Audit finds TRIBE v2 brain model captures motion selectivity but not retinotopy or tonotopy
An in-silico audit of the multimodal fMRI prediction model reports a sharp split: motion selectivity in area MT+ is well reproduced, while visual-field and frequency maps are absent.
In plain words
TRIBE v2 is a model that predicts where activity appears on the brain's surface when a person watches video with sound and text. A researcher tested whether the model's predictions preserve the brain's orderly maps of sight, sound and motion. Motion maps held up, but the sight and sound maps did not.
What was announced
The audit's author fitted standard brain-mapping stimuli to TRIBE v2's predictions and compared the resulting maps with human atlases using spin tests. The results show no contralateral hemifield response and no polar-angle map, robustly across stimulus controls, and no tonotopic gradient in primary auditory cortex; best frequencies spanned an interquartile range of only 0.14 octaves across core and belt, and the null persisted with loudness-equalized tones. An eccentricity-like gradient appeared only when the ring geometry changed, which the author reads as consistent with global image statistics rather than a positional code.
Motion selectivity was the exception: 22% of the 1% most motion-selective vertices fell in MT+, an area covering 0.42% of cortex. The author attributes the retinotopy failure to architecture, noting that TRIBE v2's video features are averaged over space before the response head, which should discard visual-field position. The conclusion is that in-silico experiments depending on feature identity appear viable with the model, while those depending on visual-field or frequency position are not supported by the current model.
What it does not show
- The source is an arXiv abstract only; the full paper's methods and peer-review status are not covered, and the audit is a single-author in-silico analysis, not independent confirmation of TRIBE v2's original validation.
For security and IT teams
- Scope
- The findings concern TRIBE v2's predicted fMRI responses as audited in silico, not new human neuroimaging measurements.
Sources
Source-support review completed 17 Nov 2026. This story was drafted with AI assistance and checked against retained source passages. Automated checking is not independent verification.
Coverage: arXiv: abstract.
Updates and corrections
No corrections so far.
First published 17 Nov 2026
New to the subject? AI from the Ground Up starts from the beginning.