation framework, covering three key dimensions: random-image locality, no-image locality, and consistent-image locality, operationalized through seven distinct data types, enabling a detailed and structured analysis of multimodal edits. We introduce De-VQA, a dynamic evaluation for visual question answering, uncovering a phenomenon we term transient blindness, overfitting to edit-similar text while ignoring visuals. Token analysis shows edits disproportionately affect textual tokens. We propose locality-aware adversarial losses to balance cross-modal representations. Empirical results demonstrate that our approach consistently outperforms existing baselines, reducing transient blindness and improving locality by 17% on average.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 22 of its 33 references that are in the map. First posted on arXiv 17 Nov 2025.

Builds on

22 of 33 references.

References · 22

  1. 2017Attention Is All You Needcited by 60,361
  2. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,170
  3. 2021Editing Factual Knowledge in Language Modelscited by 553
  4. 2022Memory-Based Model Editing at Scalecited by 447
  5. 2022Multimodal Learning with Transformers: A Surveycited by 400
  6. 2023Can We Edit Factual Knowledge by In-Context Learning?cited by 285
  7. 2023Transformer-Patcher: One Mistake worth One Neuroncited by 197
  8. 2020Editable Neural Networkscited by 155
  9. 2024Strengthening Multimodal Large Language Model with Bootstrapped Preference Optimizationcited by 75
  10. 2023Detecting Edit Failures In Large Language Models: An Improved Specificity Benchmarkcited by 62
  11. 2024Learning to Edit: Aligning LLMs with Knowledge Editingcited by 43
  12. 2024Lifelong Knowledge Editing for LLMs with Retrieval-Augmented Continuous Prompt Learningcited by 40
10 more
  1. 2024VLKEB: A Large Vision-Language Model Knowledge Editing Benchmarkcited by 30
  2. 2025MMKE-Bench: A Multimodal Editing Benchmark for Diverse Visual Knowledgecited by 23
  3. 2024Atomic Fact Decomposition Helps Attributed Question Answeringcited by 21
  4. 2025Robust Disentangled Counterfactual Learning for Physical Audiovisual Commonsense Reasoningcited by 20
  5. 2024Prompting Large Language Models with Knowledge Graphs for Question Answering Involving Long-tail Factscited by 14
  6. 2024Lifelong Knowledge Editing for Vision Language Models with Low-Rank Mixture-of-Expertscited by 14
  7. 2024ComprehendEdit: A Comprehensive Dataset and Evaluation Framework for Multimodal Knowledge Editingcited by 4
  8. 2025Long-Form Information Alignment Evaluation Beyond Atomic Factscited by 4
  9. 2025Prompting Large Language Models with Partial Knowledge for Answering Questions with Unseen Entitiescited by 4
  10. 2025Memorization neq Understanding: Do Large Language Models Have the Ability of Scenario Cognition?cited by 1

Built on by

Not cited yet.