Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

The paper proposes a five-way, output-modality taxonomy — native EEG decoding, EEG-text, EEG-vision, EEG-audio, and multimodal fusion — as the organizing frame for foundation models applied to EEG analysis.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

A survey that organizes EEG foundation-model research into five output-modality categories: native EEG, text, vision, audio, and multimodal fusion, with a claim to be the first such comprehensive taxonomy.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection A genuinely useful taxonomy of EEG foundation-model work, undermined by an internal scope contradiction that needs fixing before the 'cross-domain, non-EEG' claim can stand. the 3 major comments →

arxiv 2508.15716 v2 pith:M6AFEFJN submitted 2025-08-21 cs.HC cs.AI

Foundation Models for Cross-Domain EEG Analysis Application: A Survey

Hongqi Li , Yitong Chen , Yujuan Wang , Weihang Ni , Haodong Zhang This is my paper
classification cs.HC cs.AI
keywords EEG analysisfoundation modelscross-domain transfermodality-oriented taxonomyEEG-to-textEEG-to-visionEEG-to-audiomultimodal fusion
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

This paper tries to establish that the exploding literature on foundation models applied to electroencephalography (EEG) has a simple organizing principle: what the pretrained model produces. It proposes the first modality-oriented taxonomy, sorting work into five output-modality families: native EEG decoding, EEG-text, EEG-vision, EEG-audio, and multimodal fusion. The payoff of getting the classification right is practical: researchers gain a shared map for comparing models, the recurring roles a foundation model can play (feature extractor, alignment tool, generative backbone) become visible, and open problems such as interpretability, cross-subject generalization, and missing benchmarks can be attacked systematically. The scope is deliberately limited to models pre-trained on non-EEG data and applied without EEG fine-tuning, and the survey's usefulness depends on that boundary being held consistently.

Core claim

The paper's central claim is that the heterogeneous and rapidly growing body of work using pretrained foundation models for EEG analysis is best understood by the modality of the model's output, not by the EEG task alone. It sorts the field into five output-modality families: native unimodal EEG decoding; EEG-to-text alignment, generation, and domain-specific understanding; EEG-to-vision retrieval, reconstruction, and video/3D generation; EEG-to-audio decoding, generation, and reconstruction; and multimodal EEG fusion. Within each family, the paper argues, the foundation model plays one of a small set of recurring roles — feature extractor, cross-modal alignment bridge, or generative backbon

What carries the argument

The governing device is a function-driven, modality-oriented taxonomy: five output-modality families (native EEG decoding, EEG-text, EEG-vision, EEG-audio, and multimodal fusion) plus a three-role account of foundation models as feature extractor, cross-modal alignment tool, or generative backbone. The taxonomy does the organizing work: it converts a scattered literature into a grid in which each surveyed paper can be located by what it emits, and it makes the shared EEG-encoder / alignment / decoder pipeline visible across otherwise dissimilar papers.

Load-bearing premise

The central claim collapses if any surveyed model was actually trained on brain recordings (EEG) or fine-tuned on them, because the taxonomy is explicitly about non-EEG pretraining; the paper's own unimodal section contains examples of EEG-pretrained models, so this assumption is already under strain.

What would settle it

A reader could audit every entry in Tables II through VI against the scope rule; any model pre-trained on EEG data would falsify the claim that the survey covers strictly non-EEG-pretrained models. A second test: take a representative EEG-to-text pipeline and replace the pretrained language model with a randomly initialized version of the same architecture; if downstream decoding does not degrade, then cross-domain pretraining is not doing the work the taxonomy attributes to it.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Researchers can position any EEG foundation-model paper by asking what modality it outputs, making the field's design space explicit.
  • The recurring three-stage pipeline (self-supervised pretraining, cross-modal alignment, task-specific fine-tuning) becomes a template for designing new methods.
  • The survey's named open problems — cross-subject generalization, authenticity of generated outputs, and missing standardized benchmarks — become the concrete agenda for the next wave of work.
  • The inverse direction (generating EEG from text, image, or audio) is identified as a gap, so future work can extend the taxonomy from decoding to synthesis.

Where Pith is reading between the lines

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

  • The paper's stated exclusion of EEG-pretrained models is not consistently applied: the unimodal decoding section includes models trained on EEG itself (for example, a contrastive transformer, a Mamba-based encoder, and several large-transformers), so the actual scope is broader than the advertised cross-domain-only boundary.
  • If the taxonomy is right, a direct test suggests itself: compare same-architecture models with and without large-scale non-EEG pretraining on each of the five output families; the taxonomy predicts pretrained versions should win reliably in cross-modal families and less clearly in native EEG decoding.
  • The paper's call for EEG digital twins and X-to-EEG synthesis implies the taxonomy could be extended beyond output modalities to treat synthetic EEG as a new modality, something the paper does not spell out.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. This manuscript is a survey of foundation model applications in EEG analysis. It proposes a modality-oriented taxonomy with five categories: unimodal EEG decoding, EEG-to-text, EEG-to-vision, EEG-to-audio, and multimodal EEG fusion. The stated scope (Section I) is that only foundation models pretrained on large-scale non-EEG data and directly applied to EEG tasks are covered, and that models fine-tuned on EEG datasets are excluded. The paper describes representative works in each category, discusses architectural components such as EEG encoders and cross-modal alignment modules, and closes with challenges and future directions.

Significance. If the stated scope were consistently applied, the survey would fill a useful niche: it collects a large and recent body of work, organizes it into an intuitive five-category structure, and highlights important caveats such as the reliability concerns raised by Jo et al. [67] and the scaling-law observations of Banville et al. [82]. The taxonomy is easy to grasp and could serve as an entry point for researchers new to EEG foundation models. However, the central contribution is framed as a comprehensive map of cross-domain (non-EEG pretrained) transfer, and that framing is contradicted by the paper's own included corpus. The organizing value of the taxonomy survives, but the paper's main claim of an exclusively cross-domain survey does not.

major comments (3)
  1. [§I vs. §II; Table II] The scope statement in Section I is load-bearing: the survey claims to cover 'exclusively' models pretrained on non-EEG data and to exclude models fine-tuned on EEG datasets. Section II contradicts this by including BENDR [29], which is pretrained via contrastive self-supervised learning on massive EEG data; CBraMod [32], an EEG-pretrained foundation model; EEGM2 [39], a self-supervised Mamba model for long-sequence EEG; and 'Large transformers are better EEG learners' [44], which is pretrained on EEG. These are in-domain, EEG-pretrained models, not cross-domain non-EEG models. The contradiction is internal, not a matter of divergent interpretations. The authors must either broaden the scope to include EEG-pretrained foundation models and adjust the 'cross-domain' claims accordingly, or remove/relabel the offending entries and re-audit the entire corpus against the stated criterion.
  2. [§VII, training methodology paragraph] The discussion section explicitly describes the standard pipeline as 'self- or unsupervised pretraining ... from large-scale unlabeled EEG data' followed by 'task-specific fine-tuning,' and it lists adapters [44] and prompt tuning [34] as acceptable parameter-efficient adaptations. This directly acknowledges EEG-domain pretraining and EEG fine-tuning as part of the surveyed paradigm, contradicting the Section I exclusion. If the authors intend to distinguish parameter-efficient adaptation from full fine-tuning, that distinction is never made explicit, and in any case [44] and [34] are EEG-pretrained/fine-tuned systems. The boundary between 'cross-domain transfer' and 'EEG foundation models' needs to be redefined or the paper must be repositioned as covering both.
  3. [§I and overall methodology] The paper claims to be 'the first and latest comprehensive taxonomy' but provides no explicit methodology for literature selection: no search databases, no inclusion/exclusion criteria beyond the contradicted scope sentence, and no screening process. Given the rapid growth of the field, the 'comprehensive' claim is unverifiable and arguably overreaching. A short methodology paragraph or a table of inclusion decisions, especially explaining why some EEG-pretrained models are included while others are excluded, would address this. This is not purely cosmetic: the central contribution is the corpus/taxonomy, so the selection protocol is part of the evidence.
minor comments (5)
  1. [References [116] and [117]] The reference list contains two identical entries for 'CineBrain' (J. Gao et al., arXiv:2503.06940). In Section VI.C, 'Wu et al. [117]' is cited for a hypergraph-based language model, but [117] is the duplicate CineBrain entry. This citation is incorrect and must be fixed.
  2. [Throughout] Several typos and formatting issues: 'demonstrating superior performance' (Section I) should be 'demonstrated'; 'Reconstrution' in the Section IV.B heading; stray spaces in arXiv identifiers such as 'arXiv: 2411.15395' and 'arXiv: 2501.17489'; inconsistent capitalization of 'modality oriented' in the index terms.
  3. [Reference [57]] The BART paper is by Lewis et al., not 'M. Lewi.' Please correct the author name.
  4. [§I and abstract] The abstract and introduction promise a 'rigorous' analysis of 'theoretical foundations' and 'architectural innovations,' but the body is mostly descriptive. The only equation is the InfoNCE-style contrastive loss in Section III.A. Consider tempering the wording or adding a comparative analysis that goes beyond citing representative architectures.
  5. [Taxonomy definition] The five categories are not mutually exclusive: EEG-to-vision reconstruction can be considered a multimodal perception task (Section VI.A), and EEG-to-text generation is also a cross-modal task. The paper would benefit from explicit rules for assigning a paper to exactly one category, especially because the taxonomy is the main contribution.

Circularity Check

0 steps flagged

No significant circularity: the survey organizes external literature and does not reduce to its own inputs.

full rationale

This is a survey paper whose central output is a modality-based taxonomy of foundation-model applications to EEG. The taxonomy is an organizational scheme over externally published methods; it is not derived from the authors' own equations or fitted parameters, and it does not make quantitative predictions that could be forced by construction. The few self-citations ([1], [2], [5], [124]) appear as background motivation, prior-survey framing, and a peripheral forward-looking mention; none of them carries the load of the taxonomy's definitions or categories. The paper does contain an internal scope inconsistency: it states that it includes 'exclusively' models pre-trained on non-EEG data and excludes models fine-tuned on EEG datasets, yet Section II and Table II include EEG-pretrained models such as BENDR, CBraMod, EEGM2, and Large transformers, and Section VII describes fine-tuning as part of the standard pipeline. That is a correctness/scope-consistency concern, not a circularity: the included items do not make the taxonomy's organizing criterion equivalent to its output. Since no equation is shown to be equal to another by definition, and no fitted parameter is relabeled as a prediction, the appropriate circularity score is 0.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The survey has no fitted parameters or invented entities. Its dependencies are the literature corpus and the categorization assumptions listed.

axioms (2)
  • domain assumption The modality-based partition (native EEG, EEG-text, EEG-vision, EEG-audio, multimodal) is a valid and non-overlapping way to organize the literature.
    This is the paper's chosen organizing principle; it is not derived from data and is presented as a framing premise.
  • domain assumption All included works actually satisfy the stated scope: pretrained on non-EEG data and not fine-tuned on EEG datasets.
    Section I defines the scope, but Section II includes EEG-pretrained models (BENDR, CBraMod), so this premise is violated.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Foundation Models for Cross-Domain EEG Analysis Application: A Survey." pith.science (2026). https://pith.science/paper/M6AFEFJN

@misc{pith2026250815716,
  author       = {Pith},
  title        = {Pith review of: Foundation Models for Cross-Domain EEG Analysis Application: A Survey},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/M6AFEFJN}},
  note         = {Machine review of arXiv:2508.15716}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Electroencephalography (EEG) analysis stands at the forefront of neuroscience and artificial intelligence research, where foundation models are reshaping the traditional EEG analysis paradigm by leveraging their powerful representational capacity and cross-modal generalization. However, the rapid proliferation of these techniques has led to a fragmented research landscape, characterized by diverse model roles, inconsistent architectures, and a lack of systematic categorization. To bridge this gap, this study presents the first comprehensive modality-oriented taxonomy for foundation models in EEG analysis, systematically organizing research advances based on output modalities of the native EEG decoding, EEG-text, EEG-vision, EEG-audio, and broader multimodal frameworks. We rigorously analyze each category's research ideas, theoretical foundations, and architectural innovations, while highlighting open challenges such as model interpretability, cross-domain generalization, and real-world applicability in EEG-based systems. By unifying this dispersed field, our work not only provides a reference framework for future methodology development but accelerates the translation of EEG foundation models into scalable, interpretable, and online actionable solutions.

discussion (0)

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

Forward citations

Cited by 1 Pith paper

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

  1. EEG-FM-Audit: A Systematic Evaluation and Analysis Pipeline for EEG Foundation Models

    cs.LG 2026-05 unverdicted novelty 7.0

    EEG-FM-Audit is an evaluation pipeline showing that properly tuned supervised baselines can match or outperform EEG foundation models with far fewer parameters on public datasets.

Reference graph

Works this paper leans on

125 extracted references · 37 linked inside Pith · cited by 1 Pith paper

  1. [1]

    Dual-TSST: A dual-branch temporal- spectral-spatial Transformer model for EEG decoding,

    H. Li, H. Zhang and Y . Chen, “Dual-TSST: A dual-branch temporal- spectral-spatial Transformer model for EEG decoding,” J. Biomed. Heath Inform., doi: 10.1109/JBHI.2025.3577611

    arXiv 2025

  2. [2]

    Noninvasive EEG-based intelligent mobile robots: a systematic review,

    H. Li, X. Li and J. d. R. Mill ´an, “Noninvasive EEG-based intelligent mobile robots: a systematic review,” IEEE Trans. Autom. Sci. Eng., vol. 22, pp. 6291-6315, 2025

    2025

  3. [3]

    Neural decoding of EEG signals with machine learning: a systematic review,

    M. Saeidi et al. , “Neural decoding of EEG signals with machine learning: a systematic review,” Brain Sci. , vol. 11, no. 11, pp. 1524- 1569, 2021

    2021

  4. [4]

    Deep learning-based electroencephalography analysis: A systematic review,

    Yannick et al., “Deep learning-based electroencephalography analysis: A systematic review,” J. Neural Eng. , vol. 16, 2019, Art. no. 051001

    2019

  5. [5]

    Transformer-based EEG decoding: A survey,

    H. Zhang, and H. Li, “Transformer-based EEG decoding: A survey,” 2025, arXiv:2507.02320

    Pith/arXiv arXiv 2025

  6. [6]

    On the opportunities and risks of foundation models,

    R. Bommasani, et al. , “On the opportunities and risks of foundation models,” 2021, arXiv:2108.07258

    Pith/arXiv arXiv 2021

  7. [7]

    Improving language understanding by generative pre-training,

    A. Radford, and K. Narasimhan, “Improving language understanding by generative pre-training,” 2018

    2018

  8. [8]

    LLaMA: Open and efficient foundation language models

    H. Touvron et al. , “LLaMA: Open and efficient foundation language models.” 2023, arXiv:2302.13971

    Pith/arXiv arXiv 2023

  9. [9]

    An image is worth 16×16 words: Transformers for image recognition at scale,

    A. Dosovitskiy et al., “An image is worth 16×16 words: Transformers for image recognition at scale,” in Proc. Int. Conf. Learn. Represent. , 2021, pp. 1-21

    2021

  10. [10]

    Swin Transformer: Hierarchical vision transformer using shifted windows,

    Z. Liu et al., “Swin Transformer: Hierarchical vision transformer using shifted windows,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. , 2021, pp. 10012-10022

    2021

  11. [11]

    Wav2Vec: Un- supervised pre training for speech recognition,

    S. Schneider, A. Baevski, R. Collobert, and M. Auli, “Wav2Vec: Un- supervised pre training for speech recognition,” in Proc. Interspeech, 2019, pp. 3465-3469

    2019

  12. [12]

    Whisper: Robust speech recognition via large scale weak supervision,

    A. Radford et al., “Whisper: Robust speech recognition via large scale weak supervision,” arXiv:2212.04356, 2022

    Pith/arXiv arXiv 2022

  13. [13]

    Learning transferable visual models from natural language supervision,

    A. Radford et al. , “Learning transferable visual models from natural language supervision,” in Proc. Int. Conf. Mach. Learn. , 2021, pp. 8748–8763

    2021

  14. [14]

    Flamingo: a visual language model for few-shot learning,

    J.B. Alayrac et al. , “Flamingo: a visual language model for few-shot learning,” Adv. Neural Inf. Process. Syst. , vol. 35, 2022, pp. 23716- 23736

    2022

  15. [15]

    Gemini: Efficient multimodal pretraining for vision and language,

    N. Goyal et al. , “Gemini: Efficient multimodal pretraining for vision and language,” 2024, arXiv:2401.01234

    Pith/arXiv arXiv 2024

  16. [16]

    PaLM-E: An embodied multimodal language model,

    D. Driess et al., “PaLM-E: An embodied multimodal language model,” 2023, arXiv:2303.13971

    Pith/arXiv arXiv 2023

  17. [17]

    A survey on bridging EEG signals and generative AI: From image and text to beyond,

    S. Shukla et al. , “A survey on bridging EEG signals and generative AI: From image and text to beyond,” 2025, arXiv:2502.12048

    arXiv 2025

  18. [18]

    Decoding linguistic representations of human brain,

    Y . Wang, et al., “Decoding linguistic representations of human brain,” 2024, arXiv:2407.20622

    Pith/arXiv arXiv 2024

  19. [19]

    Unveiling thoughts: A review of ad- vancements in EEG brain signal decoding into text,

    S. A. Murad and N. Rahimi, “Unveiling thoughts: A review of ad- vancements in EEG brain signal decoding into text,”IEEE Trans. Cogn. Dev. Syst., vol. 17, no. 1, pp. 61-76, Feb. 2025

    2025

  20. [20]

    Decoding natural images from EEG for object recogni- tion,

    Y . Song et al., “Decoding natural images from EEG for object recogni- tion,” in Proc. Int. Conf. Learn. Represent. (ICLR), May 2024, pp. 1–6

    2024

  21. [21]

    Decoding speech perception from non-invasive brain recordings,

    A. D ´efossez et al. , “Decoding speech perception from non-invasive brain recordings,” Nat. Mach. Intell. , vol. 5, no. 10, pp. 1097-1107, Oct. 2023

    2023

  22. [22]

    Self-supervised Learning for Electroencephalog- raphy,

    M. H. Rafiei, et al., “Self-supervised Learning for Electroencephalog- raphy,” IEEE Trans. Neural Netw. Learn. Syst. , V ol. 35, no. 2, 2024, pp. 1457-1471

    2024

  23. [23]

    E2H: A two-stage non-invasive neural signal driven humanoid robotic whole-body control framework,

    Y . Duan et al. , “E2H: A two-stage non-invasive neural signal driven humanoid robotic whole-body control framework,” 2024, arXiv:2410.02141

    Pith/arXiv arXiv 2024

  24. [24]

    ChatBCI: A P300 speller BCI leveraging large language models for improved sentence composition in realistic scenarios,

    J. Hong et al., “ChatBCI: A P300 speller BCI leveraging large language models for improved sentence composition in realistic scenarios,” 2024, arXiv: 2411.15395

    arXiv 2024

  25. [25]

    Sequential best-arm identification with application to P300 Speller,

    X. Zhou et al., “Sequential best-arm identification with application to P300 Speller,” Transact. Mach. Learn. Res. , Aug. 2024

    2024

  26. [26]

    Neural spelling: A spell-based BCI system for language neural decoding,

    X. Jiang et al., “Neural spelling: A spell-based BCI system for language neural decoding,” 2025, arXiv:2501.17489

    arXiv 2025

  27. [27]

    Human-AI teaming using large language models: Boosting brain-computer interfacing (BCI) and brain research,

    M. Kapitnova and T. Ball, “Human-AI teaming using large language models: Boosting brain-computer interfacing (BCI) and brain research,” 2025, arXiv: 2501.01451

    Pith/arXiv arXiv 2025

  28. [28]

    Classification of non-invasive EEG signals during motor imagery tasks using a large language model,

    D. Lim, M. Cho and H. Kim, “Classification of non-invasive EEG signals during motor imagery tasks using a large language model,” in Proc. Int. Conf. Cyberworlds (CW) , Oct. 2024, pp. 378-379

    2024

  29. [29]

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

    D. Kostas, S. Aroca-Ouellette and F. Rudzicz, “BENDR: Using trans- formers and a contrastive self-supervised learning task to learn from massive amounts of EEG data,” Front. Hum. Neurosci. , vol. 15, pp. 253, Jun. 2021

    2021

  30. [30]

    From word embedding to reading embedding using large language model, EEG and eye-tracking,

    Y . Zhang et al. , “From word embedding to reading embedding using large language model, EEG and eye-tracking,” inProc. IEEE Eng. Med. Biol. Soc. (EMBC) , Jul. 2024, pp. 1-4

    2024

  31. [31]

    Integrating large language model, EEG, and eye- tracking for word-level neural state classification in reading com- prehension,

    Y . Zhang et al. , “Integrating large language model, EEG, and eye- tracking for word-level neural state classification in reading com- prehension,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 32, pp. 3465- 3475, Aug. 2024

    2024

  32. [32]

    CBraMod: A criss-cross brain foundation model for EEG decoding,

    J. Wang et al. , “CBraMod: A criss-cross brain foundation model for EEG decoding,” in Proc. Int. Conf. Learn. Represent. (ICLR) , Apr. 2025, pp. 62056-62092

    2025

  33. [33]

    NeuroChat: A neuroadaptive AI chatbot for cus- tomizing learning experiences,

    D. Baradari et al. , “NeuroChat: A neuroadaptive AI chatbot for cus- tomizing learning experiences,” 2025, arXiv:2503.07599

    arXiv 2025

  34. [34]

    EEG emotion copilot: Pruning LLMs for emotional EEG interpretation with assisted medical record generation,

    H. Chen et al. , “EEG emotion copilot: Pruning LLMs for emotional EEG interpretation with assisted medical record generation,” Nerual Netw., 107848, pp. 1-17, Jul. 2025

    2025

  35. [35]

    Exploring large-scale language models to evaluate EEG- based multimodal data for mental health,

    Y . Hu et al., “Exploring large-scale language models to evaluate EEG- based multimodal data for mental health,” in Proc. ACM Int. Jt. Conf. Pervasive Ubiquitous Comput. (UbiComp) , Oct. 2024, pp. 412-417

    2024

  36. [36]

    LLM-enhanced multi-teacher knowledge distillation for modality-incomplete emotion recognition in daily healthcare,

    Y . Zhang et al. , “LLM-enhanced multi-teacher knowledge distillation for modality-incomplete emotion recognition in daily healthcare,” IEEE J. Biomed. Heath Inform. , pp. 1–11, Sep. 2024

    2024

  37. [37]

    Advancing semi-supervised EEG emotion recognition through feature extraction with mixup and large language models,

    S. Yao et al. , “Advancing semi-supervised EEG emotion recognition through feature extraction with mixup and large language models,” in Proc. IEEE Int. Conf. Bioinform. Biomed. (BIBM) , Dec. 2024, pp. 2772-2779

    2024

  38. [38]

    Emotion analysis AI model for sensing architecture using EEG,

    S.-Y . Ji, M.-K. Kim and H.-J. Jun, “Emotion analysis AI model for sensing architecture using EEG,” Appl. Sci. , vol. 15, no.5, pp. 1-21, Mar. 2025

    2025

  39. [39]

    EEGM2: An efficient mamba-2-based self- supervised framework for long-sequence EEG modeling,

    J. Hong et al. , “EEGM2: An efficient mamba-2-based self- supervised framework for long-sequence EEG modeling,” 2025, arXiv:2502.17813

    Pith/arXiv arXiv 2025

  40. [40]

    BERT learns from electroencephalograms about Parkinson’s disease: Transformer-based models for aid diagnosis,

    A. Nogales et al. , “BERT learns from electroencephalograms about Parkinson’s disease: Transformer-based models for aid diagnosis,” IEEE Access, vol. 10, pp. 101672-101682, Aug. 2022

    2022

  41. [41]

    When neural implant meets multi- modal LLM: A dual-loop system for neuromodulation and naturalistic neuralbehav-ioral research,

    E. H. Wang and C. X. Wen, “When neural implant meets multi- modal LLM: A dual-loop system for neuromodulation and naturalistic neuralbehav-ioral research,” 2025, arXiv:2503.12334

    Pith/arXiv arXiv 2025

  42. [42]

    BrainWave: A brain signal foundation model for clinical applications,

    Z. Yuan et al. , “BrainWave: A brain signal foundation model for clinical applications,” 2024, arXiv:2402.10251

    Pith/arXiv arXiv 2024

  43. [43]

    Clinical grade prediction of therapeutic dosage for electroconvulsive therapy (ECT) based on patient’s pre-ictal EEG using fuzzy causal transformers,

    C. Kalantarpour et al., “Clinical grade prediction of therapeutic dosage for electroconvulsive therapy (ECT) based on patient’s pre-ictal EEG using fuzzy causal transformers,” in Proc. 3rd Int. Conf. Elect. Comput. Commun. Mechatronics Eng. (ICECCME) , Jul. 2023, pp. 1-6

    2023

  44. [44]

    Large transformers are better EEG learners,

    B. Wang et al. , “Large transformers are better EEG learners,” 2023, arXiv:2308.11654

    Pith/arXiv arXiv 2023

  45. [45]

    EEG-GPT: Exploring capabilities of large language models for EEG classification and interpretation,

    J. W. Kim, A. Alaa and D. Bernardo, “EEG-GPT: Exploring capabilities of large language models for EEG classification and interpretation,” 2024, arXiv:2401.18006

    Pith/arXiv arXiv 2024

  46. [46]

    NeuroLM: A universal multi-task foundation model for bridging the gap between language and EEG signals,

    W. B. Jiang et al., “NeuroLM: A universal multi-task foundation model for bridging the gap between language and EEG signals,” in Proc. Int. Conf. Learn. Represent. (ICLR) , Apr. 2025, pp. 46232-46253

    2025

  47. [47]

    A spatial-temporal transformer architecture using mul- ti-channel signals for sleep stage classification,

    H. Yao et al., “A spatial-temporal transformer architecture using mul- ti-channel signals for sleep stage classification,” IEEE Trans. Neural Syst. Rehabil. Eng. , vol. 31, pp. 3353-3362, Aug. 2023

    2023

  48. [48]

    Exploration of LLMs, EEG and behavioral data to measure and support attention and sleep,

    A. Sano et al. , “Exploration of LLMs, EEG and behavioral data to measure and support attention and sleep,” 2024, arXiv:2408.07822

    Pith/arXiv arXiv 2024

  49. [49]

    Are foundation models useful feature extractors for electroencephalography analysis?,

    ¨O. Turgut et al., “Are foundation models useful feature extractors for electroencephalography analysis?,” 2025, arXiv:2502.21086

    arXiv 2025

  50. [50]

    LEAD: Large foundation model for EEG-based alz- heimer’s disease detection,

    Y . Wang et al. , “LEAD: Large foundation model for EEG-based alz- heimer’s disease detection,” 2025, arXiv:2502.01678

    arXiv 2025

  51. [51]

    Can brain signals reveal inner alignment with human languages?,

    J. Qiu et al. , “Can brain signals reveal inner alignment with human languages?,” in Proc. Find. Assoc. Comput. Linguist (EMNLP) , Dec

  52. [52]

    Enhancing EEG-to-text decoding through transferable representations from pre-trained contrastive EEG-text masked auto- encoder,

    J. Wang et al., “Enhancing EEG-to-text decoding through transferable representations from pre-trained contrastive EEG-text masked auto- encoder,” in Proc. Annu. Meet. Assoc. Comput. Linguist. (ACL) , Aug. 2024, pp. 7278-7292

    2024

  53. [53]

    Aligning semantic in brain and language: A curricu- lum contrastive method for electroencephalography-to-text generation,

    X. Feng et al. , “Aligning semantic in brain and language: A curricu- lum contrastive method for electroencephalography-to-text generation,” IEEE Trans. Neural Syst. Rehabil. Eng. , vol.31, pp. 3874-3883, Sep. 2023

    2023

  54. [54]

    Towards linguistic neural representation learning and sentence retrieval from electroencephalogram recordings,

    J. Zhou et al. , “Towards linguistic neural representation learning and sentence retrieval from electroencephalogram recordings,” in Proc. Int. Workshop Brain-Comput. Interfaces (BCI) Multimed. Underst. (BCIMM), Oct. 2024, pp. 19-28

    2024

  55. [55]

    LLMs help alleviate the cross-subject variabil- ity in brain signal and language alignment,

    Y . Liu, H. Ye and S. Li, “LLMs help alleviate the cross-subject variabil- ity in brain signal and language alignment,” 2025, arXiv:2501.02621. JOURNAL OF LATEX CLASS FILES, VOL. 18, NO. 9, SEPTEMBER 2020 14

    Pith/arXiv arXiv 2025

  56. [56]

    BELT: Bootstrapped EEG-to-language training by natural language supervision,

    J. Zhou et al. , “BELT: Bootstrapped EEG-to-language training by natural language supervision,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 32, pp. 3278-3288, Aug. 2024

    2024

  57. [57]

    BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension,

    M. Lewi et al. , “BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension,” in Proc. Annu. Meet. Assoc. Comput Linguist. (ACL), Jul. 2020, pp. 7871- 7880

    2020

  58. [58]

    Open vocabulary electroencephalography-to-text decoding and zero-shot sentiment classification,

    Z. Wang, and H. Ji, “Open vocabulary electroencephalography-to-text decoding and zero-shot sentiment classification,” in Proc. AAAI Conf. Artif. Intell. (AAAI) , Jun. 2022, pp. 5350-5358

    2022

  59. [59]

    From thought to action: How a hierarchy of neural dynamics supports language production,

    J. L ´evy et al. , “From thought to action: How a hierarchy of neural dynamics supports language production,” 2025, arXiv:2502.07429

    Pith/arXiv arXiv 2025

  60. [60]

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

    J. L ´evy, et al. , “Brain-to-text decoding: A non-invasive approach via typing,” 2025, arXiv:2502.17480

    Pith/arXiv arXiv 2025

  61. [61]

    Deep representation learning for open vocabulary electroencephalography-to-text decoding,

    H. Amrani, D. Micucci and P. Napoletano, “Deep representation learning for open vocabulary electroencephalography-to-text decoding,” IEEE J. Biomed. Health Inform. , pp. 1-12, Jun. 2024

    2024

  62. [62]

    SEE: Semantically aligned EEG-to-text translation,

    Y . Tao et al., “SEE: Semantically aligned EEG-to-text translation,” in Proc. IEEE Int. Conf. Acoust. Speech Signal Process. (ICASSP) , Apr. 2025, pp. 1-5

    2025

  63. [63]

    EEG2TEXT: Open vocabulary EEG-to-text de- coding with EEG pre-Training and multi-view transformer,

    H. Liu et al. , “EEG2TEXT: Open vocabulary EEG-to-text de- coding with EEG pre-Training and multi-view transformer,” 2024, arXiv:2405.02165

    Pith/arXiv arXiv 2024

  64. [64]

    Dewave: Discrete EEG waves encoding for brain dynamics to text translation,

    Y . Duan et al. , “Dewave: Discrete EEG waves encoding for brain dynamics to text translation,” in Proc. Int. Conf. Neural Inf. Process. Syst. (NeurIPS), Dec. 2023, pp. 1-22

    2023

  65. [65]

    BELT-2: Bootstrapping EEG-to-language representa- tion alignment for multi-task brain decoding,

    J. Zhou, et al. , “BELT-2: Bootstrapping EEG-to-language representa- tion alignment for multi-task brain decoding,” 2024, arXiv:2409.00121

    Pith/arXiv arXiv 2024

  66. [66]

    EEG-CLIP: Learning EEG repre-sentations from natural language descriptions,

    T. C. Ndir and R. T. Schirrmeister, “EEG-CLIP: Learning EEG repre-sentations from natural language descriptions,” 2025, arXiv:2503.16531

    Pith/arXiv arXiv 2025

  67. [67]

    Are EEG-to-text models working?,

    H. Jo et al. , “Are EEG-to-text models working?,” 2024, arXiv:2405.06459

    Pith/arXiv arXiv 2024

  68. [68]

    Learning robust deep visual representations from EEG brain recordings,

    P. Singh et al. , “Learning robust deep visual representations from EEG brain recordings,” in Proc. IEEE Winter Conf. Appl. Comput. Vis. (WACV), Jan. 2024, pp. 7553-7562

    2024

  69. [69]

    Human-aligned image models improve visual decod- ing from the brain,

    N. Rajabi et al., “Human-aligned image models improve visual decod- ing from the brain,” 2025, arXiv: 2502.03081

    Pith/arXiv arXiv 2025

  70. [70]

    MB2C: Multimodal bidirectional cycle consistency for learning robust visual neural representations,

    Y . Wei et al., “MB2C: Multimodal bidirectional cycle consistency for learning robust visual neural representations,” in Proc. ACM Int. Conf. Multimed. (MM), Oct. 2024, pp. 8992-9000

    2024

  71. [71]

    Neural-MCRL: Neural multimodal contrastive representa- tion learning for EEG-based visual decoding,

    Y . Li et al., “Neural-MCRL: Neural multimodal contrastive representa- tion learning for EEG-based visual decoding,” 2024,arXiv:2412.17337

    Pith/arXiv arXiv 2024

  72. [72]

    Visual neural decoding via improved visual-EEG se- mantic consistency,

    H. Chen et al., “Visual neural decoding via improved visual-EEG se- mantic consistency,” 2024, arXiv:2408.06788

    arXiv 2024

  73. [73]

    RealMind: Advancing visual decoding and language in- teraction via EEG signals,

    D Li et al. , “RealMind: Advancing visual decoding and language in- teraction via EEG signals,” 2024, arXiv:2410.23754

    Pith/arXiv arXiv 2024

  74. [74]

    Visual decoding and reconstruction via EEG embeddings with guided diffusion,

    D. Li et.al., “Visual decoding and reconstruction via EEG embeddings with guided diffusion,” in Proc. 38th Int. Conf. Neural Inf. Process. Syst. (NeurIPS), Dec. 2024, pp. 1-35

    2024

  75. [75]

    DreamDiffusion: Generating high-quality images from brain EEG signals,

    Y . Bai et al. , “DreamDiffusion: Generating high-quality images from brain EEG signals,” 2023, arXiv:2306.16934

    Pith/arXiv arXiv 2023

  76. [76]

    Guess what I think: Streamlined EEG-to-image gen- eration with latent diffusion models,

    E. Lopez et al., “Guess what I think: Streamlined EEG-to-image gen- eration with latent diffusion models,” in Proc. IEEE Int. Conf. Acoust. Speech Signal Process. (ICASSP) , Apr. 2025, pp. 13-18

    2025

  77. [77]

    EEG decoding and visual reconstruction via 3D geometric with nonstationarity modelling,

    X. Xiao et al. , “EEG decoding and visual reconstruction via 3D geometric with nonstationarity modelling,” in Proc. IEEE Int. Conf. Acoust. Speech Signal Process. (ICASSP) , Apr. 2025, pp. 1-5

    2025

  78. [78]

    Seeing through the brain: Image reconstruction of visual perception from human brain signals,

    Y .-T. Lan et al. , “Seeing through the brain: Image reconstruction of visual perception from human brain signals,” 2023, arXiv:2308.02510

    Pith/arXiv arXiv 2023

  79. [79]

    BrainDecoder: Style-based visual decoding of EEG signals,

    M. Choi and H. Ishikawa, “BrainDecoder: Style-based visual decoding of EEG signals,” 2024, arXiv: 2409.05729

    Pith/arXiv arXiv 2024

  80. [80]

    BrainDreamer: Reasoning-coherent and controllable image generation from EEG brain signals via language guidance,

    L. Wang, C. Wu and L Wang, “BrainDreamer: Reasoning-coherent and controllable image generation from EEG brain signals via language guidance,” 2024, arXiv: 2409.14021

    Pith/arXiv arXiv 2024

Showing first 80 references.

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.