AI Paper Atlas

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

How Does Alignment Enhance LLMs’ Multilingual Capabilities? A Language Neurons Perspective

Shimao Zhang, Zhejian Lai, Xiang Liu, Shuaijie She, Xiao Liu, Yeyun Gong, Shujian Huang, Jiajun Chen

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Multilingual LLM data and culture · 5 citations (OpenAlex)

Read the paper

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
align with zh/de => en (MAPO DPO)MistralMathOctopus-7B · MGSM (average)accuracy: 57.8 -> 63.6Improved
align with sw/th => en (MAPO DPO)MistralMathOctopus-7B · MGSM (average)accuracy: 57.8 -> 64.7Improved
align with all 9 languages => enMistralMathOctopus-7B · MGSM (average)accuracy: 57.8 -> 62.0Improved
align with all 9 languages instead of sw/thMistralMathOctopus-7B · MGSM (average)accuracy: 64.7 -> 62.0Hurt
align with sw/th => en (MAPO DPO)MistralMathOctopus-7B · MSVAMP (average)accuracy: 64.4 -> 69.3Improved
align with zh/de => en (unseen languages)MistralMathOctopus-7B · MGSM (sw)accuracy: 50.4 -> 59.2Improved
align with zh => enMistralMathOctopus-7B · MGSM (en)accuracy: 75.6 -> 75.2No clear effect
align with zh/de => en: language-specific neurons change (trained langs)MistralMathOctopus-7B · MGSMlanguage-specific neuron count: -37Hurt
align with zh/de => en: language-related neurons change (trained langs)MistralMathOctopus-7B · MGSMlanguage-related neuron count: +232Improved
align with zh/de => en: language-related neurons (unseen langs)MistralMathOctopus-7B · MGSMlanguage-related neuron count: +205Improved
deactivate language-specific plus language-related neurons instead of only specific (ablation)MistralMathOctopus-7B · MGSMaccuracy/PPL change: more pronounced effectImproved
deactivate English language neurons (ablation)MistralMathOctopus-7B · MGSM (en)accuracyNo clear effect

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