MLLM Unlearning

MLLM: Multimodal Large Language Model

MLLM unlearning aims to remove specific information or behaviors—such as private identities, copyrighted content, or unsafe associations—from models that combine language with images, audio, or other inputs. Methods include fine-tuning and adapter edits, activation steering, representation projection, and distillation, often designed to target a concept while protecting related knowledge. Benchmarks test forgetting alongside retained utility, using varied prompts, cross-modal queries, held-out examples, and adversarial recovery attempts. A central challenge is distinguishing genuine removal from refusal or suppression, while avoiding collateral damage and cumulative degradation during repeated unlearning.

Latest papers

37 papers5 papers in the last four weeks, against 1 the four weeks before. 0.0% of all new papers.

Before Forgetting, Learn to Remember: Revisiting Foundational Learning Failures in LVLM Unlearning Benchmarks
May 5, 2026JuneHyoung Kwon, MiHyeon Kim, Eunju Lee +3VLM Evaluation VLM Unlearning
Department of Artificial Intelligence, Chung-Ang University · KT Corporation · Graduate School of Advanced Imaging Sciences, Multimedia and Film, Chung-Ang University
Erase Persona, Forget Lore: Benchmarking Multimodal Copyright Unlearning in Large Vision Language Models
May 5, 2026JuneHyoung Kwon, JungMin Yun, YoungBin KimVLM Evaluation VLM Unlearning
Department of Artificial Intelligence, Chung-Ang University · Graduate School of Advanced Imaging Sciences, Multimedia and Film, Chung-Ang University
Metric Unreliability in Multimodal Machine Unlearning: A Systematic Analysis and Principled Unified Score
May 4, 2026Abdullah Ahmad Khan, Hamid Laga, Ferdous SohelVLM Evaluation VLM Unlearning
Murdoch University
Suppression Is Not Forgetting: Residual Recoverability in Visual Concept Unlearning for VLMs
Apr 3, 2026Zhangyun Tan, Zeliang Zhang, Jiani Liu +4VLM Evaluation VLM Unlearning
University of Rochester
SALMUBench: A Benchmark for Sensitive Association-Level Multimodal Unlearning
Mar 27, 2026Cai Selvas-Sala, Lei Kang, Lluis GomezVLM Unlearning Machine Unlearning Evaluation
Computer Vision Center · Universitat Polit`ecnica de Catalunya · Universitat Aut`onoma de Barcelona
Relationship-Aware Safety Unlearning for Multimodal LLMs
Mar 15, 2026Vishnu Narayanan Anilkumar, Abhijith Sreesylesh Babu, Trieu Hai Vo +2MLLM Unlearning LLM Safety
Florida International University
Towards Benign Memory Forgetting for Selective Multimodal Large Language Model Unlearning
Nov 25, 2025Zhen Zeng, Leijiang Gu, Zhangling Duan +4VLM Unlearning Multimodal Large Language Models
Hefei University of Technology, China · Institute of Artificial Intelligence, Hefei Comprehensive National Science Center, China · University of Amsterdam, the Netherlands

May 5, 2026J/K move · Enter open · S save

Before Forgetting, Learn to Remember: Revisiting Foundational Learning Failures in LVLM Unlearning Benchmarks

JuneHyoung Kwon, MiHyeon Kim, Eunju Lee, JungMin Yun+2

Department of Artificial Intelligence, Chung-Ang University · KT Corporation · Graduate School of Advanced Imaging Sciences, Multimedia and Film, Chung-Ang University

VLM Evaluation VLM Unlearning Vision-Language Models Memorization in Language Models Large Vision-Language Models Memorization in Generative Models+4

Open paper AskPDF arXiv

While Large Vision-Language Models (LVLMs) offer powerful capabilities, they pose privacy risks by unintentionally memorizing sensitive personal information. Current unlearning benchmarks attempt to mitigate this using fictitious identities but overlook a critical stage 1 failure: models fail to effectively memorize target information initially, rendering subsequent unlearning evaluations unreliable. Diagnosing under-memorization and the multi-hop curse as root causes, we introduce ReMem, a Reliable Multi-hop and Multi-image Memorization Benchmark. ReMem ensures robust foundational learning through principled data scaling, reasoning-aware QA pairs, and diverse visual contexts. Additionally, we propose a novel Exposure metric to quantify the depth of information erasure from the model's internal probability distribution. Extensive experiments demonstrate that ReMem provides a rigorous and trustworthy framework for diagnosing both learning and unlearning behaviors in LVLMs.