aiqbench analysis
This study critically exposes the fragility of similarity scores used to validate claims about brain-model alignment or cross-lingual transfer, demonstrating that these metrics often reflect tool limitations rather than genuine structural correspondence. By showing that probes fail in distant languages and that baseline scores persist even when brain-target correspondence is destroyed, the research urges a rigorous re-evaluation of how we interpret evidence for shared cognitive structures in AI.
The Score Is Not the Structure: Brain Alignment and Cross-Lingual Transfer
arXiv:2610.03827v1 Announce Type: cross Abstract: Researchers often support the claim that a model shares structure with the brain, or across languages, by reporting a similarity score. We ask what such a score reads when the shared structure is absent, or when the tool that measures it does not work. We check two settings, and in both the score is not what it appears. First, a probe trained to tell grammatical from ungrammatical sentences in one language transfers worse to more distant languages, the usual evidence for shared structure. But the probe itself gets worse along the same axis: in four of seventeen languages it performs at chance, so 64 of 272 language pairs are scored with a tool that does not work. Dropping those languages halves the strength of the relationship, but they are also the most distant, and this design cannot separate the two effects. Counting matters too: the same data give p = 0.0006 when the 272 pairs are treated as independent and p = 0.155 when the seventeen languages are, which is the correct unit. Second, training a language model to match human brain responses raises its similarity score from 0.10 to 0.34, against a ceiling of 0.54. A model trained on a target whose correspondence to the brain was destroyed still scores 0.31, so only 0.028 to 0.068 of the rise is specific to the brain. With no model at all, a destroyed target already sits at 0.204 from the real one, a floor that tracks the target's rank divided by the number of sentences. Finally, steering a language along its own direction works (+6.2 over a random direction in sixteen of seventeen languages) yet shows no effect that varies with language distance. Before asking whether a correspondence helps, ask how much of the score would survive without it.
Abstract:Researchers often support the claim that a model shares structure with the brain, or across languages, by reporting a similarity score. We ask what such a score reads when the shared structure is absent, or when the tool that measures it does not work. We check two settings, and in both the score is not what it appears. First, a probe trained to tell grammatical from ungrammatical sentences in one language transfers worse to more distant languages, the usual evidence for shared structure. But the probe itself gets worse along the same axis: in four of seventeen languages it performs at chance, so 64 of 272 language pairs are scored with a tool that does not work. Dropping those languages halves the strength of the relationship, but they are also the most distant, and this design cannot separate the two effects. Counting matters too: the same data give p = 0.0006 when the 272 pairs are treated as independent and p = 0.155 when the seventeen languages are, which is the correct unit. Second, training a language model to match human brain responses raises its similarity score from 0.10 to 0.34, against a ceiling of 0.54. A model trained on a target whose correspondence to the brain was destroyed still scores 0.31, so only 0.028 to 0.068 of the rise is specific to the brain. With no model at all, a destroyed target already sits at 0.204 from the real one, a floor that tracks the target's rank divided by the number of sentences. Finally, steering a language along its own direction works (+6.2 over a random direction in sixteen of seventeen languages) yet shows no effect that varies with language distance. Before asking whether a correspondence helps, ask how much of the score would survive without it.
12 pages, 2 figures, 1 table. Accepted as a poster at the NeurIPS 2026 - Linguistic Principles for Foundation Models (LP4FM). Code: this https URL
Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
arXiv:2610.03827 [cs.CL]
(or arXiv:2610.03827v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2610.03827 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 428 words
Related coverage
Cross-lingual brain-language model alignment is robust but challenges hierarchical and computational accounts
arXiv · Sep 11, 2026
Cross-Modal Ultrasound-MRI Learning for Fetal Brain Ventricular Volumetry and Abnormality Screening
arXiv · Aug 19, 2026
Cross-Task Generalization in Handwriting-Based Alzheimer's Screening via Vision Language Adaptation
arXiv · Sep 25, 2026
Concrete Evaluations to Investigate the OpenAI Model That Hacked Hugging Face
AI Alignment Forum · Aug 3, 2026