Cong Wang, Tat-Seng Chua, Fei Shen, Weixiang Zhao, Zifeng Cheng, Yonghui Yang, Xianhui Zhang, Chengyu Xie, Linxia Zhu, 1+5 +5, H Ms-Neurons, ๐๐ ๐๐๐๐๐๐๐๐๐ฌ๐ฌ๐ฌ๐ฌ๐๐๐ฌ๐ฌ, and 1 more
arXiv ยท PDF ยท Open in the Atlas
We have not lifted any functions out of this paper's repositories yet, so there is nothing we have run. The repositories linked to it are listed below.
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
Multilingual safety remains significantly imbalanced, leaving non-high-resource (NHR) languages vulnerable compared to robust highresource (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 crosslingual 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 finetuning 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 at https://github.com/1518630367/SS-Neuron-Expansion. Warning: this paper contains examples with unsafe content.
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2602.01283")
get_code_for_paper("2602.01283")
have("2602.01283")
The run record, dated, one paper per request, free:
curl https://syntology.ai/api/ran/2602.01283.json
A badge for a README (the split and the date, never a ratio):
[](https://syntology.ai/paper/2602.01283)
Connect an agent โ have() is free.