arxiv: 2604.21780 · v1 · submitted 2026-04-23 · 🧬 q-bio.NC

Recognition: unknown

Only Brains Align with Brains: Cross-Region Alignment Patterns Expose Limits of Normative Models

Larissa H\"ofling , Matthias Tangemann , Lotta Piefke , Susanne Keller , Katrin Franke , Matthias Bethge

Authors on Pith no claims yet

Pith reviewed 2026-05-08 13:09 UTC · model grok-4.3

classification 🧬 q-bio.NC

keywords brain-model alignmentalignment patternsfMRIvision modelsvisual cortexrepresentational similaritymodel evaluation

0 comments

The pith

Brain regions show stable cross-region alignment patterns that even top vision models fail to match.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces alignment patterns as the characteristic profiles of how each visual brain region relates functionally to all others. Standard benchmarks that measure how well model activations predict brain responses or match representational geometry often rank many different models as roughly equivalent. In contrast, alignment pattern analysis requires that a model aligned to one region also reproduces that region's specific pattern of relations to the others. These patterns prove consistent across different human subjects, yet top-performing models do not reproduce them. The work therefore separates the use of models as prediction tools from any claim that they capture the brain's computational organization.

Core claim

Alignment patterns are defined as the characteristic functional relationship profiles of each brain region to all others. When applied to the BOLD Moments video fMRI dataset across visual ROIs, these patterns remain highly stable across subjects. A broad range of vision models that perform well under conventional alignment measures nevertheless fail to reproduce the observed patterns, showing that standard benchmarks lack the discriminative power to establish deeper structural similarity.

What carries the argument

Alignment pattern analysis (APA), a second-order test that checks whether a model aligned to a given ROI reproduces that ROI's characteristic cross-region alignment profile.

If this is right

Conventional alignment benchmarks are insufficient to discriminate models on structural grounds.
Models can remain useful as predictive tools even when they do not match relational brain patterns.
Claims of computational similarity to human visual cortex require evidence beyond first-order response prediction.
Evaluation standards should differ depending on whether a model is intended as a tool or as a model of brain computation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Model training procedures could add explicit losses that penalize mismatches in cross-region profiles.
The same second-order test could be applied to other sensory systems or brain networks to check whether relational consistency is a general requirement.
The distinction between predictive utility and computational equivalence suggests separate validation pipelines for different scientific goals.

Load-bearing premise

Reproducing the stable cross-region relational profiles is necessary to claim that a model is computationally or algorithmically similar to the brain.

What would settle it

Identifying or constructing a vision model that reproduces the measured cross-region alignment patterns for multiple ROIs while retaining high predictive accuracy on individual regions would contradict the reported limits of current models.

Figures

Figures reproduced from arXiv: 2604.21780 by Katrin Franke, Larissa H\"ofling, Lotta Piefke, Matthias Bethge, Matthias Tangemann, Susanne Keller.

**Figure 1.** Figure 1: Alignment pattern analysis to distinguish between equivalently aligned models. Left: Standard brain-alignment benchmarks rank models according to their alignment to a brain region under some similarity transform. Comparing models’ alignment scores to a reference derived from brain-brain alignment scores aids interpreting model scores (e.g. the NeuroAI Turing Test Feather et al., 2025), but leaves open the … view at source ↗

**Figure 2.** Figure 2: Diverse models achieve comparable alignment scores on the BOLD Moments Dataset. (a) Subject-averaged alignment scores (RSA/LP) across ROIs; errorbars are standard deviation across ROIs. (b) Subject-averaged alignment scores for individual ROIs (V1, V8, MST); errorbars indicate bootstrapped 95% confidence-intervals around the mean. Effectively equivalent models (see Methods Sec. 3.4) are highlighted in bold… view at source ↗

**Figure 3.** Figure 3: Brain-brain alignment patterns are consistent across subjects and characteristic for ROIs. (a) RSA-based fMRI-derived AP for example ROIs, mean ± SEM across subjects. Shaded areas indicate where predictor and target ROI coincide. (b) Dark gray bars: APS between fMRIderived AP and structural connectivity-derived AP, horizontal lines indicate 95% percentile of the null distribution of APS. Red for ROIs wher… view at source ↗

**Figure 4.** Figure 4: Alignment patterns differentiate between models that appear effectively equivalent. (a) RSA-based fMRI-derived (solid lines, ROI-color mapping as in Fig. S3.1) and model-derived ((dash-)dotted lines, color mapping as in view at source ↗

read the original abstract

Neuroscientists and computer vision researchers use model-brain alignment benchmarks to compare artificial and biological vision systems. These benchmarks rank models according to alignment measures such as the similarity of representational geometry or the predictability of neural responses from model activations. However, recent works have identified a number of problems with these rankings, among them their lack of discriminative power and robustness, raising the conceptual question of what it means for a model to be brain-aligned. Here we introduce alignment patterns -- characteristic functional relationship profiles of each brain region to all others -- and propose that models should reproduce these patterns to qualify as brain-aligned. First, we apply a standard benchmarking pipeline to a broad spectrum of vision models of the BOLD Moments video fMRI dataset across visual regions of interest (ROIs). We find diverse models appear equivalent in their brain alignment, reflecting the lack of discriminative power of conventional alignment benchmarking pipelines. In contrast, alignment pattern analysis (APA) is a second-order structural consistency test: a model aligned to a given ROI should reproduce that ROI's characteristic cross-region alignment profile. Applying APA, we find that, while these patterns are highly stable across brains of different subjects, even top-ranked models often fail to capture them. Finally, we argue for a clearer distinction between the criteria a model must meet to serve as a tool versus as a computational model for human visual cortex. Conventional alignment measures may be sufficient for identifying neurally predictive models, but claims about computational or algorithmic similarity may require a stronger basis of evidence, including the reproducibility of relational alignment patterns.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper shows that standard model-brain alignment scores fail to discriminate models while their new alignment pattern analysis finds stable cross-region profiles that top models miss.

read the letter

The main takeaway is that conventional alignment benchmarks on the BOLD Moments dataset make many vision models look roughly equivalent, but alignment pattern analysis (APA) exposes that even the better ones do not reproduce the stable relational profiles that hold across human subjects' visual regions. APA treats each ROI's pattern of alignment to all other ROIs as a second-order signature that a model aligned to one area should recover for the others. The authors apply a standard pipeline first, then show the contrast with APA, and close by separating models that are merely predictive tools from those that could count as computational accounts of cortex. That distinction is the clearest part of the argument and is stated without overclaiming. The empirical contrast on one video fMRI dataset is straightforward and the stability claim across subjects is presented as the key positive result for the brain side. The paper does not invent new alignment metrics from scratch; it reframes existing similarity measures into a consistency test across regions. That reframing is the actual novelty and it is cleanly motivated by the known weakness of first-order scores. The soft spot is the absence of concrete numbers in the abstract on failure rates, dataset scale, or statistical controls for the APA results. Without those, it is difficult to judge how decisive the model failures are or how sensitive the patterns are to preprocessing choices. The conceptual section on tools versus models is careful and does not over-reach, but it remains an argument rather than a new empirical demonstration. This paper is aimed at researchers who run or evaluate brain-alignment benchmarks in computational neuroscience and AI. Anyone who has noticed that many models cluster together on standard scores will find the critique useful even if they ultimately disagree with APA as the fix. It deserves peer review because the core observation is testable and the distinction it draws is worth sharpening with data and methods details.

Referee Report

2 major / 2 minor

Summary. The paper claims that conventional model-brain alignment benchmarks lack discriminative power, as diverse vision models yield equivalent scores on the BOLD Moments video fMRI dataset across visual ROIs. It introduces alignment patterns—characteristic cross-region relational profiles—and shows via alignment pattern analysis (APA) that these patterns are stable across human subjects yet are not reproduced by even top-ranked models. The authors argue that standard alignment measures suffice for identifying predictive tools but that computational models of visual cortex require reproduction of these relational structures.

Significance. If the APA results hold, the work supplies a concrete, second-order consistency test that could sharpen the distinction between neurally predictive models and those claiming computational or algorithmic similarity to cortex. It directly addresses documented weaknesses in current benchmarking pipelines and supplies an empirical basis for requiring models to recover inter-region functional relationships rather than isolated ROI predictions.

major comments (2)

[Abstract and §3] Abstract and §3 (methods): the central claim that 'even top-ranked models often fail to capture' the patterns is load-bearing yet the abstract supplies no quantitative failure rates, effect sizes, or statistical controls (e.g., permutation baselines, multiple-comparison correction, or subject-level variance). Without these numbers it is impossible to judge whether the reported failures exceed what would be expected from noise or from the limited discriminative power already acknowledged in the conventional benchmarks.
[§4 and discussion] §4 (results) and discussion: the stability of alignment patterns across subjects is asserted as 'highly stable,' but the manuscript must report the precise inter-subject correlation coefficients, the number of subjects, and the cross-validation scheme used to establish this stability. If these values are modest or if the patterns are derived from the same BOLD Moments data used to rank the models, the contrast between brain stability and model failure risks circularity.

minor comments (2)

[Abstract] Notation: the term 'alignment patterns' is introduced without an explicit equation or pseudocode definition in the abstract; a compact formal definition (e.g., a vector of pairwise alignment scores) would improve reproducibility.
[Figures] Figure clarity: any figures showing cross-region matrices should include subject-averaged and model-averaged versions side-by-side with the same color scale and a clear legend for the alignment metric used.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which have prompted us to strengthen the quantitative presentation and clarify potential concerns about circularity in our analysis. We address each major comment below and have revised the manuscript accordingly.

read point-by-point responses

Referee: [Abstract and §3] Abstract and §3 (methods): the central claim that 'even top-ranked models often fail to capture' the patterns is load-bearing yet the abstract supplies no quantitative failure rates, effect sizes, or statistical controls (e.g., permutation baselines, multiple-comparison correction, or subject-level variance). Without these numbers it is impossible to judge whether the reported failures exceed what would be expected from noise or from the limited discriminative power already acknowledged in the conventional benchmarks.

Authors: We agree that the abstract and methods would benefit from explicit quantitative support for the failure claim. In the revised manuscript we have updated the abstract to report specific failure rates among top-ranked models (expressed as percentages exceeding permutation baselines), along with effect sizes and references to the statistical controls. Section 3 has been expanded to detail the permutation testing procedure, FDR-corrected multiple-comparison adjustments, and subject-level variance estimates. These additions confirm that the observed model failures are statistically distinguishable from noise while preserving the original findings. revision: yes
Referee: [§4 and discussion] §4 (results) and discussion: the stability of alignment patterns across subjects is asserted as 'highly stable,' but the manuscript must report the precise inter-subject correlation coefficients, the number of subjects, and the cross-validation scheme used to establish this stability. If these values are modest or if the patterns are derived from the same BOLD Moments data used to rank the models, the contrast between brain stability and model failure risks circularity.

Authors: We have revised §4 to explicitly state the inter-subject correlation coefficients, the exact number of subjects, and the cross-validation scheme (subject-wise hold-out) used to quantify stability. In the discussion we have added a dedicated paragraph clarifying that model rankings rely on standard first-order alignment scores, whereas the stability metric is computed via independent subject splits that do not overlap with the model-evaluation folds. This separation prevents circularity and ensures the reported contrast between brain consistency and model performance is valid. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper computes alignment patterns directly from BOLD Moments brain data across subjects and ROIs, then tests whether models reproduce those independently measured profiles. No equations or definitions reduce the APA test result to a model-fitted quantity or self-referential input. Conventional benchmarks are applied as standard external measures. The conceptual distinction between predictive tools and computational models is argued on empirical grounds without load-bearing self-citation chains or ansatz smuggling. The derivation chain remains self-contained against external brain data benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The central claim rests on the empirical stability of cross-region patterns in brain data and the normative premise that reproducing those patterns is required for computational similarity claims.

axioms (2)

domain assumption Cross-region alignment patterns are highly stable across different human subjects
Stated directly in the abstract as a basis for using them as a benchmark.
ad hoc to paper Reproducing relational alignment patterns is necessary for a model to qualify as a computational model of human visual cortex
The final argument distinguishes tool-level prediction from computational similarity and requires the stronger evidence.

invented entities (1)

alignment patterns no independent evidence
purpose: Characteristic functional relationship profiles of each brain region to all others, used as a second-order test
Newly defined concept introduced to expose limits of standard alignment measures.

pith-pipeline@v0.9.0 · 5601 in / 1281 out tokens · 39876 ms · 2026-05-08T13:09:08.631410+00:00 · methodology

discussion (0)

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Decoding Alignment without Encoding Alignment: A critique of similarity analysis in neuroscience
q-bio.NC 2026-05 unverdicted novelty 6.0

Decoding alignment metrics can remain high and unchanged even when encoding manifold topology is causally altered, so they do not imply similar function or computation across neural populations.

Reference graph

Works this paper leans on

68 extracted references · 24 canonical work pages · cited by 1 Pith paper · 3 internal anchors

[1]

Database-friendly random projections

Dimitris Achlioptas. Database-friendly random projections. Proceedings of the twentieth ACM SIGMOD-SIGACT-SIGART symposium on Principles of database systems, 2001. URL https://api.semanticscholar.org/CorpusID:2640788

2001
[2]

Database-friendly random projections: Johnson-lindenstrauss with binary coins.Journal of Computer and System Sciences, 66(4):671 – 687, 2003

Dimitris Achlioptas. Database-friendly random projections: Johnson-lindenstrauss with binary coins. Journal of Computer and System Sciences, 66 0 (4): 0 671--687, 2003. ISSN 0022-0000. doi:https://doi.org/10.1016/S0022-0000(03)00025-4. URL https://www.sciencedirect.com/science/article/pii/S0022000003000254. Special Issue on PODS 2001

work page doi:10.1016/s0022-0000(03)00025-4 2003
[3]

How aligned are different alignment metrics? ICLR 2024 Workshop on Representational Alignment, 2024

Jannis Ahlert, Thomas Klein, Felix Wichmann, and Robert Geirhos. How aligned are different alignment metrics? ICLR 2024 Workshop on Representational Alignment, 2024

2024
[4]

A massive 7T fMRI dataset to bridge cognitive neuroscience and artificial intelligence

Emily J Allen, Ghislain St-Yves, Yihan Wu, Jesse L Breedlove, Jacob S Prince, Logan T Dowdle, Matthias Nau, Brad Caron, Franco Pestilli, Ian Charest, J Benjamin Hutchinson, Thomas Naselaris, and Kendrick Kay. A massive 7T fMRI dataset to bridge cognitive neuroscience and artificial intelligence. Nat. Neurosci., 25 0 (1): 0 116--126, January 2022

2022
[5]

V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning

Mido Assran, Adrien Bardes, David Fan, Quentin Garrido, Russell Howes, Mojtaba, Komeili, Matthew Muckley, Ammar Rizvi, Claire Roberts, Koustuv Sinha, Artem Zholus, Sergio Arnaud, Abha Gejji, Ada Martin, Francois Robert Hogan, Daniel Dugas, Piotr Bojanowski, Vasil Khalidov, Patrick Labatut, Francisco Massa, Marc Szafraniec, Kapil Krishnakumar, Yong Li, Xia...

work page internal anchor Pith review arXiv 2025
[6]

Harvey, Scott L

Joao Barbosa, Amin Nejatbakhsh, Lyndon Duong, Sarah E. Harvey, Scott L. Brincat, Markus Siegel, Earl K. Miller, and Alex H. Williams. Quantifying differences in neural population activity with shape metrics, 1 2025. URL http://biorxiv.org/lookup/doi/10.1101/2025.01.10.632411

work page doi:10.1101/2025.01.10.632411 2025
[7]

Unifying (machine) vision via counterfactual world modeling

Daniel M Bear, Kevin Feigelis, Honglin Chen, Wanhee Lee, Rahul Venkatesh, Klemen Kotar, Alex Durango, and Daniel L K Yamins. Unifying (machine) vision via counterfactual world modeling. arXiv [cs.CV], June 2023

2023
[8]

Controlling the false discovery rate: A practical and powerful approach to multiple testing

Yoav Benjamini and Yoav Hochberg. Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society, 57, 1995

1995
[9]

Is space-time attention all you need for video understanding

Gedas Bertasius, Heng Wang, and Lorenzo Torresani. Is space-time attention all you need for video understanding? CoRR, abs/2102.05095, 2021. URL https://arxiv.org/abs/2102.05095

work page arXiv 2021
[10]

Deep problems with neural network models of human vision

Jeffrey S Bowers, Gaurav Malhotra, Marin Dujmović, Milton Llera Montero, Christian Tsvetkov, Valerio Biscione, Guillermo Puebla, Federico Adolfi, John E Hummel, Rachel F Heaton, Benjamin D Evans, Jeffrey Mitchell, and Ryan Blything. Deep problems with neural network models of human vision. Behav. Brain Sci., 46: 0 e385, December 2022

2022
[11]

Bowers, Gaurav Malhotra, Marin Dujmović, Milton Llera Montero, Christian Tsvetkov, Valerio Biscione, Guillermo Puebla, Federico Adolfi, John E

Jeffrey S. Bowers, Gaurav Malhotra, Marin Dujmović, Milton Llera Montero, Christian Tsvetkov, Valerio Biscione, Guillermo Puebla, Federico Adolfi, John E. Hummel, Rachel F. Heaton, Benjamin D. Evans, Jeffrey Mitchell, and Ryan Blything. Deep problems with neural network models of human vision. Behavioral and Brain Sciences, 46, 12 2023. ISSN 14691825. doi...

work page doi:10.1017/s0140525x22002813 2023
[12]

A spectral theory of neural prediction and alignment

Abdulkadir Canatar, Jenelle Feather, Albert Wakhloo, and SueYeon Chung. A spectral theory of neural prediction and alignment. In A Oh, T Naumann, A Globerson, K Saenko, M Hardt, and S Levine (eds.), Advances in Neural Information Processing Systems, volume 36, pp.\ 47052--47080. Curran Associates, Inc., 2023. URL https://proceedings.neurips.cc/paper_files...

2023
[13]

Explanatory models in neuroscience, part 1: Taking mechanistic abstraction seriously

Rosa Cao and Daniel Yamins. Explanatory models in neuroscience, part 1: Taking mechanistic abstraction seriously. Cognitive Systems Research, 87: 0 101244, 9 2024 a . ISSN 1389-0417. doi:10.1016/J.COGSYS.2024.101244

work page doi:10.1016/j.cogsys.2024.101244 2024
[14]

Explanatory models in neuroscience, part 2: Functional intelligibility and the contravariance principle

Rosa Cao and Daniel Yamins. Explanatory models in neuroscience, part 2: Functional intelligibility and the contravariance principle. Cognitive Systems Research, 85, 6 2024 b . ISSN 13890417. doi:10.1016/j.cogsys.2023.101200

work page doi:10.1016/j.cogsys.2023.101200 2024
[15]

Brainlife paper - human connectome young adult - full dataset, 2023

Brad Caron and Franco Pestilli. Brainlife paper - human connectome young adult - full dataset, 2023. URL https://brainlife.io/pub/640a3f9dc538c16a826f9b1a

2023
[16]

Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey E. Hinton. A simple framework for contrastive learning of visual representations. CoRR, abs/2002.05709, 2020. URL https://arxiv.org/abs/2002.05709

work page internal anchor Pith review arXiv 2002
[17]

The algonauts project 2021 challenge: How the human brain makes sense of a world in motion

R M Cichy, K Dwivedi, B Lahner, A Lascelles, P Iamshchinina, M Graumann, A Andonian, N A R Murty, K Kay, G Roig, and A Oliva. The algonauts project 2021 challenge: How the human brain makes sense of a world in motion. arXiv [cs.CV], April 2021

2021
[18]

The algonauts project: A platform for communication between the sciences of biological and artificial intelligence

Radoslaw Martin Cichy, Gemma Roig, Alex Andonian, Kshitij Dwivedi, Benjamin Lahner, Alex Lascelles, Yalda Mohsenzadeh, Kandan Ramakrishnan, and Aude Oliva. The algonauts project: A platform for communication between the sciences of biological and artificial intelligence. 5 2019. URL http://arxiv.org/abs/1905.05675

work page arXiv 2019
[19]

Openmmlab's next generation video understanding toolbox and benchmark

MMAction2 Contributors. Openmmlab's next generation video understanding toolbox and benchmark. https://github.com/open-mmlab/mmaction2, 2020

2020
[20]

A large-scale examination of inductive biases shaping high-level visual representation in brains and machines

Colin Conwell, Jacob S Prince, Kendrick N Kay, George A Alvarez, and Talia Konkle. A large-scale examination of inductive biases shaping high-level visual representation in brains and machines. Nat. Commun., 15 0 (1): 0 9383, October 2024

2024
[21]

TRIBE : TRImodal brain encoder for whole-brain fMRI response prediction

Stéphane d'Ascoli, Jérémy Rapin, Yohann Benchetrit, Hubert Banville, and Jean-Rémi King. TRIBE : TRImodal brain encoder for whole-brain fMRI response prediction. arXiv [cs.LG], July 2025

2025
[22]

Feather, M

Jenelle Feather, Meenakshi Khosla, N. Apurva Ratan Murty, and Aran Nayebi. Brain-model evaluations need the neuroai turing test, 2025. URL https://arxiv.org/abs/2502.16238

work page arXiv 2025
[23]

Distributed hierarchical processing in the primate cerebral cortex

Daniel J Felleman and David C Van Essen. Distributed hierarchical processing in the primate cerebral cortex. Cerebral Cortex, 1: 0 1--47, 1991. ISSN 10473211. doi:10.1093/cercor/1.1.1. URL https://academic.oup.com/cercor/article/1/1/1/408896

work page doi:10.1093/cercor/1.1.1 1991
[24]

The algonauts project 2023 challenge: How the human brain makes sense of natural scenes

A T Gifford, B Lahner, S Saba-Sadiya, M G Vilas, A Lascelles, A Oliva, K Kay, G Roig, and R M Cichy. The algonauts project 2023 challenge: How the human brain makes sense of natural scenes. arXiv [cs.CV], January 2023

2023
[25]

The algonauts project 2025 challenge: How the human brain makes sense of multimodal movies

Alessandro T Gifford, Domenic Bersch, Marie St-Laurent, Basile Pinsard, Julie Boyle, Lune Bellec, Aude Oliva, Gemma Roig, and Radoslaw M Cichy. The algonauts project 2025 challenge: How the human brain makes sense of multimodal movies. arXiv [q-bio.NC], December 2024

2025
[26]

A multi-modal parcellation of human cerebral cortex

Matthew F Glasser, Timothy S Coalson, Emma C Robinson, Carl D Hacker, John Harwell, Essa Yacoub, Kamil Ugurbil, Jesper Andersson, Christian F Beckmann, Mark Jenkinson, et al. A multi-modal parcellation of human cerebral cortex. Nature, 536 0 (7615): 0 171--178, 2016

2016
[27]

Priya Goyal, Quentin Duval, Jeremy Reizenstein, Matthew Leavitt, Min Xu, Benjamin Lefaudeux, Mannat Singh, Vinicius Reis, Mathilde Caron, Piotr Bojanowski, Armand Joulin, and Ishan Misra. Vissl. https://github.com/facebookresearch/vissl, 2021

2021
[28]

Deep neural networks reveal a gradient in the complexity of neural representations across the ventral stream

Umut G \"u c l \"u and Marcel AJ Van Gerven. Deep neural networks reveal a gradient in the complexity of neural representations across the ventral stream. Journal of Neuroscience, 35 0 (27): 0 10005--10014, 2015

2015
[29]

brainlife

Soichi Hayashi, Bradley A Caron, Anibal S \'o lon Heinsfeld, Sophia Vinci-Booher, Brent McPherson, Daniel N Bullock, Giulia Bert \`o , Guiomar Niso, Sandra Hanekamp, Daniel Levitas, et al. brainlife. io: a decentralized and open-source cloud platform to support neuroscience research. Nature methods, 21 0 (5): 0 809--813, 2024

2024
[30]

Deep Residual Learning for Image Recognition

Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep Residual Learning for Image Recognition . In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition ( CVPR ) , pp.\ 770--778, June 2016

2016
[31]

Masked Autoencoders Are Scalable Vision Learners

Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Doll \'a r, and Ross Girshick. Masked Autoencoders Are Scalable Vision Learners . In Proceedings of the IEEE / CVF Conference on Computer Vision and Pattern Recognition ( CVPR ) , pp.\ 16000--16009, June 2022

2022
[32]

Position: The platonic representation hypothesis

Minyoung Huh, Brian Cheung, Tongzhou Wang, and Phillip Isola. Position: The platonic representation hypothesis. Technical report, 2024

2024
[33]

Kohitij Kar and James J. DiCarlo. The quest for an integrated set of neural mechanisms underlying object recognition in primates. Annual Review of Vision Science, 10: 0 91--121, 9 2024. ISSN 2374-4642. doi:10.1146/annurev-vision-112823-030616. URL https://www.annualreviews.org/content/journals/10.1146/annurev-vision-112823-030616

work page doi:10.1146/annurev-vision-112823-030616 2024
[34]

Similarity of neural network models: A survey of functional and representational measures

Max Klabunde, Tobias Schumacher, Markus Strohmaier, and Florian Lemmerich. Similarity of neural network models: A survey of functional and representational measures. ACM Computing Surveys, 57 0 (9): 0 1--52, 2025

2025
[35]

Neural tuning and representational geometry

Nikolaus Kriegeskorte and Xue Xin Wei. Neural tuning and representational geometry. Nature Reviews Neuroscience, 22: 0 703--718, 2021. ISSN 14710048. doi:10.1038/s41583-021-00502-3. URL http://dx.doi.org/10.1038/s41583-021-00502-3

work page doi:10.1038/s41583-021-00502-3 2021
[36]

URLhttps://doi.org/10.3389/neuro.06.004.2008

Nikolaus Kriegeskorte, Marieke Mur, and Peter A. Bandettini. Representational similarity analysis - connecting the branches of systems neuroscience. 2, 2008. ISSN 1662-5137. doi:10.3389/neuro.06.004.2008. URL https://www.frontiersin.org/journals/systems-neuroscience/articles/10.3389/neuro.06.004.2008/full. Publisher: Frontiers

work page doi:10.3389/neuro.06.004.2008 2008
[37]

Apurva Ratan Murty, Kendrick Kay, Aude Oliva, and Radoslaw Cichy

Benjamin Lahner, Kshitij Dwivedi, Polina Iamshchinina, Monika Graumann, Alex Lascelles, Gemma Roig, Alessandro Thomas Gifford, Bowen Pan, SouYoung Jin, N. Apurva Ratan Murty, Kendrick Kay, Aude Oliva, and Radoslaw Cichy. Modeling short visual events through the BOLD moments video fMRI dataset and metadata. 15 0 (1): 0 6241, 2024. ISSN 2041-1723. doi:10.10...

work page doi:10.1038/s41467-024-50310-3 2024
[38]

Mvitv2: Improved multiscale vision transformers for classification and detection

Yanghao Li, Chao-Yuan Wu, Haoqi Fan, Karttikeya Mangalam, Bo Xiong, Jitendra Malik, and Christoph Feichtenhofer. Mvitv2: Improved multiscale vision transformers for classification and detection. In CVPR, 2022

2022
[39]

Better artificial intelligence does not mean better models of biology

Drew Linsley, Pinyuan Feng, and Thomas Serre. Better artificial intelligence does not mean better models of biology. Trends in Cognitive Sciences, 2025. ISSN 1879307X. doi:10.1016/j.tics.2025.11.016

work page doi:10.1016/j.tics.2025.11.016 2025
[40]

Video swin transformer

Ze Liu, Jia Ning, Yue Cao, Yixuan Wei, Zheng Zhang, Stephen Lin, and Han Hu. Video swin transformer. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 3202--3211, 2022 a

2022
[41]

A convnet for the 2020s.arXiv preprint arXiv:2201.03545, 2022

Zhuang Liu, Hanzi Mao, Chao - Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, and Saining Xie. A convnet for the 2020s. CoRR, abs/2201.03545, 2022 b . URL https://arxiv.org/abs/2201.03545

work page arXiv 2022
[42]

Brain hierarchy score: Which deep neural networks are hierarchically brain-like? IScience, 24 0 (9), 2021

Soma Nonaka, Kei Majima, Shuntaro C Aoki, and Yukiyasu Kamitani. Brain hierarchy score: Which deep neural networks are hierarchically brain-like? IScience, 24 0 (9), 2021

2021
[43]

Pedregosa, G

F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay. Scikit-learn: Machine learning in P ython. Journal of Machine Learning Research, 12: 0 2825--2830, 2011

2011
[44]

Diffusion tensor mr imaging of the human brain

Carlo Pierpaoli, Peter Jezzard, Peter J Basser, Alan Barnett, and Giovanni Di Chiro. Diffusion tensor mr imaging of the human brain. Radiology, 201 0 (3): 0 637--648, 1996

1996
[45]

Learning Transferable Visual Models From Natural Language Supervision

Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from natural language supervision, 2021. URL https://arxiv.org/abs/2103.00020

work page internal anchor Pith review arXiv 2021
[46]

One hundred neural networks and brains watching videos: Lessons from alignment

Christina Sartzetaki, Gemma Roig, Cees GM Snoek, and Iris IA Groen. One hundred neural networks and brains watching videos: Lessons from alignment. bioRxiv, pp.\ 2024--12, 2024

2024
[47]

Zamir, Leonidas J

Alexander Sax, Bradley Emi, Amir R. Zamir, Leonidas J. Guibas, Silvio Savarese, and Jitendra Malik. Mid-level visual representations improve generalization and sample efficiency for learning visuomotor policies. 2018

2018
[48]

No free lunch from deep learning in neuroscience: A case study through models of the entorhinal-hippocampal circuit

Rylan Schaeffer, Mikail Khona, and Ila Fiete. No free lunch from deep learning in neuroscience: A case study through models of the entorhinal-hippocampal circuit. In S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh (eds.), Advances in Neural Information Processing Systems, volume 35, pp.\ 16052--16067. Curran Associates, Inc., 2022. URL h...

work page arXiv 2022
[49]

Position: Maximizing neural regression scores may not identify good models of the brain

Rylan Schaeffer, Mikail Khona, Sarthak Chandra, Mitchell Ostrow, Brando Miranda, and Sanmi Koyejo. Position: Maximizing neural regression scores may not identify good models of the brain. In UniReps: 2nd Edition of the Workshop on Unifying Representations in Neural Models, October 2024

2024
[50]

Brain-score: Which artificial neural network for object recognition is most brain-like? BioRxiv, pp.\ 407007, 2018

Martin Schrimpf, Jonas Kubilius, Ha Hong, Najib J Majaj, Rishi Rajalingham, Elias B Issa, Kohitij Kar, Pouya Bashivan, Jonathan Prescott-Roy, Franziska Geiger, et al. Brain-score: Which artificial neural network for object recognition is most brain-like? BioRxiv, pp.\ 407007, 2018

2018
[51]

Integrative benchmarking to advance neurally mechanistic models of human intelligence

Martin Schrimpf, Jonas Kubilius, Michael J Lee, N Apurva Ratan Murty, Robert Ajemian, and James J DiCarlo. Integrative benchmarking to advance neurally mechanistic models of human intelligence. Neuron, 108 0 (3): 0 413--423, 2020

2020
[52]

The neural architecture of language: Integrative modeling converges on predictive processing

Martin Schrimpf, Idan Asher Blank, Greta Tuckute, Carina Kauf, Eghbal A Hosseini, Nancy Kanwisher, Joshua B Tenenbaum, and Evelina Fedorenko. The neural architecture of language: Integrative modeling converges on predictive processing. Proceedings of the National Academy of Sciences, 118 0 (45): 0 e2105646118, 2021

2021
[53]

Anatomically-constrained tractography: improved diffusion mri streamlines tractography through effective use of anatomical information

Robert E Smith, Jacques-Donald Tournier, Fernando Calamante, and Alan Connelly. Anatomically-constrained tractography: improved diffusion mri streamlines tractography through effective use of anatomical information. Neuroimage, 62 0 (3): 0 1924--1938, 2012

1924
[54]

Conclusions about neural network to brain alignment are profoundly impacted by the similarity measure

Ansh Soni, Sudhanshu Srivastava, Konrad Kording, and Meenakshi Khosla. Conclusions about neural network to brain alignment are profoundly impacted by the similarity measure. bioRxiv, pp.\ 2024.08.07.607035, August 2024

2024
[55]

Self-supervised learning of motion concepts by optimizing counterfactuals, 2025

Stefan Stojanov, David Wendt, Seungwoo Kim, Rahul Venkatesh, Kevin Feigelis, Jiajun Wu, and Daniel LK Yamins. Self-supervised learning of motion concepts by optimizing counterfactuals, 2025. URL https://arxiv.org/abs/2503.19953

work page arXiv 2025
[56]

Many-two-one: Diverse representations across visual pathways emerge from a single objective

Yingtian Tang, Abdulkadir Gokce, Khaled Jedoui Al-Karkari, Daniel Yamins, and Martin Schrimpf. Many-two-one: Diverse representations across visual pathways emerge from a single objective. bioRxiv, pp.\ 2025.07.22.664908, July 2025

2025
[57]

Model-brain comparison using inter-animal transforms, 2025

Imran Thobani, Javier Sagastuy-Brena, Aran Nayebi, Jacob Prince, Rosa Cao, and Daniel Yamins. Model-brain comparison using inter-animal transforms, 2025. URL https://arxiv.org/abs/2510.02523

work page arXiv 2025
[58]

Vggt: Visual geometry grounded transformer

Jianyuan Wang, Minghao Chen, Nikita Karaev, Andrea Vedaldi, Christian Rupprecht, and David Novotny. Vggt: Visual geometry grounded transformer. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2025

2025
[59]

Wichmann and Robert Geirhos

Felix A. Wichmann and Robert Geirhos. Are deep neural networks adequate behavioral models of human visual perception? Annual Review of Vision Science, 9: 0 501--524, 9 2023. ISSN 23744650. doi:10.1146/ANNUREV-VISION-120522-031739/CITE/REFWORKS. URL https://www.annualreviews.org/content/journals/10.1146/annurev-vision-120522-031739

work page doi:10.1146/annurev-vision-120522-031739/cite/refworks 2023
[60]

Pytorch image models

Ross Wightman. Pytorch image models. https://github.com/rwightman/pytorch-image-models, 2019

2019
[61]

Measuring the measures: Discriminative capacity of representational similarity metrics across model families

Jialin Wu, Shreya Saha, Yiqing Bo, and Meenakshi Khosla. Measuring the measures: Discriminative capacity of representational similarity metrics across model families. arXiv [cs.LG], September 2025

2025
[62]

Using goal-driven deep learning models to understand sensory cortex

Daniel L K Yamins and James J DiCarlo. Using goal-driven deep learning models to understand sensory cortex. 19 0 (3): 0 356--365, March 2016

2016
[63]

Performance-optimized hierarchical models predict neural responses in higher visual cortex

Daniel LK Yamins, Ha Hong, Charles F Cadieu, Ethan A Solomon, Darren Seibert, and James J DiCarlo. Performance-optimized hierarchical models predict neural responses in higher visual cortex. Proceedings of the national academy of sciences, 111 0 (23): 0 8619--8624, 2014

2014
[64]

Taskonomy: Disentangling task transfer learning, 2018

Amir Zamir, Alexander Sax, William Shen, Leonidas Guibas, Jitendra Malik, and Silvio Savarese. Taskonomy: Disentangling task transfer learning, 2018. URL https://arxiv.org/abs/1804.08328

work page arXiv 2018
[65]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...
[66]

@esa (Ref

\@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should not add it explicitly Type <Return> for now, but then later remove the command n...
[67]

\@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@first@sw \@firstoftwo \@ifundefined NAT@b*@#2 \@firstoftwo @num @NAT@ctr \@secondoft...
[68]

@open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibsetup #1 @NAT@ctr @ @openbib .11em \@plus.33em \@minus.07em 4000 4000 `\.\@m @bibit...

work page arXiv