English

Stroke Lesions as a Rosetta Stone for Language Model Interpretability

Machine Learning 2026-02-05 v1 Computation and Language

Abstract

Large language models (LLMs) have achieved remarkable capabilities, yet methods to verify which model components are truly necessary for language function remain limited. Current interpretability approaches rely on internal metrics and lack external validation. Here we present the Brain-LLM Unified Model (BLUM), a framework that leverages lesion-symptom mapping, the gold standard for establishing causal brain-behavior relationships for over a century, as an external reference structure for evaluating LLM perturbation effects. Using data from individuals with chronic post-stroke aphasia (N = 410), we trained symptom-to-lesion models that predict brain damage location from behavioral error profiles, applied systematic perturbations to transformer layers, administered identical clinical assessments to perturbed LLMs and human patients, and projected LLM error profiles into human lesion space. LLM error profiles were sufficiently similar to human error profiles that predicted lesions corresponded to actual lesions in error-matched humans above chance in 67% of picture naming conditions (p < 10^{-23}) and 68.3% of sentence completion conditions (p < 10^{-61}), with semantic-dominant errors mapping onto ventral-stream lesion patterns and phonemic-dominant errors onto dorsal-stream patterns. These findings open a new methodological avenue for LLM interpretability in which clinical neuroscience provides external validation, establishing human lesion-symptom mapping as a reference framework for evaluating artificial language systems and motivating direct investigation of whether behavioral alignment reflects shared computational principles.

Keywords

Cite

@article{arxiv.2602.04074,
  title  = {Stroke Lesions as a Rosetta Stone for Language Model Interpretability},
  author = {Julius Fridriksson and Roger D. Newman-Norlund and Saeed Ahmadi and Regan Willis and Nadra Salman and Kalil Warren and Xiang Guan and Yong Yang and Srihari Nelakuditi and Rutvik Desai and Leonardo Bonilha and Jeff Charney and Chris Rorden},
  journal= {arXiv preprint arXiv:2602.04074},
  year   = {2026}
}

Comments

45 pages, 17 figures

R2 v1 2026-07-01T09:35:09.855Z