Stroke Lesions as a Rosetta Stone for Language Model Interpretability
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910011053047808 |
|---|---|
| author | Fridriksson, Julius Newman-Norlund, Roger D. Ahmadi, Saeed Willis, Regan Salman, Nadra Warren, Kalil Guan, Xiang Yang, Yong Nelakuditi, Srihari Desai, Rutvik Bonilha, Leonardo Charney, Jeff Rorden, Chris |
| author_facet | Fridriksson, Julius Newman-Norlund, Roger D. Ahmadi, Saeed Willis, Regan Salman, Nadra Warren, Kalil Guan, Xiang Yang, Yong Nelakuditi, Srihari Desai, Rutvik Bonilha, Leonardo Charney, Jeff Rorden, Chris |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_04074 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Stroke Lesions as a Rosetta Stone for Language Model Interpretability Fridriksson, Julius Newman-Norlund, Roger D. Ahmadi, Saeed Willis, Regan Salman, Nadra Warren, Kalil Guan, Xiang Yang, Yong Nelakuditi, Srihari Desai, Rutvik Bonilha, Leonardo Charney, Jeff Rorden, Chris Machine Learning Computation and Language I.2.6; I.2.7; I.6.4; J.3 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. |
| title | Stroke Lesions as a Rosetta Stone for Language Model Interpretability |
| topic | Machine Learning Computation and Language I.2.6; I.2.7; I.6.4; J.3 |
| url | https://arxiv.org/abs/2602.04074 |