← Search

Linxia Zhu

1 accepted papers

2026

Who Transfers Safety? Identifying and Targeting Cross-Lingual Shared Safety Neurons

ICML 2026poster

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…

Cited by 0SourceScholar