Multimodal Unlearning Across Vision, Language, Video, and Audio: Survey of Methods, Datasets, and Benchmarks
Nobin Sarwar, Shubhashis Roy Dipta, Zheyuan Liu, Vaidehi Patil
Why It Matters
What makes this one worth your time
Understanding and implementing multimodal unlearning is crucial for maintaining privacy, compliance, and ethical standards in AI systems that integrate diverse data types.
A comprehensive survey on multimodal unlearning techniques and challenges.
Summary
The paper surveys methods, datasets, and benchmarks for multimodal unlearning, focusing on removing specific knowledge from models across vision, language, audio, and video modalities while maintaining overall utility.
Key contributions
- Provides a unified, system-oriented view of multimodal unlearning.
- Offers a taxonomy for systematic comparison of unlearning methods.
- Releases a curated repository for multimodal unlearning resources.
Notable insights
- The paper introduces a taxonomy for comparing multimodal unlearning methods across different model architectures and modalities.
- It highlights the trade-offs involved in deletion strength, retention, efficiency, reversibility, and robustness.
Possible limitations
- Not stated in the abstract
Abstract
arXiv:2607.07907v1 Announce Type: cross Abstract: With the growing adoption of VLMs, DMs, LLMs, and AFMs, these multimodal foundation models can inadvertently encode sensitive, copyrighted, biased, or unsafe cross-modal associations that originate from their training data. Retraining after deletion requests or policy updates is often impractical, and targeted forgetting remains difficult because knowledge is distributed across shared representations. Multimodal unlearning addresses this challenge by enabling selective removal across modalities while retaining overall utility. This survey offers a unified, system-oriented view of multimodal unlearning across vision, language, audio, and video, grounded in recent advances, emerging applications, and open problems. Our taxonomy enables systematic comparison across model architectures and modalities, clarifying trade-offs among deletion strength, retention, efficiency, reversibility, and robustness. This survey highlights open problems and practical considerations to support future research and deployment of multimodal unlearning. We release a curated repository: https://smsnobin77.github.io/Awesome-Multimodal-Unlearning/