AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

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

Xianhui Zhang, Chengyu Xie, Linxia Zhu, Yonghui Yang, Weixiang Zhao, Zifeng Cheng, Cong Wang, Fei Shen, Tat-Seng Chua

ICML 2026 · Large language models › Language, safety and interpretability · approach: Multilingual LLM data and culture · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add English-MS-neuron-only SS-Neuron expansion tuningGemma2-9B-it · MultiJailavg ASR (%): 10.93 -> 1.59Improved
add SS-Neuron expansion tuningLlama3.1-8B-it · AdvBench-xASR (%): 30.22 -> 0.22Improved
add NHR translated parallel data to HR-only training (ablation)SS-Neuron expansion · MultiJail (Gemma2-9B-it)ASR (%): 23.49 -> 1.59Improved
add NHR translated parallel data to HR-only training (ablation)SS-Neuron expansion · AdvBench-x (Qwen3-8B)ASR (%): 12.20 -> 1.13Improved
select English MS-Neurons instead of random neurons (ablation)SS-Neuron expansion · MultiJail (Gemma2-9B-it)ASR (%): 5.12 -> 1.59Improved
select English MS-Neurons instead of random neurons (ablation)SS-Neuron expansion · AdvBench-x (Llama3.1-8B-it)ASR (%): 5.19 -> 0.22Improved
mask MS-Neurons (causal check)Qwen3-8B · AdvBench-xASR (%): 15.60 -> 41.46Hurt
mask randomly chosen neurons (control)Qwen3-8B · AdvBench-xASR (%): 15.60 -> 15.38No clear effect
mask SS-Neurons on NHR languagesQwen3-8B · MultiJailASR (%): 22.12 -> 33.97Hurt
use sparse neuron tuning instead of LoRAGemma2-9B-it · MultiJailASR (%): 33.06 -> 1.59Improved
use neuron tuning instead of full trainingGemma2-9B-it · MMLUaccuracy: 12.88 -> 79.36Improved
use neuron tuning instead of LoRAQwen3-8B · MMLUaccuracy: 78.40 -> 84.08Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy