# Relationship-Aware Safety Unlearning for Multimodal LLMs

arXiv: https://arxiv.org/abs/2603.14185 · HTML: https://mlsanity.com/paper/2603.14185/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in MLLM jailbreaks and adversarial robustness, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 24 of its 30 references that are in the map.
- First posted on arXiv 15 Mar 2026.

## Placed in

- Multimodal LLMs / vision-language models / [MLLM jailbreaks and adversarial robustness](https://mlsanity.com/mllm/mllm-jailbreaks-and-adversarial-robustness/)
- LLM safety and security / [Vision-language and text-to-image safety](https://mlsanity.com/safety/vision-language-and-text-to-image-safety/)

## Builds on (most cited first, 24 of 24)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Erasing Concepts from Diffusion Models](https://mlsanity.com/paper/2303.07345/) (2023)
- [Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning](https://mlsanity.com/paper/2404.05868/) (2024)
- [Knowledge Unlearning for Mitigating Privacy Risks in Language Models](https://mlsanity.com/paper/2210.01504/) (2022)
- [A Survey of Machine Unlearning](https://mlsanity.com/paper/2209.02299/) (2022)
- [Large Language Model Unlearning](https://mlsanity.com/paper/2310.10683/) (2023)
- [MUSE: Machine Unlearning Six-Way Evaluation for Language Models](https://mlsanity.com/paper/2407.06460/) (2024)
- [Simplicity Prevails: Rethinking Negative Preference Optimization for LLM Unlearning](https://mlsanity.com/paper/2410.07163/) (2024)
- [One-Dimensional Adapter to Rule Them All: Concepts, Diffusion Models and Erasing Applications](https://mlsanity.com/paper/2312.16145/) (2023)
- [Machine Unlearning via Algorithmic Stability](https://mlsanity.com/paper/2102.13179/) (2021)
- [SOUL: Unlocking the Power of Second-Order Optimization for LLM Unlearning](https://mlsanity.com/paper/2404.18239/) (2024)
- [Direct Unlearning Optimization for Robust and Safe Text-to-Image Models](https://mlsanity.com/paper/2407.21035/) (2024)
- [Dissecting Language Models: Machine Unlearning via Selective Pruning](https://mlsanity.com/paper/2403.01267/) (2024)
- [Digital Forgetting in Large Language Models: A Survey of Unlearning Methods](https://mlsanity.com/paper/2404.02062/) (2024)
- [A Review on Machine Unlearning](https://mlsanity.com/paper/2411.11315/) (2024)
- [Separable Multi-Concept Erasure from Diffusion Models](https://mlsanity.com/paper/2402.05947/) (2024)
- [Modality-Aware Neuron Pruning for Unlearning in Multimodal Large Language Models](https://mlsanity.com/paper/2502.15910/) (2025)
- [Fine-Grained Erasure in Text-to-Image Diffusion-based Foundation Models](https://mlsanity.com/paper/2503.19783/) (2025)
- [MultiDelete for Multimodal Machine Unlearning](https://mlsanity.com/paper/2311.12047/) (2023)
- [CLIPErase: Efficient Unlearning of Visual-Textual Associations in CLIP](https://mlsanity.com/paper/2410.23330/) (2024)
- [Machine Unlearning in Learned Databases: An Experimental Analysis](https://mlsanity.com/paper/2311.17276/) (2023)
- [A Survey on Machine Unlearning: Techniques and New Emerged Privacy Risks](https://mlsanity.com/paper/2406.06186/) (2024)
- [SoK: Machine Unlearning for Large Language Models](https://mlsanity.com/paper/2506.09227/) (2025)
- [Erasing Concepts, Steering Generations: A Comprehensive Survey of Concept Suppression](https://mlsanity.com/paper/2505.19398/) (2025)

## Abstract (arXiv)

Generative multimodal models can exhibit safety failures that are inherently relational: two benign concepts can become unsafe when linked by a specific action or relation (e.g., child-drinking-wine). Existing unlearning and concept-erasure approaches often target isolated concepts or image-text pairs, which can cause collateral damage to benign uses of the same objects and relations. We propose relationship-aware safety unlearning: a framework that explicitly represents unsafe object-relation-object (O-R-O) tuples and applies targeted parameter-efficient edits (LoRA) to suppress unsafe tuples while preserving object marginals and safe neighboring relations. We include CLIP-based experiments and robustness evaluation under paraphrase, contextual, and out-of-distribution image attacks.