pith. sign in

arxiv: 2509.21671 · v2 · submitted 2025-09-25 · 💻 cs.LG · q-bio.NC

Neuroprobe: Evaluating Intracranial Brain Responses to Naturalistic Stimuli

Andrii Zahorodnii , Christopher Wang , Geeling Chau , Bennett Stankovits , Charikleia Moraitaki , Eli Gross , Alexander Brady , Andrei Barbu
show 2 more authors
Boris Katz Ila R Fiete
This is my paper

Pith reviewed 2026-05-18 13:32 UTC · model grok-4.3

classification 💻 cs.LG q-bio.NC
keywords intracranial EEGlanguage decodingnaturalistic stimulibrain-computer interfacesmulti-modal processingtemporal dynamicsspatial mapping
0
0 comments X

The pith

Neuroprobe introduces decoding tasks to map when and where the brain computes language features from intracranial recordings.

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

The paper presents Neuroprobe as a set of tasks that measure how well auditory and linguistic features can be decoded from high-resolution brain signals recorded while people watch movies. This setup lets researchers track the timing and brain locations of different processing stages without relying on artificial lab tasks. A reader would care because accurate decoding could reveal the sequence from basic sound properties to grammar and meaning, and it supplies a shared testbed for testing new models on real neural data. The work shows information moving from temporal areas to frontal regions and the gradual shift toward more abstract features.

Core claim

Neuroprobe is a suite of decoding tasks built on intracranial EEG recordings from subjects engaged in naturalistic movie viewing that allows systematic measurement of when and where each aspect of multi-modal language processing occurs by assessing feature decodability across time and electrode sites, while also serving as a benchmark for comparing model architectures.

What carries the argument

Neuroprobe, a collection of decoding tasks that quantify the decodability of auditory features such as pitch and volume and linguistic features such as part of speech from intracranial EEG signals across time and all electrode locations.

If this is right

  • Information can be shown flowing from language and audio sites in the superior temporal gyrus to locations in the prefrontal cortex.
  • Processing advances over time from simple auditory properties to more abstract linguistic properties in a data-driven way.
  • Different model architectures and training methods for neural data can be compared on the same standardized tasks.
  • Neuroscience questions about the spatial and temporal organization of language computations can be addressed directly from the labeled recordings.

Where Pith is reading between the lines

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

  • The same decoding approach could be extended to test whether the observed flow patterns hold during other everyday activities besides movie watching.
  • Reliable decoding of these features might identify candidate sites for future brain-computer interface applications that target language.
  • Results from this intracranial data could be compared with non-invasive recordings to check which processing stages are visible at larger scales.

Load-bearing premise

Brain responses collected during movie viewing represent general multi-modal language processing rather than being shaped mainly by movie-specific content or labeling inaccuracies.

What would settle it

Decoding accuracy for complex language features such as part of speech remains at chance level across subjects, time windows, and electrode locations even after controlling for basic auditory confounds.

Figures

Figures reproduced from arXiv: 2509.21671 by Alexander Brady, Andrei Barbu, Andrii Zahorodnii, Bennett Stankovits, Boris Katz, Charikleia Moraitaki, Christopher Wang, Eli Gross, Geeling Chau, Ila R Fiete.

Figure 1
Figure 1. Figure 1: Overview of Neuroprobe’s goals. Neuroprobe consists of classification tasks derived from human intracranial recordings aligned with annotated stimuli. It serves two critical roles: first, by performing a decoding analysis for each task, we can localize various aspects of multimodal language processing in the brain and discover their time evolution. Second, Neuroprobe is a rigorous, standardized benchmark f… view at source ↗
Figure 2
Figure 2. Figure 2: From raw data to decoding tasks. As part of the BrainTreebank dataset, 26 movies (left) are watched by 10 patients with stereoelectroencephalography intracranial electrodes implanted in various brain regions (middle), and the local field potential from the implanted electrodes is recorded (right). Neuroprobe turns this dataset into a standardized evaluation benchmark by segmenting the aligned data into var… view at source ↗
Figure 3
Figure 3. Figure 3: Neuroprobe allows for evaluating decoding within and across recording sessions and subjects. We perform analyses on three different types of splits (top row). In the within-session split, we train on data from one subject and one movie segment, and evaluate on the same subject, but another segment of the same movie. Performance is measured via cross-validation. In the cross-session split, we train and eval… view at source ↗
Figure 4
Figure 4. Figure 4: Neuroprobe enables the visualization of how multimodal stimuli are processed throughout the brain. This figure shows performance of linear decoders trained separately for every electrode’s data on the cross-session split, averaged across all recording sessions of every subject. Color denotes AUROC on a logarithmic scale to show trends for tasks that have lower decodability. Sentence Onset is decodable thro… view at source ↗
Figure 5
Figure 5. Figure 5: Tracking multimodal sensory processing in the brain across time. Here, we show the mean performance of the most decodable 100 electrodes per each task across time (top), where t = 0 corresponds to word onset. A linear model is fit on spectrograms of 250ms-long sliding windows of activity. Shaded regions denote s.e.m. across electrodes. We extract the peak of each decoding curve to obtain an approximate tim… view at source ↗
Figure 6
Figure 6. Figure 6: Time evolution of speech onset decodability across brain areas. The ‘sentence onset’ task is most decodable in the superior temporal gyrus at the first word’s onset (t = 0). Note that the decoding performance is above chance even before the speech onset, highlighting the predictive nature of sensory processing in the brain. As time progresses, speech becomes more decodable in the frontal areas of the brain… view at source ↗
Figure 7
Figure 7. Figure 7: Performance of baseline models on the 15 tasks of Neuroprobe (cross-session). The performance of four models is displayed: (1) logistic regression either from raw voltage signal of all electrodes to the labels, or (2) from the spectrogram of the signal to the labels, including laplacian re-referencing (3), as well as (4) BrainBERT (Wang et al., 2023) and (5) PopulationTransformer (Chau et al., 2024). For a… view at source ↗
read the original abstract

High-resolution neural datasets enable foundation models for the next generation of brain-computer interfaces and neurological treatments. The community requires rigorous benchmarks to discriminate between competing modeling approaches, yet no standardized evaluation frameworks exist for intracranial EEG (iEEG) recordings. To address this gap, we present Neuroprobe: a suite of decoding tasks for studying multi-modal language processing in the brain. Unlike scalp EEG, intracranial EEG requires invasive surgery to implant electrodes that record neural activity directly from the brain with minimal signal distortion. Neuroprobe is built on the BrainTreebank dataset, which consists of over 40 hours of iEEG recordings from 10 human subjects performing a naturalistic movie viewing task. Neuroprobe serves two critical functions. First, it is a source from which neuroscience insights can be drawn. The high temporal and spatial resolution of the labeled iEEG allows researchers to systematically determine when and where computations for each aspect of language processing occur in the brain by measuring the decodability of each feature across time and all electrode locations. Using Neuroprobe, we visualize how information flows from key language and audio processing sites in the superior temporal gyrus to sites in the prefrontal cortex. We also demonstrate the time evolution of processing from simple auditory features (e.g., pitch and volume) to more complex language features (e.g., part of speech) in a purely data-driven manner. Second, as the field moves toward neural foundation models trained on large-scale datasets, Neuroprobe provides a rigorous framework for comparing competing architectures and training protocols. We make the code for Neuroprobe openly available, aiming to enable rapid progress in the field of iEEG foundation models. Public leaderboard: https://neuroprobe.dev/

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.

Referee Report

2 major / 2 minor

Summary. The manuscript presents Neuroprobe, a suite of decoding tasks for intracranial EEG (iEEG) recordings from the BrainTreebank dataset (over 40 hours from 10 subjects during naturalistic movie viewing). It functions as both a resource for neuroscience insights—via decodability analyses to determine when and where multi-modal language computations occur—and a benchmark framework for comparing neural foundation models and training protocols. The authors report visualizations of information flow from superior temporal gyrus sites to prefrontal cortex and a data-driven time evolution from simple auditory features (pitch, volume) to complex linguistic features (part of speech), with code and a public leaderboard made available.

Significance. If the central demonstrations hold after addressing potential confounds, Neuroprobe would provide a valuable open benchmark and high-resolution dataset for advancing iEEG-based brain-computer interfaces and foundation models. The public leaderboard and open code release are strengths that could accelerate community progress in model evaluation. The data-driven approach to mapping neural processing hierarchies in naturalistic settings offers potential for novel insights into brain computations, though its impact depends on robustness to stimulus correlations.

major comments (2)
  1. [Abstract] Abstract: The claim to demonstrate 'the time evolution of processing from simple auditory features (e.g., pitch and volume) to more complex language features (e.g., part of speech) in a purely data-driven manner' is load-bearing for the neuroscience contribution but lacks support against stimulus confounds. In naturalistic audiovisual movies, low-level acoustic features statistically co-vary with linguistic features (e.g., pitch contours with syntactic boundaries or lexical stress), so later decodability of complex features could reflect leakage of simpler variance rather than a genuine neural hierarchy. The manuscript should report stimulus-feature correlation matrices and apply residualization or partial-correlation controls to validate the timeline.
  2. [Abstract] Abstract: The described visualizations of information flow and time evolution lack accompanying quantitative metrics, error bars, statistical tests, or full methods details (e.g., decoding procedures, feature extraction, cross-validation), as noted in the soundness assessment. This makes it difficult to evaluate the robustness of the reported findings.
minor comments (2)
  1. The manuscript would benefit from a dedicated methods section or supplementary table explicitly listing all decoding features, their extraction pipelines, and any preprocessing to improve reproducibility.
  2. Verify and prominently display all links to code, data, and the leaderboard (https://neuroprobe.dev/) to ensure accessibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed feedback. We have carefully reviewed the concerns regarding stimulus confounds and the presentation of quantitative details. Below we respond point by point and indicate the revisions we will make to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim to demonstrate 'the time evolution of processing from simple auditory features (e.g., pitch and volume) to more complex language features (e.g., part of speech) in a purely data-driven manner' is load-bearing for the neuroscience contribution but lacks support against stimulus confounds. In naturalistic audiovisual movies, low-level acoustic features statistically co-vary with linguistic features (e.g., pitch contours with syntactic boundaries or lexical stress), so later decodability of complex features could reflect leakage of simpler variance rather than a genuine neural hierarchy. The manuscript should report stimulus-feature correlation matrices and apply residualization or partial-correlation controls to validate the timeline.

    Authors: We agree that potential correlations between low-level acoustic and higher-level linguistic features in naturalistic stimuli represent an important interpretational concern. Our decoding analyses measure the temporal profile of each feature's decodability independently, which already provides a data-driven characterization of when information becomes available in the neural signals. To further isolate neural contributions from stimulus statistics, we will add stimulus-feature correlation matrices and perform residualization (or partial-correlation) controls in the revised manuscript. These additions will allow us to test whether the reported temporal progression persists after removing shared variance with simpler features. revision: yes

  2. Referee: [Abstract] Abstract: The described visualizations of information flow and time evolution lack accompanying quantitative metrics, error bars, statistical tests, or full methods details (e.g., decoding procedures, feature extraction, cross-validation), as noted in the soundness assessment. This makes it difficult to evaluate the robustness of the reported findings.

    Authors: The full manuscript contains a dedicated Methods section that details the decoding procedures, feature extraction pipelines, cross-validation scheme, and statistical testing approach. To improve clarity and address the referee's concern directly, we will revise the relevant results figures and accompanying text to include quantitative metrics, error bars, and statistical significance indicators for the information-flow and time-evolution visualizations. We will also add concise methodological summaries to the figure captions and ensure the abstract references the full methods. revision: yes

Circularity Check

0 steps flagged

Neuroprobe benchmark is self-contained data release with no derivation chain

full rationale

The manuscript defines a suite of decoding tasks on the BrainTreebank iEEG dataset and reports empirical visualizations of information flow and feature decodability timelines. These results are obtained directly from applying standard decoding methods to the provided stimulus features and neural recordings. No equations, fitted parameters, or self-citations are invoked to derive or justify the core claims; the work contains no mathematical derivation that could reduce to its inputs by construction. The central contributions are task definition and data-driven observations, which remain externally falsifiable against the released dataset and code.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper relies on the existing BrainTreebank dataset and standard decoding methods without introducing new free parameters, axioms beyond domain assumptions, or invented entities.

axioms (1)
  • domain assumption The BrainTreebank iEEG recordings and stimulus labels are sufficiently accurate and representative for multi-modal language processing studies.
    Invoked when using the dataset to draw neuroscience insights and build the benchmark.

pith-pipeline@v0.9.0 · 5870 in / 1194 out tokens · 41878 ms · 2026-05-18T13:32:07.034079+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

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

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

28 extracted references · 28 canonical work pages

  1. [1]

    Guillaume Alain and Yoshua Bengio

    Accessed: 2024-11-23. Guillaume Alain and Yoshua Bengio. Understanding intermediate layers using linear classifier probes. InInternational Conference on Learning Representations (ICLR),

    work page 2024

  2. [2]

    ISBN 979-10-95546-34-4

    European Language Resources Association. ISBN 979-10-95546-34-4. URLhttps://aclanthology.org/2020.lrec-1.15/. Jonathan R Brennan and John T Hale. Hierarchical structure guides rapid linguistic predictions during naturalistic listening.PloS one, 14(1):e0207741,

    work page 2020

  3. [3]

    Linnea Evanson, Christine Bulteau, Mathilde Chipaux, Georg Dorfmüller, Sarah Ferrand-Sorbets, Emmanuel Raffo, Sarah Rosenberg, Pierre Bourdillon, and Jean-Rémi King

    doi: 10.1016/j.neuroimage.2006.01.021. Linnea Evanson, Christine Bulteau, Mathilde Chipaux, Georg Dorfmüller, Sarah Ferrand-Sorbets, Emmanuel Raffo, Sarah Rosenberg, Pierre Bourdillon, and Jean-Rémi King. Emergence of language in the de- veloping brain.Meta AI Research,

    work page doi:10.1016/j.neuroimage.2006.01.021 2006

  4. [4]

    URL https://doi.org/10.1038/s41597-021-01102-7

    doi: 10.1038/s41597-021-01102-7. URL https://doi.org/10.1038/s41597-021-01102-7. Martin N Hebart, Oliver Contier, Lina Teichmann, Adam H Rockter, Charles Y Zheng, Alexis Kidder, Anna Corriveau, Maryam Vaziri-Pashkam, and Chris I Baker. Things-data, a multimodal collection of large-scale datasets for investigating object representations in human brain and ...

    work page doi:10.1038/s41597-021-01102-7

  5. [5]

    Large brain model for learning generic representations with tremendous eeg data in bci.arXiv preprint arXiv:2405.18765,

    10 Neuroprobe: Evaluating Intracranial Brain Responses to Naturalistic Stimuli Wei-Bang Jiang, Li-Ming Zhao, and Bao-Liang Lu. Large brain model for learning generic representations with tremendous eeg data in bci.arXiv preprint arXiv:2405.18765,

    work page arXiv

  6. [6]

    Jean-Philippe Lachaux, Nikolai Axmacher, Florian Mormann, Eric Halgren, and Nathan E

    URLhttps://arxiv.org/abs/2507.11783. Jean-Philippe Lachaux, Nikolai Axmacher, Florian Mormann, Eric Halgren, and Nathan E. Crone. High-frequency neural activity and human cognition: Past, present and possible future of intracranial eeg research.Progress in Neurobiology, 98(3):279–301,

    work page arXiv

  7. [7]

    Trung Le and Eli Shlizerman

    doi: 10.1016/j.pneurobio.2012.06.008. Trung Le and Eli Shlizerman. STNDT: Modeling Neural Population Activity with a Spatiotemporal Transformer, June

    work page doi:10.1016/j.pneurobio.2012.06.008 2012

  8. [8]

    doi: 10.1038/s41597-023-02437-z

    doi: 10.1038/s41597-023-02437-z. URL https://doi.org/10.1038/ s41597-023-02437-z. G. Li, S. Jiang, S. E. Paraskevopoulou, M. Wang, Y . Xu, Z. Wu, L. Chen, D. Zhang, and G. Schalk. Optimal referencing for stereo-electroencephalographic (seeg) recordings.NeuroImage, 183:327–335, Dec

    work page doi:10.1038/s41597-023-02437-z

  9. [9]

    Epub 2018 Aug

    doi: 10.1016/j.neuroimage.2018.08.020. Epub 2018 Aug

    work page doi:10.1016/j.neuroimage.2018.08.020 2018

  10. [10]

    URL https://www.nature.com/articles/ s41597-022-01625-7

    doi: 10.1038/s41597-022-01625-7. URLhttps://doi.org/10.1038/s41597-022-01625-7. Jan-Matthis Lueckmann, Alexander Immer, Alex Bo-Yuan Chen, Peter H Li, Mariela D Petkova, Nirmala A Iyer, Luuk Willem Hesselink, Aparna Dev, Gudrun Ihrke, Woohyun Park, et al. Zapbench: A benchmark for whole-brain activity prediction in zebrafish.arXiv preprint arXiv:2503.02618,

    work page doi:10.1038/s41597-022-01625-7

  11. [11]

    2020, Results in Physics, 16, 102918, doi:10.1016/j

    ISSN 0896-6273. doi: https://doi.org/10.1016/j. neuron.2024.02.011. URL https://www.sciencedirect.com/science/article/pii/S0896627324001211. Samuel A. Nastase, Ariel Goldstein, and Uri Hasson. Keep it real: Rethinking the primacy of experimental control in cognitive neuroscience.NeuroImage, 222:117254,

    work page doi:10.1016/j 2024

  12. [12]

    narratives

    doi: 10.1016/j.neuroimage.2020.117254. URL https://doi.org/10.1016/j.neuroimage.2020.117254. Open access under CC license. Samuel A. Nastase, Yung-Fang Liu, Harrison Hillman, et al. The “narratives” fmri dataset for evaluating models of naturalistic language comprehension.Scientific Data, 8:250,

    work page doi:10.1016/j.neuroimage.2020.117254 2020

  13. [13]

    Narratives

    doi: 10.1038/s41597-021-01033-3. URL https://doi.org/10.1038/s41597-021-01033-3. Petr Nejedly, Vaclav Kremen, Vladimir Sladky, Jan Cimbalnik, Petr Klimes, Filip Plesinger, Filip Mivalt, V ojtech Travnicek, Ivo Viscor, Martin Pail, et al. Multicenter intracranial eeg dataset for classification of graphoelements and artifactual signals.Scientific data, 7(1):179,

    work page doi:10.1038/s41597-021-01033-3

  14. [14]

    2002 , month =

    ISBN 9780195050387. doi: 10.1093/acprof:oso/9780195050387.001.0001. Iyad Obeid and Joseph Picone. The temple university hospital eeg data corpus.Frontiers in neuroscience, 10:196,

    work page doi:10.1093/acprof:oso/9780195050387.001.0001

  15. [15]

    URLhttps://doi.org/10.1016/j.conb.2018.04.007

    doi: 10.1016/j.conb.2018.04.007. URLhttps://doi.org/10.1016/j.conb.2018.04.007. Copyright © 2018 Elsevier Ltd. All rights reserved. Josef Parvizi and Sabine Kastner. Promises and limitations of human intracranial electroencephalography.Nature Neuroscience, 21(4):474–483,

    work page doi:10.1016/j.conb.2018.04.007 2018

  16. [16]

    URL https://doi.org/10.1038/ s41593-018-0108-2

    doi: 10.1038/s41593-018-0108-2. URL https://doi.org/10.1038/ s41593-018-0108-2. Felix Pei, Joel Ye, David M. Zoltowski, Anqi Wu, Raeed H. Chowdhury, Hansem Sohn, Joseph E. O’Doherty, Krishna V . Shenoy, Matthew T. Kaufman, Mark Churchland, Mehrdad Jazayeri, Lee E. Miller, Jonathan Pillow, Il Memming Park, Eva L. Dyer, and Chethan Pandarinath. Neural laten...

    work page doi:10.1038/s41593-018-0108-2

  17. [17]

    11 Neuroprobe: Evaluating Intracranial Brain Responses to Naturalistic Stimuli Matthew G

    URLhttps://arxiv.org/abs/2109.04463. 11 Neuroprobe: Evaluating Intracranial Brain Responses to Naturalistic Stimuli Matthew G. Perich, Lee E. Miller, Mehdi Azabou, and Eva L. Dyer. Long-term recordings of motor and premotor cortical spiking activity during reaching in monkeys. Data set,

    work page arXiv

  18. [18]

    000688/0.250122.1735

    URL https://doi.org/10.48324/dandi. 000688/0.250122.1735. Steven M Peterson, Zoe Steine-Hanson, Nathan Davis, Rajesh PN Rao, and Bingni W Brunton. Generalized neural decoders for transfer learning across participants and recording modalities.Journal of Neural Engineering, 18(2): 026014,

    work page doi:10.48324/dandi

  19. [19]

    Matthias Schurz, Joaquim Radua, Markus Aichhorn, Fabio Richlan, and Josef Perner

    doi: 10.1371/journal.pbio.1000610. Matthias Schurz, Joaquim Radua, Markus Aichhorn, Fabio Richlan, and Josef Perner. Fractionating theory of mind: A meta-analysis of functional brain imaging studies.Neuroscience & Biobehavioral Reviews, 42:9–34,

    work page doi:10.1371/journal.pbio.1000610

  20. [20]

    arXiv preprint arXiv:2403.11207 (2024)

    Paul S Scotti, Mihir Tripathy, Cesar Kadir Torrico Villanueva, Reese Kneeland, Tong Chen, Ashutosh Narang, Charan Santhirasegaran, Jonathan Xu, Thomas Naselaris, Kenneth A Norman, et al. Mindeye2: Shared-subject models enable fmri-to-image with 1 hour of data.arXiv preprint arXiv:2403.11207,

    work page arXiv

  21. [21]

    arXiv preprint arXiv:1905.05950 , year =

    Ian Tenney, Dipanjan Das, and Ellie Pavlick. Bert rediscovers the classical nlp pipeline.arXiv preprint arXiv:1905.05950,

    work page arXiv 1905

  22. [22]

    doi: https://doi.org/10.1016/j.jneumeth.2021.109089

    ISSN 0165-0270. doi: https://doi.org/10.1016/j.jneumeth.2021.109089. URL https://www.sciencedirect. com/science/article/pii/S0165027021000248. Franck Vidal, Boris Burle, Laure Spieser, Laurence Carbonnell, Cédric Meckler, Laurence Casini, and Thierry Has- broucq. Linking eeg signals, brain functions and mental operations: Advantages of the laplacian trans...

    work page doi:10.1016/j.jneumeth.2021.109089 2021

  23. [23]

    doi: https://doi.org/10.1016/j.ijpsycho

    ISSN 0167-8760. doi: https://doi.org/10.1016/j.ijpsycho. 2015.04.022. URL https://www.sciencedirect.com/science/article/pii/S0167876015001737. On the benefits of using surface Laplacian (current source density) methodology in electrophysiology. Christopher Wang, Vighnesh Subramaniam, Adam Uri Yaari, Gabriel Kreiman, Boris Katz, Ignacio Cases, and Andrei B...

    work page doi:10.1016/j.ijpsycho 2015

  24. [24]

    Leila Wehbe, Brian Murphy, Partha Talukdar, Alona Fyshe, Aaditya Ramdas, and Tom Mitchell

    URL https://arxiv.org/ abs/2411.08343. Leila Wehbe, Brian Murphy, Partha Talukdar, Alona Fyshe, Aaditya Ramdas, and Tom Mitchell. Simultaneously uncovering the patterns of brain regions involved in different story reading subprocesses.PLOS ONE, 9(11):e112575, November

    work page arXiv

  25. [25]

    doi: 10.1371/journal.pone.0112575

    ISSN 1932-6203. doi: 10.1371/journal.pone.0112575. URL http://dx.plos.org/10.1371/ journal.pone.0112575. Francis R Willett, Erin M Kunz, Chaofei Fan, Donald T Avansino, Guy H Wilson, Eun Young Choi, Foram Kamdar, Matthew F Glasser, Leigh R Hochberg, Shaul Druckmann, et al. A high-performance speech neuroprosthesis.Nature, 620(7976):1031–1036,

    work page doi:10.1371/journal.pone.0112575 1932

  26. [26]

    Brainwave: A brain signal foundation model for clinical applications.arXiv preprint arXiv:2402.10251,

    Zhizhang Yuan, Fanqi Shen, Meng Li, Yuguo Yu, Chenhao Tan, and Yang Yang. Brainwave: A brain signal foundation model for clinical applications.arXiv preprint arXiv:2402.10251,

    work page arXiv

  27. [27]

    12 Neuroprobe: Evaluating Intracranial Brain Responses to Naturalistic Stimuli Daoze Zhang, Zhizhang Yuan, Yang Yang, Junru Chen, Jingjing Wang, and Yafeng Li

    doi: 10.1038/s41597-025-03994-7. 12 Neuroprobe: Evaluating Intracranial Brain Responses to Naturalistic Stimuli Daoze Zhang, Zhizhang Yuan, Yang Yang, Junru Chen, Jingjing Wang, and Yafeng Li. Brant: Foundation Model for Intracranial Neural Signal. InThirty-Seventh Conference on Neural Information Processing Systems, November

    work page doi:10.1038/s41597-025-03994-7

  28. [28]

    encoding the mask pattern into the representation

    doi: 10.1109/TAMD.2015.2431497. 13 Neuroprobe: Evaluating Intracranial Brain Responses to Naturalistic Stimuli A Splits Neuroprobe includes 3 different types of splits. Within-SessionIn this split, models are trained and tested within the same subject and the same movie session. To avoid temporal data leakage, we are using 2-fold cross-validation using no...

    work page doi:10.1109/tamd.2015.2431497 2015