AI Paper Atlas

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

Focusing on Language: Revealing and Exploiting Language Attention Heads in Multilingual Large Language Models

Xin Liu, Qiyang Song, Qihang Zhou, Haichao Du, Shaowen Xu, Wenbo Jiang, Weijuan Zhang, Xiaoqi Jia

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Task vectors and steering · 2 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
train language-specific head mask (14-20 params) on top 2% LAHIS headsMistral-7B · XQuAD (avg)accuracy: 22.53 -> 29.03Improved
train language-specific head maskAya-23-8B · XQuAD (avg)accuracy: 55.28 -> 61.10Improved
train language-specific head maskLlama-3.2-3B · XQuAD (avg)accuracy: 32.98 -> 36.78Improved
use LAHIS-selected heads instead of random heads for the mask (ablation)Aya-23-8B · XQuAD (avg)accuracy: 56.15 -> 61.10Improved
use random head mask instead of vanilla (ablation)Aya-23-8B · XQuAD (avg)accuracy: 55.28 -> 56.15No clear effect
deactivate language-general heads (ablation)Aya-23-8B · XL-Sum (avg)BERTScore F1: 78.5 -> 66.2Hurt
deactivate language-general heads (ablation)Llama-3.2-3B · XL-Sum (avg)BERTScore F1: 66.3 -> 49.6Hurt
deactivate the same number of random heads (ablation)Llama-3.2-3B · XL-Sum (avg)BERTScore F1: 66.3 -> 66.8No clear effect
deactivate language-specific heads (ablation)Aya-23-8B · Wikipedia PPL (zh)perplexity: 8.66 -> 14.1Hurt
deactivate language-specific heads (ablation)Mistral-7B · Wikipedia PPL (th)perplexity: 4.35 -> 7.70Hurt
deactivate random heads of same count (ablation)Aya-23-8B · Wikipedia PPL (zh)perplexity: 8.66 -> 8.94No clear effect
deactivate English heads to cut off-target English outputMistral-7B · XL-Sum Vietnameselanguage accuracy: 0.35 -> 1.00Improved

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