Computational Lesions in Multilingual Language Models Separate Shared and Language-specific Brain Alignment

2026-04-12

AI Summary

This study investigates whether the human brain employs shared and language-specific neural mechanisms in multilingual processing and establishes an interpretable alignment between multilingual large language models and brain activity. By applying targeted “computational lesions”—selectively zeroing out either shared or language-specific model parameters—and evaluating changes in the models’ ability to predict fMRI responses from 112 native speakers listening to English, Chinese, and French narratives, the research causally disentangles shared core components from language-specialized modules within six multilingual models. Lesioning the shared core reduced whole-brain encoding performance by 60.32%, whereas impairing language-specific components selectively degraded prediction accuracy for the corresponding native language while preserving cross-lingual representational separation. These findings reveal a “shared backbone plus language specialization” architecture underlying multilingual representation, offering a novel paradigm for cognitive modeling.

Technology Category

• Natural Language Processing: Language Grounding & Multi-modal NLP
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: (Computational) Cognitive Architectures

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

How the brain supports language across different languages is a basic question in neuroscience and a useful test for multilingual artificial intelligence. Neuroimaging has identified language-responsive brain regions across languages, but it cannot by itself show whether the underlying processing is shared or language-specific. Here we use six multilingual large language models (LLMs) as controllable systems and create targeted ``computational lesions'' by zeroing small parameter sets that are important across languages or especially important for one language. We then compare intact and lesioned models in predicting functional magnetic resonance imaging (fMRI) responses during 100 minutes of naturalistic story listening in native English, Chinese and French (112 participants). Lesioning a compact shared core reduces whole-brain encoding correlation by 60.32% relative to intact models, whereas language-specific lesions preserve cross-language separation in embedding space but selectively weaken brain predictivity for the matched native language. These results support a shared backbone with embedded specializations and provide a causal framework for studying multilingual brain-model alignment.

Problem

• multilingual
• brain alignment
• language-specific
• shared processing
• neuroscience

Innovation

• computational lesions
• multilingual language models
• brain-model alignment
• fMRI prediction
• shared vs. language-specific processing