arxiv: 2605.08495 · v1 · submitted 2026-05-08 · 💻 cs.LG · q-bio.NC

Recognition: 2 theorem links

· Lean Theorem

NeuralBench: A Unifying Framework to Benchmark NeuroAI Models

Antoine Ratouchniak, Elisa Cascardi, Hubert Banville, Jarod L\'evy, Jean-R\'emi King, J\'er\'emy Rapin, Katelyn Begany, Marl\`ene Careil, Mingfang (Lucy) Zhang, Saarang Panchavati, Simon Dahan, St\'ephane d'Ascoli, Teon Brooks, Yohann Benchetrit

Pith reviewed 2026-05-12 02:07 UTC · model grok-4.3

classification 💻 cs.LG q-bio.NC

keywords NeuroAIEEGbenchmarking frameworkfoundation modelsbrain recordingsdeep learningstandardized evaluationcognitive decoding

0 comments

The pith

NeuralBench introduces a standardized framework for evaluating AI models on brain recordings, showing foundation models only marginally outperform task-specific ones.

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

The paper establishes NeuralBench as a unifying open framework to benchmark models that process brain signals, addressing inconsistent preprocessing and narrow task sets in existing work. It pairs this with an initial EEG release covering 36 tasks and 14 architectures across 94 datasets accessed uniformly. Evaluations using the framework demonstrate that large foundation models deliver only small gains over simpler task-specific approaches. A wide range of tasks, including cognitive decoding and clinical predictions, stay difficult even for the strongest models tested. The structure supports straightforward additions of new tasks, datasets, models, and modalities such as MEG or fMRI.

Core claim

NeuralBench is a unified framework for benchmarking AI models of brain activity, accompanied by NeuralBench-EEG v1.0 which evaluates 14 deep learning architectures on 36 EEG tasks across 94 datasets through a standardized interface. This shows foundation models only marginally outperform task-specific models and that many tasks remain highly challenging.

What carries the argument

NeuralBench framework with its standardized interface for tasks, datasets, models, and preprocessing pipelines, which enables consistent evaluation and extensibility to new modalities.

Load-bearing premise

That the chosen preprocessing pipelines, task definitions, and dataset selection through the standardized interface produce unbiased relative rankings of models without favoring certain architectures or data characteristics.

What would settle it

Re-evaluating the same models on the same datasets but with alternative reasonable preprocessing pipelines that reverses the observed performance ordering between foundation models and task-specific models.

read the original abstract

Deep learning and large public datasets have recently catalyzed the proliferation of AI models for processing brain recordings. However, systematically evaluating these models remains a challenge: not only do the preprocessing pipelines, training and finetuning approaches largely vary across studies, but their downstream evaluation is often limited to small sets of tasks and/or datasets. Here, we present NeuralBench: a unified framework for benchmarking AI models of brain activity. We accompany this framework with NeuralBench-EEG v1.0 -- a large EEG benchmark that includes 36 electroencephalography (EEG) tasks and 14 deep learning architectures, and is evaluated on 94 datasets accessed through a standardized interface. This first EEG-focused release already highlights two main findings. First, current foundation models only marginally outperform task-specific models. Second, a large set of tasks (e.g. cognitive decoding, clinical predictions) remain highly challenging, even for the best models. Critically, NeuralBench is designed for the integration of new tasks, datasets, models, and neuroimaging modalities, as illustrated by preliminary extensions to MEG and fMRI datasets and models. Through this white paper, we invite the community to expand this open-source framework and work together toward a unified benchmarking standard for neuroimaging models.

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

NeuralBench gives NeuroAI a practical large-scale EEG benchmark with 36 tasks across 94 datasets, but its claims of marginal foundation-model gains rest on untested standardization and lack statistical backing.

read the letter

NeuralBench introduces a unified open framework and its first EEG release that standardizes evaluation across 36 tasks, 14 architectures, and 94 public datasets. It also sketches quick extensions to MEG and fMRI. That scale and the shared interface are bigger than most prior single-study comparisons, and the design for community additions is a clear practical step forward. The observation that many cognitive and clinical tasks remain hard even for the strongest models is useful for focusing future work. The paper does a decent job pulling existing data together under one API and making the code available so others can reproduce or extend the setup without starting from scratch. The two headline findings are new in this combined form, even if they build on earlier scattered results. The softer parts sit in the analysis. The manuscript gives little detail on how tasks were chosen or how dataset biases were checked, and it reports no statistical tests, confidence intervals, or robustness checks on the performance gaps. A single preprocessing and task pipeline could easily favor certain model families, yet there are no ablations across alternatives to test that. Without those steps the “marginally outperform” and “highly challenging” statements stay descriptive rather than quantified. This is for NeuroAI researchers who need a common starting point for model comparisons or who want to add their own tasks and modalities. A reader already working on brain-signal models would get immediate value from the resource itself. I would send it to peer review because the benchmark and its scale are new enough to deserve referee input, especially on tightening the empirical claims and documenting curation choices.

Referee Report

3 major / 2 minor

Summary. The manuscript introduces NeuralBench, a unifying open-source framework for benchmarking NeuroAI models on brain recordings, with an initial EEG-focused release (NeuralBench-EEG v1.0) that standardizes access to 94 datasets, defines 36 tasks, and evaluates 14 deep learning architectures (including foundation models and task-specific ones). The central empirical claims are that foundation models only marginally outperform task-specific models and that a substantial subset of tasks (e.g., cognitive decoding, clinical predictions) remain highly challenging even for the best-performing models. The framework is presented as extensible to additional modalities such as MEG and fMRI.

Significance. If the standardization proves robust, this work offers a valuable contribution to NeuroAI by addressing the fragmentation of preprocessing pipelines, training protocols, and evaluation tasks that currently hinders systematic model comparison. The scale of the benchmark (94 datasets, 36 tasks) and its open design with community-invitation elements provide a practical foundation for reproducible progress; the explicit release of a common interface is a concrete strength that could accelerate falsifiable comparisons across architectures.

major comments (3)

[§4] §4 (Experiments and Results): The headline claim that foundation models 'only marginally outperform task-specific models' is presented without statistical quantification of the deltas (e.g., no paired tests, bootstrap CIs, or multiple-comparison correction across the 36 tasks), leaving the 'marginal' qualifier unanchored and vulnerable to the possibility that observed differences fall within noise.
[§3.2] §3.2 (Task and Dataset Interface): No ablation or sensitivity analysis is reported on alternative preprocessing pipelines, normalization choices, or task-definition variants; without these, it is impossible to rule out that the reported relative rankings and identification of 'highly challenging' tasks are artifacts of the single chosen standardization rather than intrinsic model properties.
[§2 and §3.1] §2 and §3.1 (Benchmark Construction): The criteria used to select the 36 tasks and 94 datasets are described at a high level but lack explicit documentation of inclusion/exclusion rules, potential dataset biases (e.g., class imbalance, recording quality), or coverage of the broader EEG literature, which directly affects the generalizability of the 'remain highly challenging' conclusion.

minor comments (2)

[Figures/Tables] Figure 2 and Table 1: Axis labels and legend entries are too small for readability in print; consider increasing font size and adding error bars or statistical annotations directly on the performance plots.
[§5] §5 (Discussion): The extensibility claims for MEG/fMRI are illustrated only with preliminary results; a short dedicated subsection with concrete code snippets or interface examples would strengthen the invitation for community contributions.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. We address each major comment below and describe the revisions we will make to improve the manuscript.

read point-by-point responses

Referee: [§4] §4 (Experiments and Results): The headline claim that foundation models 'only marginally outperform task-specific models' is presented without statistical quantification of the deltas (e.g., no paired tests, bootstrap CIs, or multiple-comparison correction across the 36 tasks), leaving the 'marginal' qualifier unanchored and vulnerable to the possibility that observed differences fall within noise.

Authors: We agree that statistical quantification is needed to support the 'marginal' claim. In the revised manuscript, we will add paired non-parametric tests (Wilcoxon signed-rank) comparing foundation models against task-specific models across tasks, report bootstrap confidence intervals on the performance deltas, and apply multiple-comparison correction (FDR) across the 36 tasks. These changes will provide a rigorous anchor for the observed differences. revision: yes
Referee: [§3.2] §3.2 (Task and Dataset Interface): No ablation or sensitivity analysis is reported on alternative preprocessing pipelines, normalization choices, or task-definition variants; without these, it is impossible to rule out that the reported relative rankings and identification of 'highly challenging' tasks are artifacts of the single chosen standardization rather than intrinsic model properties.

Authors: We acknowledge that sensitivity analyses would strengthen the robustness claims. A full ablation across all 94 datasets and 36 tasks is computationally prohibitive for this initial release. In revision, we will add a targeted sensitivity analysis on a representative subset of tasks for key choices (normalization and filtering), justify the selected pipeline against existing standards, and add an explicit limitations discussion noting that broader sensitivity testing is planned as future community-driven work. revision: partial
Referee: [§2 and §3.1] §2 and §3.1 (Benchmark Construction): The criteria used to select the 36 tasks and 94 datasets are described at a high level but lack explicit documentation of inclusion/exclusion rules, potential dataset biases (e.g., class imbalance, recording quality), or coverage of the broader EEG literature, which directly affects the generalizability of the 'remain highly challenging' conclusion.

Authors: We will expand Sections 2 and 3.1 with explicit inclusion/exclusion criteria (public availability, minimum subject count, task relevance to NeuroAI). We will also report dataset-level statistics on class balance and recording quality where available, and add a discussion comparing our selection to recent EEG literature surveys to clarify coverage and potential biases. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical benchmark results are computed from external datasets and models

full rationale

The paper introduces a benchmarking framework and reports performance numbers obtained by executing published models on public EEG datasets through a fixed preprocessing and task interface. No equations derive predictions from first principles, no parameters are fitted to the reported deltas, and no self-citation chain is invoked to justify uniqueness or force the central claims. The two headline findings (marginal foundation-model gains and persistently hard tasks) are direct empirical outputs of the benchmark run, not algebraic rearrangements of the input data or prior author results. The standardization choices affect absolute numbers but do not create a definitional loop or fitted-input prediction; they are methodological decisions whose validity can be tested by re-running the open framework on alternative pipelines.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claims rest on the representativeness of the selected tasks and the fairness of the standardized preprocessing and evaluation protocol; no free parameters or invented entities are introduced.

axioms (1)

domain assumption A single standardized preprocessing and evaluation interface produces comparable and unbiased model rankings across diverse EEG datasets and tasks.
Invoked when claiming that the benchmark reveals true relative performance differences.

pith-pipeline@v0.9.0 · 5591 in / 1238 out tokens · 47537 ms · 2026-05-12T02:07:58.297612+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
NeuralBench is built on top of three main Python packages... NeuralFetch... NeuralSet... NeuralTrain... task configuration through lightweight YAML files
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear
We accompany this framework with NeuralBench-EEG v1.0 — a large EEG benchmark that includes 36 electroencephalography (EEG) tasks and 14 deep learning architectures

Reference graph

Works this paper leans on

290 extracted references · 290 canonical work pages · 4 internal anchors

[1]

Machine learning for neuroimaging with scikit-learn

Alexandre Abraham, Fabian Pedregosa, Michael Eickenberg, Philippe Gervais, Andreas Mueller, Jean Kossaifi, Alexandre Gramfort, Bertrand Thirion, and Ga \"e l Varoquaux. Machine learning for neuroimaging with scikit-learn. Frontiers in neuroinformatics, 8: 0 14, 2014

work page 2014
[2]

Gaze-independent BCI-spelling using rapid serial visual presentation ( RSVP )

Laura Acqualagna and Benjamin Blankertz. Gaze-independent BCI-spelling using rapid serial visual presentation ( RSVP ). Clinical Neurophysiology, 124 0 (5): 0 901--908, May 2013. ISSN 1388-2457. doi:10.1016/j.clinph.2012.12.050. http://dx.doi.org/10.1016/j.clinph.2012.12.050

work page doi:10.1016/j.clinph.2012.12.050 2013
[3]

Albrecht, James A

Matthew A. Albrecht, James A. Waltz, James F. Cavanagh, Michael J. Frank, and James M. Gold. Increased Conflict-Induced slowing, but no differences in Conflict-Induced positive or negative prediction error learning in patients with schizophrenia. Neuropsychologia, 123: 0 131--140, February 2019. ISSN 1873-3514. doi:10.1016/j.neuropsychologia.2018.04.031

work page doi:10.1016/j.neuropsychologia.2018.04.031 2019
[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, et al. A massive 7T fMRI dataset to bridge cognitive neuroscience and artificial intelligence. Nature neuroscience, 25 0 (1): 0 116--126, 2022

work page 2022
[5]

Physics-informed attention temporal convolutional network for eeg-based motor imagery classification

Hamdi Altaheri, Ghulam Muhammad, and Mansour Alsulaiman. Physics-informed attention temporal convolutional network for eeg-based motor imagery classification. IEEE Transactions on Industrial Informatics, 2022. doi:10.1109/TII.2022.3197419

work page doi:10.1109/tii.2022.3197419 2022
[6]

Influence of P300 latency jitter on event related potential-based brain--computer interface performance

Pietro Aricò, Fabio Aloise, Francesca Schettini, Serenella Salinari, Donatella Mattia, and Febo Cincotti. Influence of P300 latency jitter on event related potential-based brain--computer interface performance. Journal of Neural Engineering, 11 0 (3): 0 035008, 2014. doi:10.1088/1741-2560/11/3/035008

work page doi:10.1088/1741-2560/11/3/035008 2014
[7]

Mother of all BCI benchmarks, 2025

Bruno Aristimunha, Igor Carrara, Pierre Guetschel, Sara Sedlar, Pedro Rodrigues, Jan Sosulski, Divyesh Narayanan, Erik Bjareholt, Quentin Barthelemy, Robin Tibor Schirrmeister, Reinmar Kobler, Emmanuel Kalunga, Ludovic Darmet, Cattan Gregoire, Ali Abdul Hussain, Ramiro Gatti, Vladislav Goncharenko, Anton Andreev, Jordy Thielen, Thomas Moreau, Yannick Roy,...

work page 2025
[8]

Braindecode : toolbox for decoding raw electrophysiological brain data with deep learning models, 2026

Bruno Aristimunha, Pierre Guetschel, Martin Wimpff, Lukas Gemein, Cedric Rommel, Hubert Banville, Maciej Sliwowski, Daniel Wilson, Simon Brandt, Th \'e o Gnassounou, Joseph Paillard, Aman Srivastava, Bruna Junqueira Lopes , Sara Sedlar, Thomas Moreau, Sylvain Chevallier, Alexandre Gramfort, and Robin Tibor Schirrmeister. Braindecode : toolbox for decoding...

work page 2026
[9]

A unified, scalable framework for neural population decoding

Mehdi Azabou, Vinam Arora, Venkataramana Ganesh, Ximeng Mao, Santosh Nachimuthu, Michael Mendelson, Blake Richards, Matthew Perich, Guillaume Lajoie, and Eva Dyer. A unified, scalable framework for neural population decoding. Advances in Neural Information Processing Systems, 36: 0 44937--44956, 2023

work page 2023
[10]

wav2vec 2.0: A framework for self-supervised learning of speech representations

Alexei Baevski, Yuhao Zhou, Abdelrahman Mohamed, and Michael Auli. wav2vec 2.0: A framework for self-supervised learning of speech representations. Advances in neural information processing systems, 33: 0 12449--12460, 2020

work page 2020
[11]

Magnetoencephalography for brain electrophysiology and imaging

Sylvain Baillet. Magnetoencephalography for brain electrophysiology and imaging. Nature neuroscience, 20 0 (3): 0 327--339, 2017

work page 2017
[13]

Robust control of an effector by an asynchronous EEG brain-machine interface

Alexandre Barachant. Robust control of an effector by an asynchronous EEG brain-machine interface . Theses, Universit e de Grenoble , March 2012. https://theses.hal.science/tel-01196752

work page 2012
[14]

A plug&play P300 BCI using information geometry

Alexandre Barachant and Marco Congedo. A plug&play P300 BCI using information geometry. arXiv preprint arXiv:1409.0107, 2014

work page arXiv 2014
[15]

Multiclass brain--computer interface classification by Riemannian geometry

Alexandre Barachant, St \'e phane Bonnet, Marco Congedo, and Christian Jutten. Multiclass brain--computer interface classification by Riemannian geometry. IEEE Transactions on Biomedical Engineering, 59 0 (4): 0 920--928, 2011

work page 2011
[16]

Brain decoding: toward real-time reconstruction of visual perception

Yohann Benchetrit, Hubert Banville, and Jean-R \'e mi King. Brain decoding: toward real-time reconstruction of visual perception. In ICLR 2024, 2024

work page 2024
[17]

SpeechBrain-MOABB : An open-source python library for benchmarking deep neural networks applied to EEG signals

Davide Borra, Francesco Paissan, and Mirco Ravanelli. SpeechBrain-MOABB : An open-source python library for benchmarking deep neural networks applied to EEG signals. Computers in Biology and Medicine, 182: 0 109097, 2024

work page 2024
[18]

Hierarchical structure guides rapid linguistic predictions during naturalistic listening

Jonathan R Brennan and John T Hale. Hierarchical structure guides rapid linguistic predictions during naturalistic listening. PloS one, 14 0 (1): 0 e0207741, 2019

work page 2019
[19]

Power failure: why small sample size undermines the reliability of neuroscience

Katherine S Button, John PA Ioannidis, Claire Mokrysz, Brian A Nosek, Jonathan Flint, Emma SJ Robinson, and Marcus R Munaf \`o . Power failure: why small sample size undermines the reliability of neuroscience. Nature reviews neuroscience, 14 0 (5): 0 365--376, 2013

work page 2013
[20]

Burst c-VEP based BCI : Optimizing stimulus design for enhanced classification with minimal calibration data and improved user experience

Kalou Cabrera Castillos, Simon Ladouce, Ludovic Darmet, and Frédéric Dehais. Burst c-VEP based BCI : Optimizing stimulus design for enhanced classification with minimal calibration data and improved user experience. NeuroImage, 284: 0 120446, December 2023. ISSN 1053-8119. doi:10.1016/j.neuroimage.2023.120446. http://dx.doi.org/10.1016/j.neuroimage.2023.120446

work page doi:10.1016/j.neuroimage.2023.120446 2023
[21]

BrainLM : A foundation model for brain activity recordings

Josue Ortega Caro, Antonio H de O Fonseca, Christopher Averill, Syed A Rizvi, Matteo Rosati, James L Cross, Prateek Mittal, Emanuele Zappala, Daniel Levine, Rahul M Dhodapkar, et al. BrainLM : A foundation model for brain activity recordings. BioRxiv, pages 2023--09, 2023

work page 2023
[22]

Grégoire Cattan, Anton Andreev, Pedro L. C. Rodrigues, and Marco Congedo. Dataset of an EEG-based BCI experiment in virtual reality and on a personal computer, 2019. https://zenodo.org/record/2605204

work page arXiv 2019
[23]

Ricardo Chavarriaga and José del R. Millan. Learning from EEG Error-Related potentials in noninvasive Brain-Computer interfaces. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 18 0 (4): 0 381--388, August 2010. ISSN 1558-0210. doi:10.1109/tnsre.2010.2053387. http://dx.doi.org/10.1109/TNSRE.2010.2053387

work page doi:10.1109/tnsre.2010.2053387 2010
[24]

A large finer-grained affective computing EEG dataset

Jingjing Chen, Xiaobin Wang, Chen Huang, Xin Hu, Xinke Shen, and Dan Zhang. A large finer-grained affective computing EEG dataset. Scientific Data, 10 0 (1), October 2023. ISSN 2052-4463. doi:10.1038/s41597-023-02650-w. http://dx.doi.org/10.1038/s41597-023-02650-w

work page doi:10.1038/s41597-023-02650-w 2023
[25]

EEG datasets for motor imagery brain computer interface

Hohyun Cho, Minkyu Ahn, Sangtae Ahn, Moonyoung Kwon, and Chan Jun, Sung. Supporting data for "EEG datasets for motor imagery brain computer interface", 2017. http://gigadb.org/dataset/100295

work page 2017
[27]

EEG and MEG : relevance to neuroscience

Fernando Lopes da Silva. EEG and MEG : relevance to neuroscience. Neuron, 80 0 (5): 0 1112--1128, 2013

work page 2013
[28]

BIDS CHB-MIT scalp EEG database, December 2023

Jonathan Dan and Ali Shoeb. BIDS CHB-MIT scalp EEG database, December 2023. https://doi.org/10.5281/zenodo.10259996

work page doi:10.5281/zenodo.10259996 2023
[29]

A foundation model of vision, audition, and language for in-silico neuroscience

St \'e phane d'Ascoli, J \'e r \'e my Rapin, Yohann Benchetrit, Teon Brookes, Katelyn Begany, Jos \'e phine Raugel, Hubert Banville, and Jean-R \'e mi King. A foundation model of vision, audition, and language for in-silico neuroscience. https://ai.meta.com/research/publications/a-foundation-model-of-vision-audition-and-language-for-in-silico-neuroscience/, 2026

work page 2026
[31]

BERT : Pre-training of deep bidirectional transformers for language understanding

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. BERT : Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers), pages 4171--4186, 2019

work page 2019
[32]

LUNA : Efficient and topology-agnostic foundation model for EEG signal analysis

Berkay D \"o ner, Thorir Mar Ingolfsson, Luca Benini, and Yawei Li. LUNA : Efficient and topology-agnostic foundation model for EEG signal analysis. In The Thirty-Ninth Annual Conference on Neural Information Processing Systems (NeurIPS), 2025. https://openreview.net/forum?id=uazfjnFL0G

work page 2025
[33]

Brain- JEPA : Brain dynamics foundation model with gradient positioning and spatiotemporal masking

Zijian Dong, Ruilin Li, Yilei Wu, Thuan T Nguyen, Joanna S Chong, Fang Ji, Nathanael R Tong, Christopher L Chen, and Juan H Zhou. Brain- JEPA : Brain dynamics foundation model with gradient positioning and spatiotemporal masking. Advances in Neural Information Processing Systems, 37: 0 86048--86073, 2024

work page 2024
[34]

Dornhege, B

G. Dornhege, B. Blankertz, G. Curio, and K.-R. Muller. Boosting bit rates in noninvasive EEG single-trial classifications by feature combination and multiclass paradigms. IEEE Transactions on Biomedical Engineering, 51 0 (6): 0 993--1002, June 2004. ISSN 1558-2531. doi:10.1109/tbme.2004.827088. http://dx.doi.org/10.1109/TBME.2004.827088

work page doi:10.1109/tbme.2004.827088 2004
[35]

A large EEG database with users’ profile information for motor imagery brain-computer interface research

Pauline Dreyer, Aline Roc, Léa Pillette, Sébastien Rimbert, and Fabien Lotte. A large EEG database with users’ profile information for motor imagery brain-computer interface research. Scientific Data, 10 0 (1), September 2023. ISSN 2052-4463. doi:10.1038/s41597-023-02445-z. http://dx.doi.org/10.1038/s41597-023-02445-z

work page doi:10.1038/s41597-023-02445-z 2023
[36]

Towards decoding individual words from non-invasive brain recordings

St \'e phane d’Ascoli, Corentin Bel, J \'e r \'e my Rapin, Hubert Banville, Yohann Benchetrit, Christophe Pallier, and Jean-R \'e mi King. Towards decoding individual words from non-invasive brain recordings. Nature Communications, 16 0 (1): 0 10521, 2025

work page 2025
[38]

REVE : A foundation model for EEG -- adapting to any setup with large-scale pretraining on 25,000 subjects

Yassine El Ouahidi, Jonathan Lys, Philipp Th \"o lke, Nicolas Farrugia, Bastien Pasdeloup, Vincent Gripon, Karim Jerbi, and Giulia Lioi. REVE : A foundation model for EEG -- adapting to any setup with large-scale pretraining on 25,000 subjects. In The Thirty-Ninth Annual Conference on Neural Information Processing Systems (NeurIPS), 2025. https://openrevi...

work page 2025
[39]

PyTorch Lightning , March 2019

William Falcon and The PyTorch Lightning team . PyTorch Lightning , March 2019. https://github.com/Lightning-AI/lightning

work page 2019
[40]

Autocalibration and recurrent adaptation: Towards a plug and play online ERD-BCI

Josef Faller, Carmen Vidaurre, Teodoro Solis-Escalante, Christa Neuper, and Reinhold Scherer. Autocalibration and recurrent adaptation: Towards a plug and play online ERD-BCI . IEEE Transactions on Neural Systems and Rehabilitation Engineering, 20 0 (3): 0 313--319, May 2012. ISSN 1558-0210. doi:10.1109/tnsre.2012.2189584. http://dx.doi.org/10.1109/tnsre....

work page doi:10.1109/tnsre.2012.2189584 2012
[41]

Lukas A. W. Gemein, Robin T. Schirrmeister, Patryk Chrabaszcz, Daniel Wilson, Joschka Boedecker, Andreas Schulze-Bonhage, Frank Hutter, and Tonio Ball. Machine-learning-based diagnostics of EEG pathology. NeuroImage, 220: 0 117021, 2020

work page 2020
[42]

You Snooze , You Win : The PhysioNet / Computing in Cardiology Challenge 2018

Mohammad M Ghassemi, Benjamin E Moody, Li-wei H Lehman, Christopher Song, Qiao Li, Haoqi Sun, Roger G Mark, M Brandon Westover, and Gari D Clifford. You Snooze , You Win : The PhysioNet / Computing in Cardiology Challenge 2018. Computing in cardiology, 45: 0 10.22489/cinc.2018.049, September 2018. ISSN 2325-8861. doi:10.22489/cinc.2018.049

work page doi:10.22489/cinc.2018.049 2018
[43]

Gifford, Kshitij Dwivedi, Gemma Roig, and Radoslaw M

Alessandro T. Gifford, Kshitij Dwivedi, Gemma Roig, and Radoslaw M. Cichy. A large and rich EEG dataset for modeling human visual object recognition. Neuroimage, 264: 0 119754, December 2022 a . ISSN 1053-8119. doi:10.1016/j.neuroimage.2022.119754. https://pmc.ncbi.nlm.nih.gov/articles/PMC9771828/

work page doi:10.1016/j.neuroimage.2022.119754 2022
[44]

A large and rich EEG dataset for modeling human visual object recognition

Alessandro T Gifford, Kshitij Dwivedi, Gemma Roig, and Radoslaw M Cichy. A large and rich EEG dataset for modeling human visual object recognition. NeuroImage, 264: 0 119754, 2022 b

work page 2022
[47]

Engemann, Daniel Strohmeier, Christian Brodbeck, Roman Goj, Mainak Jas, Teon Brooks, Lauri Parkkonen, and Matti Hämäläinen

Alexandre Gramfort, Martin Luessi, Eric Larson, Denis A. Engemann, Daniel Strohmeier, Christian Brodbeck, Roman Goj, Mainak Jas, Teon Brooks, Lauri Parkkonen, and Matti Hämäläinen. MEG and EEG data analysis with MNE-Python . Frontiers in Neuroscience, 7, December 2013. ISSN 1662-453X. doi:10.3389/fnins.2013.00267

work page doi:10.3389/fnins.2013.00267 2013
[48]

Grosse-Wentrup, C

M. Grosse-Wentrup, C. Liefhold, K. Gramann, and M. Buss. Beamforming in noninvasive Brain-Computer interfaces. IEEE Transactions on Biomedical Engineering, 56 0 (4): 0 1209--1219, April 2009. ISSN 1558-2531. doi:10.1109/tbme.2008.2009768. http://dx.doi.org/10.1109/TBME.2008.2009768

work page doi:10.1109/tbme.2008.2009768 2009
[49]

How many people are able to control a P300 -based brain-computer interface ( BCI )? Neuroscience Letters, 462 0 (1): 0 94--98, September 2009

Christoph Guger, Shahab Daban, Eric Sellers, Clemens Holzner, Gunther Krausz, Roberta Carabalona, Furio Gramatica, and Guenter Edlinger. How many people are able to control a P300 -based brain-computer interface ( BCI )? Neuroscience Letters, 462 0 (1): 0 94--98, September 2009. ISSN 0304-3940. doi:10.1016/j.neulet.2009.06.045. http://dx.doi.org/10.1016/j...

work page doi:10.1016/j.neulet.2009.06.045 2009
[50]

Hamid, K

A. Hamid, K. Gagliano, S. Rahman, N. Tulin, V. Tchiong, I. Obeid, and J. Picone. The Temple University Artifact Corpus : An annotated corpus of EEG artifacts. pages 1--4, 2020

work page 2020
[51]

The TUH EEG Corpus: A big data resource for automated EEG interpretation

Amir Harati, Silvia Lopez, I Obeid, J Picone, MP Jacobson, and S Tobochnik. The TUH EEG Corpus: A big data resource for automated EEG interpretation . In 2014 IEEE signal processing in medicine and biology symposium (SPMB), pages 1--5. IEEE, 2014

work page 2014
[52]

Improved EEG event classification using differential energy

Amir Harati, Meysam Golmohammadi, Silvia Lopez, Iyad Obeid, and Joseph Picone. Improved EEG event classification using differential energy. pages 1--4, 2015

work page 2015
[53]

EEG potentials predict upcoming emergency brakings during simulated driving

Stefan Haufe, Matthias S Treder, Manfred F Gugler, Max Sagebaum, Gabriel Curio, and Benjamin Blankertz. EEG potentials predict upcoming emergency brakings during simulated driving. Journal of Neural Engineering, 8 0 (5): 0 056001, July 2011. ISSN 1741-2552. doi:10.1088/1741-2560/8/5/056001. http://dx.doi.org/10.1088/1741-2560/8/5/056001

work page doi:10.1088/1741-2560/8/5/056001 2011
[55]

Hinss, Emilie S

Marcel F. Hinss, Emilie S. Jahanpour, Bertille Somon, Lou Pluchon, Frédéric Dehais, and Raphaëlle N. Roy. COG-BCI database: A multi-session and multi-task EEG cognitive dataset for passive brain-computer interfaces, July 2022

work page 2022
[56]

Hinss, Emilie S

Marcel F. Hinss, Emilie S. Jahanpour, Bertille Somon, Lou Pluchon, Frédéric Dehais, and Raphaëlle N. Roy. Open multi-session and multi-task EEG cognitive dataset for passive brain-computer interface applications. Scientific Data, 10 0 (1), February 2023. ISSN 2052-4463. doi:10.1038/s41597-022-01898-y. http://dx.doi.org/10.1038/s41597-022-01898-y

work page doi:10.1038/s41597-022-01898-y 2023
[57]

Ridge regression: Biased estimation for nonorthogonal problems

Arthur E Hoerl and Robert W Kennard. Ridge regression: Biased estimation for nonorthogonal problems. Technometrics, 12 0 (1): 0 55--67, 1970

work page 1970
[58]

An efficient P300 -based brain-computer interface for disabled subjects

Ulrich Hoffmann, Jean-Marc Vesin, Touradj Ebrahimi, and Karin Diserens. An efficient P300 -based brain-computer interface for disabled subjects. Journal of Neuroscience Methods, 167 0 (1): 0 115--125, January 2008. ISSN 0165-0270. doi:10.1016/j.jneumeth.2007.03.005. http://dx.doi.org/10.1016/j.jneumeth.2007.03.005

work page doi:10.1016/j.jneumeth.2007.03.005 2008
[59]

ZuCo, a simultaneous EEG and eye-tracking resource for natural sentence reading

Nora Hollenstein, Jonathan Rotsztejn, Marius Troendle, Andreas Pedroni, Ce Zhang, and Nicolas Langer. ZuCo, a simultaneous EEG and eye-tracking resource for natural sentence reading . Scientific Data, 5 0 (1), December 2018. ISSN 2052-4463

work page 2018
[60]

LoRA : Low-rank adaptation of large language models

Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Liang Wang, Weizhu Chen, et al. LoRA : Low-rank adaptation of large language models. Iclr, 1 0 (2): 0 3, 2022

work page 2022
[61]

A novel 9-class auditory ERP paradigm driving a predictive text entry system

Johannes Höhne. A novel 9-class auditory ERP paradigm driving a predictive text entry system. Frontiers in Neuroscience, 5, 2011. ISSN 1662-453X. doi:10.3389/fnins.2011.00099. http://dx.doi.org/10.3389/fnins.2011.00099

work page doi:10.3389/fnins.2011.00099 2011
[62]

Learning From Label Proportions In Brain-Computer Interfaces

David Hübner. EEG data for: "Learning From Label Proportions In Brain-Computer Interfaces" , 2016. https://zenodo.org/record/192684

work page 2016
[63]

Learning from label proportions in brain-computer interfaces: Online unsupervised learning with guarantees

David Hübner, Thibault Verhoeven, Konstantin Schmid, Klaus-Robert Müller, Michael Tangermann, and Pieter-Jan Kindermans. Learning from label proportions in brain-computer interfaces: Online unsupervised learning with guarantees. PLOS ONE, 12 0 (4): 0 e0175856, April 2017. ISSN 1932-6203. doi:10.1371/journal.pone.0175856. http://dx.doi.org/10.1371/journal....

work page doi:10.1371/journal.pone.0175856 2017
[64]

Ping-Keng Jao, Ricardo Chavarriaga, and José del R. Millán. EEG-Based online regulation of difficulty in simulated flying. IEEE Transactions on Affective Computing, 14 0 (1): 0 394--405, January 2023. ISSN 2371-9850. doi:10.1109/taffc.2021.3059688. http://dx.doi.org/10.1109/TAFFC.2021.3059688

work page doi:10.1109/taffc.2021.3059688 2023
[65]

MOABB : trustworthy algorithm benchmarking for bcis

Vinay Jayaram and Alexandre Barachant. MOABB : trustworthy algorithm benchmarking for bcis. Journal of neural engineering, 15 0 (6): 0 066011, 2018

work page 2018
[66]

Large brain model for learning generic representations with tremendous EEG data in BCI

Wei-Bang Jiang, Li-Ming Zhao, and Bao-Liang Lu. Large brain model for learning generic representations with tremendous EEG data in BCI . In The Twelfth International Conference on Learning Representations (ICLR), 2024

work page 2024
[67]

Kalunga, Sylvain Chevallier, Quentin Barthélemy, Karim Djouani, Eric Monacelli, and Yskandar Hamam

Emmanuel K. Kalunga, Sylvain Chevallier, Quentin Barthélemy, Karim Djouani, Eric Monacelli, and Yskandar Hamam. Online SSVEP-based BCI using Riemannian geometry. Neurocomputing, 191: 0 55--68, May 2016. ISSN 0925-2312. doi:10.1016/j.neucom.2016.01.007. http://dx.doi.org/10.1016/j.neucom.2016.01.007

work page doi:10.1016/j.neucom.2016.01.007 2016
[68]

Kappenman, Jaclyn L

Emily S. Kappenman, Jaclyn L. Farrens, Wendy Zhang, Andrew X. Stewart, and Steven J. Luck. ERP CORE : An open resource for human event-related potential research. NeuroImage, 225: 0 117465, January 2021. ISSN 1053-8119. doi:10.1016/j.neuroimage.2020.117465. http://dx.doi.org/10.1016/j.neuroimage.2020.117465

work page doi:10.1016/j.neuroimage.2020.117465 2021
[69]

Few-shot algorithms for consistent neural decoding ( FALCON ) benchmark

Brianna M Karpowicz, Joel Ye, Chaofei Fan, Pablo Tostado-Marcos, Fabio Rizzoglio, Clay Washington, Thiago Scodeler, Diogo de Lucena, Samuel R Nason-Tomaszewski, Matthew J Mender, et al. Few-shot algorithms for consistent neural decoding ( FALCON ) benchmark. Advances in Neural Information Processing Systems, 37: 0 76578--76615, 2024

work page 2024
[70]

EEG-Bench: A Benchmark for EEG Foundation Models in Clinical Applications

Ard Kastrati, Josua B \"u rki, Jonas Lauer, Cheng Xuan, Raffaele Iaquinto, and Roger Wattenhofer. EEG-Bench: A Benchmark for EEG Foundation Models in Clinical Applications . In NeurIPS 2025 Workshop on Foundation Models for the Brain and Body, 2025

work page 2025
[71]

Kemp, A.H

B. Kemp, A.H. Zwinderman, B. Tuk, H.A.C. Kamphuisen, and J.J.L. Oberye. Analysis of a sleep-dependent neuronal feedback loop: the slow-wave microcontinuity of the EEG . IEEE Transactions on Biomedical Engineering, 47 0 (9): 0 1185--1194, 2000

work page 2000
[72]

SwiFT: Swin 4D fMRI Transformer

Peter Kim, Junbeom Kwon, Sunghwan Joo, Sangyoon Bae, Donggyu Lee, Yoonho Jung, Shinjae Yoo, Jiook Cha, and Taesup Moon. SwiFT: Swin 4D fMRI Transformer . In A. Oh, T. Naumann, A. Globerson, K. Saenko, M. Hardt, and S. Levine, editors, Advances in Neural Information Processing Systems, volume 36, pages 42015--42037. Curran Associates, Inc., 2023

work page 2023
[73]

Jean-R \'e mi King, Teon L. Brooks, Katie Begany, Lucy Zhang, Josephine Raugel, Jarod L \'e vy, Sophia Houhamdi, Julien Gadonneix, Linnea Evanson, Corentin Bel, St \'e phane d'Ascoli, Marl \`e ne Careil, Yohann Benchetrit, Hubert Banville, and J \'e r \'e my Rapin. Neuralset: A high-performing python package for neuro-ai. 2026

work page 2026
[74]

Replication data for: Four-class ASME BCI : investigation of the feasibility and comparison of two strategies for multiclassing, 2024

Simon Kojima. Replication data for: Four-class ASME BCI : investigation of the feasibility and comparison of two strategies for multiclassing, 2024. https://dataverse.harvard.edu/citation?persistentId=doi:10.7910/DVN/1UJDV6

work page doi:10.7910/dvn/1ujdv6 2024
[75]

Replication data for: An auditory brain-computer interface based on selective attention to multiple tone streams, 2024

Simon Kojima and Shin'ichiro Kanoh. Replication data for: An auditory brain-computer interface based on selective attention to multiple tone streams, 2024. https://dataverse.harvard.edu/citation?persistentId=doi:10.7910/DVN/MQOVEY

work page doi:10.7910/dvn/mqovey 2024
[76]

Brain invaders calibration-less P300 -based BCI with modulation of flash duration dataset (bi2015a), 2019 a

Louis Korczowski, Martine Cederhout, Anton Andreev, Grégoire Cattan, Pedro Luis Coelho Rodrigues, Violette Gautheret, and Marco Congedo. Brain invaders calibration-less P300 -based BCI with modulation of flash duration dataset (bi2015a), 2019 a . https://zenodo.org/record/3266930

work page arXiv 2019
[77]

Brain invaders calibration-less P300 -based BCI using dry EEG electrodes dataset (bi2014a), 2019 b

Louis Korczowski, Ekaterina Ostaschenko, Anton Andreev, Grégoire Cattan, Pedro Luis Coelho Rodrigues, Violette Gautheret, and Marco Congedo. Brain invaders calibration-less P300 -based BCI using dry EEG electrodes dataset (bi2014a), 2019 b . https://zenodo.org/record/3266223

work page arXiv 2019
[78]

Brain invaders solo versus collaboration: Multi-User P300 -based Brain-Computer interface dataset (bi2014b), 2019 c

Louis Korczowski, Ekaterina Ostaschenko, Anton Andreev, Grégoire Cattan, Pedro Luis Coelho Rodrigues, Violette Gautheret, and Marco Congedo. Brain invaders solo versus collaboration: Multi-User P300 -based Brain-Computer interface dataset (bi2014b), 2019 c . https://zenodo.org/record/3267301

work page arXiv 2019
[79]

BENDR : Using transformers and a contrastive self-supervised learning task to learn from massive amounts of EEG data

Demetres Kostas, Stephane Aroca-Ouellette, and Frank Rudzicz. BENDR : Using transformers and a contrastive self-supervised learning task to learn from massive amounts of EEG data. Frontiers in Human Neuroscience, 15: 0 653659, 2021

work page 2021
[80]

EEG and EMG dataset for the detection of errors introduced by an active orthosis device

Niklas Kueper, Kartik Chari, Judith Bütefür, Julia Habenicht, Tobias Rossol, Su Kyoung Kim, Marc Tabie, Frank Kirchner, and Elsa Andrea Kirchner. EEG and EMG dataset for the detection of errors introduced by an active orthosis device. Frontiers in Human Neuroscience, 18, January 2024. ISSN 1662-5161. doi:10.3389/fnhum.2024.1304311

work page doi:10.3389/fnhum.2024.1304311 2024
[81]

Lawhern, Amelia J

Vernon J. Lawhern, Amelia J. Solon, Nicholas R. Waytowich, Stephen M. Gordon, Chou P. Hung, and Brent J. Lance. EEGNet : a compact convolutional neural network for EEG -based brain--computer interfaces. Journal of Neural Engineering, 15 0 (5): 0 056013, 2018

work page 2018
[82]

EEG dataset and OpenBMI toolbox for three BCI paradigms: an investigation into BCI illiteracy

Min-Ho Lee, O-Yeon Kwon, Yong-Jeong Kim, Hong-Kyung Kim, Young-Eun Lee, John Williamson, Siamac Fazli, and Seong-Whan Lee. EEG dataset and OpenBMI toolbox for three BCI paradigms: an investigation into BCI illiteracy. GigaScience, 8 0 (5), January 2019. ISSN 2047-217X. doi:10.1093/gigascience/giz002. http://dx.doi.org/10.1093/gigascience/giz002

work page doi:10.1093/gigascience/giz002 2019
[83]

Brain-Computer communication: Motivation, aim, and impact of exploring a virtual apartment

Robert Leeb, Felix Lee, Claudia Keinrath, Reinhold Scherer, Horst Bischof, and Gert Pfurtscheller. Brain-Computer communication: Motivation, aim, and impact of exploring a virtual apartment. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 15 0 (4): 0 473--482, December 2007. ISSN 1558-0210. doi:10.1109/tnsre.2007.906956. http://dx.doi....

work page doi:10.1109/tnsre.2007.906956 2007
[84]

The power of scale for parameter-efficient prompt tuning

Brian Lester, Rami Al-Rfou, and Noah Constant. The power of scale for parameter-efficient prompt tuning. In Proceedings of the 2021 conference on empirical methods in natural language processing, pages 3045--3059, 2021

work page 2021
[85]

Brain-to-text decoding: A non-invasive approach via typing

Jarod L \'e vy, Mingfang Zhang, Svetlana Pinet, J \'e r \'e my Rapin, Hubert Banville, St \'e phane d'Ascoli, and Jean-R \'e mi King. Brain-to-text decoding: A non-invasive approach via typing. arXiv preprint arXiv:2502.17480, 2025

work page arXiv 2025
[86]

An EEG motor imagery dataset for brain computer interface in acute stroke patients

Haijie Liu, Penghu Wei, Haochong Wang, Xiaodong Lv, Wei Duan, Meijie Li, Yan Zhao, Qingmei Wang, Xinyuan Chen, Gaige Shi, Bo Han, and Junwei Hao. An EEG motor imagery dataset for brain computer interface in acute stroke patients. Scientific Data, 11 0 (1), January 2024 a . ISSN 2052-4463. doi:10.1038/s41597-023-02787-8. http://dx.doi.org/10.1038/s41597-02...

work page doi:10.1038/s41597-023-02787-8 2024
[87]

EEG2video: Towards decoding dynamic visual perception from EEG signals

Xuan-Hao Liu, Yan-Kai Liu, Yansen Wang, Kan Ren, Hanwen Shi, Zilong Wang, Dongsheng Li, Bao-Liang Lu, and Wei-Long Zheng. EEG2video: Towards decoding dynamic visual perception from EEG signals . Advances in Neural Information Processing Systems, 37: 0 72245--72273, 2024 b

work page 2024

Showing first 80 references.