Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2603.14185 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914420857241600 |
|---|---|
| author | Anilkumar, Vishnu Narayanan Babu, Abhijith Sreesylesh Vo, Trieu Hai Kolla, Mohankrishna Cuneo, Alexander |
| author_facet | Anilkumar, Vishnu Narayanan Babu, Abhijith Sreesylesh Vo, Trieu Hai Kolla, Mohankrishna Cuneo, Alexander |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_14185 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Relationship-Aware Safety Unlearning for Multimodal LLMs Anilkumar, Vishnu Narayanan Babu, Abhijith Sreesylesh Vo, Trieu Hai Kolla, Mohankrishna Cuneo, Alexander Artificial Intelligence 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. |
| title | Relationship-Aware Safety Unlearning for Multimodal LLMs |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2603.14185 |