Who Transfers Safety? Identifying and Targeting Cross-Lingual Shared Safety Neurons
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866915767363043328 |
|---|---|
| author | Zhang, Xianhui Xie, Chengyu Zhu, Linxia Yang, Yonghui Zhao, Weixiang Cheng, Zifeng Wang, Cong Shen, Fei Chua, Tat-Seng |
| author_facet | Zhang, Xianhui Xie, Chengyu Zhu, Linxia Yang, Yonghui Zhao, Weixiang Cheng, Zifeng Wang, Cong Shen, Fei Chua, Tat-Seng |
| contents | Multilingual safety remains significantly imbalanced, leaving non-high-resource (NHR) languages vulnerable compared to robust high-resource (HR) ones. Moreover, the neural mechanisms driving safety alignment remain unclear despite observed cross-lingual representation transfer.
In this paper, we find that LLMs contain a set of cross-lingual shared safety neurons (SS-Neurons), a remarkably small yet critical neuronal subset that jointly regulates safety behavior across languages.
We first identify monolingual safety neurons (MS-Neurons) and validate their causal role in safety refusal behavior through targeted activation and suppression.
Our cross-lingual analyses then identify SS-Neurons as the subset of MS-Neurons shared between HR and NHR languages, serving as a bridge to transfer safety capabilities from HR to NHR domains.
We observe that suppressing these neurons causes concurrent safety drops across NHR languages, whereas reinforcing them improves cross-lingual defensive consistency.
Building on these insights, we propose a simple neuron-oriented training strategy that targets SS-Neurons based on language resource distribution and model architecture. Experiments demonstrate that fine-tuning this tiny neuronal subset outperforms state-of-the-art methods, significantly enhancing NHR safety while maintaining the model's general capabilities.
The code and dataset will be available athttps://github.com/1518630367/SS-Neuron-Expansion. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_01283 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Who Transfers Safety? Identifying and Targeting Cross-Lingual Shared Safety Neurons Zhang, Xianhui Xie, Chengyu Zhu, Linxia Yang, Yonghui Zhao, Weixiang Cheng, Zifeng Wang, Cong Shen, Fei Chua, Tat-Seng Computer Vision and Pattern Recognition Multilingual safety remains significantly imbalanced, leaving non-high-resource (NHR) languages vulnerable compared to robust high-resource (HR) ones. Moreover, the neural mechanisms driving safety alignment remain unclear despite observed cross-lingual representation transfer. In this paper, we find that LLMs contain a set of cross-lingual shared safety neurons (SS-Neurons), a remarkably small yet critical neuronal subset that jointly regulates safety behavior across languages. We first identify monolingual safety neurons (MS-Neurons) and validate their causal role in safety refusal behavior through targeted activation and suppression. Our cross-lingual analyses then identify SS-Neurons as the subset of MS-Neurons shared between HR and NHR languages, serving as a bridge to transfer safety capabilities from HR to NHR domains. We observe that suppressing these neurons causes concurrent safety drops across NHR languages, whereas reinforcing them improves cross-lingual defensive consistency. Building on these insights, we propose a simple neuron-oriented training strategy that targets SS-Neurons based on language resource distribution and model architecture. Experiments demonstrate that fine-tuning this tiny neuronal subset outperforms state-of-the-art methods, significantly enhancing NHR safety while maintaining the model's general capabilities. The code and dataset will be available athttps://github.com/1518630367/SS-Neuron-Expansion. |
| title | Who Transfers Safety? Identifying and Targeting Cross-Lingual Shared Safety Neurons |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2602.01283 |