Language Arithmetics: Towards Systematic Language Neuron Identification and Manipulation

ACL ARR 2025 July Submission623 Authors

Abstract:

Large language models (LLMs) exhibit strong multilingual abilities, yet the neural mechanisms behind language-specific processing remain unclear. We analyze language-specific neurons in Llama-3.1-8B, Mistral-Nemo-12B, and Aya-Expanse-8B & 32B across 21 typologically diverse languages, identifying neurons that control language behavior. Using the Language Activation Probability Entropy (LAPE) method, we show that these neurons cluster in deeper layers, with non-Latin scripts showing greater specialization. Related languages share overlapping neurons, reflecting internal representations of linguistic proximity.

Through language arithmetics, i.e. systematic activation addition and multiplication, we steer models to deactivate unwanted languages and activate desired ones, outperforming simpler replacement approaches. These interventions effectively guide behavior across five multilingual tasks: language forcing, translation, QA, comprehension, and NLI. Manipulation is more successful for high-resource languages, while typological similarity improves effectiveness. We also demonstrate that cross-lingual neuron steering enhances downstream performance and reveal internal "fallback" mechanisms for language selection when neurons are progressively deactivated.

Paper Type: Long

Research Area: Multilingualism and Cross-Lingual NLP

Research Area Keywords: multilingual representations, language change, multilingualism

Contribution Types: Model analysis & interpretability

Languages Studied: Tibetan, Maltese, Italian, Spanish, German, Japanese, Arabic, Chinese, Afrikaans, Dutch, French, Portuguese, Russian, Korean, Hindi, Turkish, Polish, Swedish, Danish, Norwegian, English

Reassignment Request Area Chair: This is not a resubmission

Reassignment Request Reviewers: This is not a resubmission

A1 Limitations Section: This paper has a limitations section.

A2 Potential Risks: N/A

B Use Or Create Scientific Artifacts: Yes

B1 Cite Creators Of Artifacts: Yes

B1 Elaboration: Section 3 and 4

B2 Discuss The License For Artifacts: N/A

B3 Artifact Use Consistent With Intended Use: Yes

B3 Elaboration: Section 3 and 4

B4 Data Contains Personally Identifying Info Or Offensive Content: N/A

B5 Documentation Of Artifacts: Yes

B5 Elaboration: Appendix B

B6 Statistics For Data: Yes

B6 Elaboration: Section 3 and 4

C Computational Experiments: Yes

C1 Model Size And Budget: Yes

C1 Elaboration: Section 3 - we report model sizes; we only perform inference experiments and do not perform any training experiments

C2 Experimental Setup And Hyperparameters: Yes

C2 Elaboration: Appendix A

C3 Descriptive Statistics: N/A

C4 Parameters For Packages: Yes

C4 Elaboration: Appendix A

D Human Subjects Including Annotators: No

D1 Instructions Given To Participants: N/A

D2 Recruitment And Payment: N/A

D3 Data Consent: N/A

D4 Ethics Review Board Approval: N/A

D5 Characteristics Of Annotators: N/A

E Ai Assistants In Research Or Writing: No

E1 Information About Use Of Ai Assistants: N/A

Author Submission Checklist: yes

Submission Number: 623