Cross-lingual robustness of LLM-brain alignment and its computational roots

2026-05-20

AI Summary

This study investigates whether the alignment between large language models (LLMs) and the human brain during multilingual comprehension exhibits cross-linguistic robustness, extends to subcortical regions, and what computational mechanisms underlie this correspondence. Using whole-brain fMRI encoding models across naturalistic story-listening tasks in Chinese, English, and French, the authors analyze the capacity of Transformer-based models to predict neural responses. They report, for the first time, that LLMs consistently predict activity across a broad network of cortical and subcortical areas, with highly congruent alignment patterns across languages. This cross-linguistic neural alignment cannot be accounted for by prediction uncertainty or representational geometry, challenging prevailing views that emphasize contextual embedding superiority or surprisal-driven processing, and instead supports lexical–semantic correspondence as the core mechanism.

Abstract

Large language models (LLMs) reliably predict neural activity during language comprehension and transformer depth has been interpreted as mirroring hierarchical cortical organization. However, it remains unclear whether such alignment extends to subcortical regions, overlaps spatially across languages, and what the computational roots of such alignment are. Here, we used a multilingual, whole-brain encoding framework to examine brain-LLM alignment across three typologically distinct languages: Mandarin, English, and French during naturalistic story listening. Our results show that across languages, transformer-based models predicted activity in a distributed landscape spanning widely distributed cortical functional networks like limbic, ventral attention, default mode network, and subcortical structures. Spatial alignment patterns showed substantial cross-linguistic overlap and remained largely stable across model layers, with limited layer progression consistent with functional cortical hierarchies. Contrary to previous evidence, contextual embeddings did not outperform static embeddings. To test candidate computational explanations, we examined whether layer-wise brain scores reflect surprisal and intrinsic dimensionality, and thereby predictive processing and information compression. Neither of these two computational metrics mirrored neural alignment profiles. Our findings suggest that brain-LLM alignment is spatially robust and cross-linguistically stable but not explainable from predictive uncertainty or representational geometry. Rather than directly reflecting shared hierarchical computation, neural predictivity may primarily arise from distributed lexical-semantic correspondences that generalize across languages.

Problem

• cross-lingual robustness
• brain-LLM alignment
• subcortical regions
• computational roots
• neural predictivity

Innovation

• cross-lingual alignment
• LLM-brain correspondence
• whole-brain encoding
• lexical-semantic mapping
• transformer hierarchy