Pith. sign in

REVIEW 2 major objections 5 minor 2 cited by

Unlocking Non-Invasive Brain-to-Text

T0 review · 2 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper claims that non-invasive brain recordings can be decoded into text at levels that beat all chance baselines, for the first time.

desk verdict Careful controls make the above-chance claim plausible, but the missing text-only baseline leaves the LLM-memorization confound unresolved; worth peer review with that control required. read the letter →

arxiv 2505.13446 v1 pith:MAL6LAGW submitted 2025-05-19 cs.LG

Dulhan Jayalath , Gilad Landau , Oiwi Parker Jones This is my paper · ORCID
classification cs.LG
keywords brain-to-textnon-invasiveBCIMEGdecodingspeechLLMrescoringdatasetpoolingchancebaselines
verification ladder T0 review T1 audit T2 compute T3 formal

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 claims that heard speech can be transcribed from non-invasive MEG recordings at a level that beats both random word selection and noise-input baselines on every standard metric. If correct, it removes a key obstacle to brain-computer interfaces that restore communication without surgery, since prior non-invasive attempts had never convincingly passed those baselines. The route is to turn a closed-vocabulary word classifier into a sentence decoder: a language model rescores beam-search candidates, a detector finds words outside the vocabulary, and an in-filling language model fills them using context. On the LibriBrain test set the best configuration reaches BLEU 0.25 against 0.07 for random selection, and selective pooling of datasets more than doubles word classification accuracy.

What carries the argument

The load-bearing mechanism is a beam search that scores each partial sentence as the sum of the neural word probability $\log P_{\textrm{model}}(w_i\mid x)$ and a weighted language-model prior $\lambda \log P_{\textrm{LLM}}(w_{1:i})$, turning a single-word classifier into a sequence decoder. Out-of-vocabulary words are detected by a gradient-boosted classifier fed with softmax-distribution statistics (the paper reports 88% AUROC) and are then filled in by a language model using surrounding context; a second in-context variant prompts the language model to both reselect from the top-five word probabilities and fill unknown positions. Cross-dataset training is made possible by a spatial-attention projection that maps heterogeneous sensor arrays into one latent space, with dataset selection guided by each dataset's standalone accuracy.

What would settle it

Take the exact beam search and in-filling recipe but feed it the stimulus text with no brain data, letting the language model decode from its own priors; if that text-only BLEU approaches the real-input value (0.25) instead of the noise-input value (0.15), the claimed brain contribution is largely language-model memory. A cleaner version is to run the full model on a newly written story that could not be in any language model's training data and check whether the real-versus-noise gap survives.

Watch

Extended reading notes

Core claim

The central claim is that a non-invasive electrophysiological brain-to-text system can exceed chance for the first time. The system decodes words from MEG segments aligned to word onsets, scores candidate sentences with a language-model beam search, and fills out-of-vocabulary positions from context; against a random-word baseline it reports BLEU 0.25 versus 0.07, with every metric significantly above both random selection and noise inputs ($p<0.001$). The paper also claims that pooling datasets by using standalone performance as a quality measure improves word classification accuracy by 2.1–2.3× and that these methods surpass all previously reported non-invasive approaches, with BLEU improved up to 2.6× over prior work. The claim is explicitly about heard speech with word-onset alignment, not inner or attempted speech.

Load-bearing premise

The noise-input control is only valid if the language models used for rescoring and in-filling cannot themselves predict the test passages, which are public-domain Sherlock Holmes stories that are likely present in the models' training data.

Editorial extensions

If this is right

  • Any future non-invasive B2T claim will need to be validated against both random-word and noise-input baselines; this paper's ablation table establishes that bar.
  • Word-level classifiers that already exist can be upgraded to sequence decoders through language-model rescoring, so progress in word decoding translates directly to text decoding.
  • Out-of-vocabulary words need not be retrained into the classifier: an uncertainty-based detector plus language-model in-filling expands the usable vocabulary.
  • Datasets can be combined by ranking them on standalone quality, so new recordings can be added to the training pool without hand-designed harmonisation.
  • The scaling behavior reported in the paper implies that current MEG datasets are still data-limited and that collecting more data should keep improving accuracy.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper never reports a text-only baseline, so how much of the 0.25-versus-0.15 gap is language-model predictability rather than brain signal remains an open test.
  • If the result survives that test, the same rescoring-and-in-filling pipeline is a natural candidate for attempted or inner speech once word onsets can be obtained, though the paper does not claim this.
  • The pooling rule suggests a practical selection criterion for any future dataset: pool with a dataset only when its standalone accuracy is high, and verify with a deliberately low-quality dataset that no improvement follows.
  • For assistive communication, exact transcription may be the wrong bar; even at high word-error rates the decoded text appears to retain semantic structure, so a metric focused on usable meaning may be more clinically relevant.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper proposes a non-invasive brain-to-text (B2T) pipeline that combines MEG/EEG word classification with LLM-based beam rescoring and predictive in-filling, together with a selective dataset-pooling strategy. On the LibriBrain MEG dataset of a subject listening to Sherlock Holmes audiobooks, the method reports BLEU 0.25 with real inputs versus 0.15 for noise inputs and 0.07 for random word selection, and claims to be the first non-invasive B2T method to surpass all critical chance baselines across standard metrics. Additional claims include a 2.1--2.3x improvement in word classification accuracy from dataset pooling and an open-vocabulary capability via an XGBoost OOV detector (AUROC 0.88). The manuscript includes detailed ablations, random baselines, session-independent train/test splits to avoid nonsense correlations, and extensive appendices describing prompts, hyperparameters, and decoding examples.

Significance. If the central claim holds, this is a meaningful milestone for non-invasive BCI research: it would demonstrate that word-level neural decoding combined with language-model priors can exceed no-signal baselines, and that heterogenous MEG/EEG datasets can be pooled to improve performance. The paper is unusually careful on several fronts: it provides noise-input and random-selection baselines, uses session-independent splits, reports standard errors over seeds, and explicitly addresses the teacher-forcing failure mode identified by Jo et al. The dataset-pooling analysis, with standalone performance as a quality measure, is a potentially useful contribution in its own right. However, the strength of the headline attribution claim is tempered by the LLM memorization risk and by cross-protocol comparisons in the main table.

major comments (2)
  1. [Section 4.1, Table 3, Appendix L] The noise-input control is a good baseline, but it does not fully address the possibility that the performance gap is amplified by LLM memorization of the public-domain Sherlock Holmes stimulus. Llama 3.2-1B and Claude 3.7 Sonnet are likely to have seen this widely reproduced text during training, and Appendix L shows that the LLM is explicitly prompted to complete and in-fill the exact predicted sequences, including <UNK> positions. Even though the real-vs-noise comparison in Table 3 holds the LLM constant, a few correct anchor words from the brain decoder could trigger the LLM to reconstruct the surrounding narrative from memory, thereby inflating BLEU and other n-gram metrics. To support the statement in Section 4.1 that "the results are attributable to decoding brain data," please add (i) a text-only baseline that runs the same beam search and in-filling with no informative neural input (e.g., an encoder whose outputs are replaced by a constant or by an uninformative distribution), and (ii) an analysis of whether the test stories appear in the training corpora of the two LLMs. The concern is partially mitigated by the observation that in-filling improves noise and real inputs by similar absolute amounts (noise +0.05, real +0.03 BLEU) and by the non-verbatim decoded examples in Appendix M, but the suggested experiments would make the attribution claim rigorous.
  2. [Table 1 and Abstract] The headline claim that BLEU is raised by 1.4--2.6x over prior work rests on cross-protocol comparisons that mix fMRI, EEG, and MEG systems with different stimuli, vocabularies, and evaluation protocols. For example, Tang et al. is an fMRI semantic reconstruction system rather than a word-level decoder, so a direct BLEU comparison is not apples-to-apples, and the "delta to random" values are computed relative to each method's own random baseline, which is not a common scale. Please restrict the state-of-the-art claim to the matched MEG comparisons in Table 2, or clearly qualify the cross-protocol comparison in the abstract and in Section 4.1. Without this qualification, the quantitative "up to 2.6x" claim overstates the evidence.
minor comments (5)
  1. [Section 4.1 / Appendix G] The abstract and Section 4.1 state p<<.001 for the comparisons in Table 3, but no test statistics or p-values are reported anywhere in the manuscript. Please report the actual p-values and state the test used (e.g., one-tailed or two-tailed Welch's t-test).
  2. [Section 4.3 / Table 1] The "Ours (best)" row in Table 1 appears to combine the best metric from different model configurations (e.g., greedy for some metrics, beam+fill for others). Please state explicitly which configuration produced each entry, or report a single configuration consistently.
  3. [Figure 3A] The reported correlation r=.95, p=.048 between standalone performance and improvement from joint training appears to be based on a very small number of datasets. Please state the number of points, the correlation type (Pearson or Spearman), and discuss the fragility of this result.
  4. [Appendix G, LLM APIs] The in-context transcription uses Claude 3.7 Sonnet through an API with 4096 thinking tokens. This introduces non-determinism and reproducibility concerns. Please provide the exact model version, inference temperature, and date of access, or consider using an open-weight model with fixed checkpoints.
  5. [Appendix F] The toy example for <UNK> insertion is useful, but the surrounding text contains a minor typo: "we include or exclude <UNK>" should read "we include or exclude the <UNK> token."

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation; the LLM-memorization concern is a validity risk, not a circularity.

full rationale

The paper's core pipeline (Section 3) trains a neural word classifier on held-out MEG splits and then applies a fixed LLM rescorer/in-filler during beam search (Eqs. 2-6). The claimed B2T result is evaluated on test stimuli (Tables 1-3) using metrics computed against the stimulus text; the model is not defined in terms of those metrics, no fitted parameter is renamed as a prediction (lambda=1.5 is a fixed hyperparameter), and the OOV detector is a supervised auxiliary model. The main benchmark LibriBrain is same-group, but Table 2 also evaluates on the independent Gwilliams MEG-MASC data, and the data-split appendix checks for nonsense correlations and stimulus leakage. The unresolved LLM training-data overlap (Appendix L) is a real external-validity threat and should be tested with a text-only baseline, but it is a confound, not a circular step: the LLM prior is identical in the real and noise conditions and no equation equates the outcome to the inputs. Hence score 0.

Assumptions & free parameters 5 free parameters · 7 assumptions · 0 invented entities

The method introduces no new physical entities. The free parameters are conventional ML hyperparameters, though lambda, temperature, vocabulary size, and OOV detector settings are all chosen by hand and affect the results. The key assumptions are domain-level: word-onset alignment, T5 embeddings as targets, noise-input validity, LLM non-memorization of the stimulus, and quality-proxy pooling. The most fragile is the LLM memorization assumption because the test text is public domain and the LLM is explicitly asked to complete those passages.

free parameters (5)
  • LLM rescoring weight λ = 1.5
    Balances the brain model and LLM scores in Equations 2-4; chosen by hand for beam search, no sensitivity analysis given.
  • Softmax sharpening temperature = 0.01
    Applied to the top-5 word probabilities in in-context transcription (Equation 10); selected because it provided much better separation (Appendix G).
  • Retrieval set size M = 250 words (default)
    Closed vocabulary size; chosen as the optimum in Figure 4 as a trade-off between text coverage (68% at 250) and classification accuracy.
  • Beam width = 5
    Beam search width for rescoring and in-filling, given in the hyperparameter table without sensitivity analysis.
  • OOV detector features and XGBoost settings = Table 9 features; lr=0.05, n_estimators=200, max_depth=4
    The detector is fit on encoder outputs from the training set; the final in-filling threshold is not specified.
assumptions (7)
  • domain assumption MEG/EEG responses to heard words are time-locked to word onsets and contain word-identity information decodable from aligned 50 Hz segments.
    The whole pipeline (Section 3.1) segments at word onsets; Appendix B shows heavy sensitivity to jitter, so this alignment assumption is load-bearing.
  • domain assumption The mean of the 12th-layer T5 word embedding is a valid target for neural word decoding.
    Adopted from d'Ascoli et al. without re-validation on the pooled datasets (Section 3.1, Appendix G).
  • domain assumption Random noise with the same mean and standard deviation as real MEG inputs is a valid null distribution for brain-derived decoding.
    Used in Table 3 following Jo et al.; validity depends on the noise matching the relevant statistics of real inputs.
  • domain assumption The LLM rescorer and filler do not encode the specific test stimuli through memorization.
    Sherlock Holmes text is public domain and likely in LLM training data; no overlap test or text-only baseline is reported (Section 3.2, Appendix L).
  • domain assumption Standalone dataset performance is a valid and transferable proxy for dataset quality when selecting pooling partners.
    Correlation r=0.95 is computed on the same four datasets used to formulate the rule (Section 4.2, Figure 3), so it is a small-sample empirical rule.
  • domain assumption Independent session splits avoid nonsense correlations between neural data and stimulus text.
    Appendix D tests overlapping versus independent splits and finds no significant difference, supporting the assumption.
  • standard math Standard statistical assumptions hold: i.i.d. test samples, Welch's t-test over 5 seeds, early stopping does not bias results.
    Used throughout training and significance testing (Appendix G).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unlocking Non-Invasive Brain-to-Text." pith.science (2026). https://pith.science/paper/MAL6LAGW

@misc{pith2026250513446,
  author       = {Pith},
  title        = {Pith review of: Unlocking Non-Invasive Brain-to-Text},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MAL6LAGW}},
  note         = {Machine review of arXiv:2505.13446}
}
abstract

Despite major advances in surgical brain-to-text (B2T), i.e. transcribing speech from invasive brain recordings, non-invasive alternatives have yet to surpass even chance on standard metrics. This remains a barrier to building a non-invasive brain-computer interface (BCI) capable of restoring communication in paralysed individuals without surgery. Here, we present the first non-invasive B2T result that significantly exceeds these critical baselines, raising BLEU by $1.4\mathrm{-}2.6\times$ over prior work. This result is driven by three contributions: (1) we extend recent word-classification models with LLM-based rescoring, transforming single-word predictors into closed-vocabulary B2T systems; (2) we introduce a predictive in-filling approach to handle out-of-vocabulary (OOV) words, substantially expanding the effective vocabulary; and (3) we demonstrate, for the first time, how to scale non-invasive B2T models across datasets, unlocking deep learning at scale and improving accuracy by $2.1\mathrm{-}2.3\times$. Through these contributions, we offer new insights into the roles of data quality and vocabulary size. Together, our results remove a major obstacle to realising practical non-invasive B2T systems.

Figures

Figures reproduced from arXiv: 2505.13446 by the authors.

Figure 1
Figure 1. Our approach outperforms all non-invasive B2T methods. Ranks are calculated from absolute improve￾ment over associated random baselines. Transcribing natural language text directly from speech￾related neural signals, known as brain-to-text decoding, remains one of neuroscience’s most challenging and clin￾ically significant frontiers. Recent breakthroughs showing remarkable accuracy (e.g. [1–5]) use surgical methods … view at source ↗
Figure 2
Figure 2. Brain-to-text decoding method. We pool data from multiple heterogeneous datasets (e.g. A and B) and align brain data segments to word onsets. Then, the segments are encoded by a signal encoder, handling dataset differences. A transformer learns the relationships between the encoded latents, embedding them with context. Its outputs are predictions of target word embeddings from a large language model. We map these pr… view at source ↗
Figure 3
Figure 3. Selectively pooling data improves accuracy by 2.3×. (A) Pairing datasets in training. Numbers show accuracy improvement on the evaluation dataset when trained additionally with the training dataset. Numbers in brackets show raw accuracy (asterisks are statistically significant against chance). The diagonal shows no paired training, i.e. standalone. Shading shows the change in accuracy relative to standalone. (B) Exp… view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Optimal vocabulary size differs by method and metric. We train our model with increasingly large vocabularies and show the effect on each decoding strategy’s performance. Methods with/without in-filling tend to converge with larger vocabularies as there are fewer words…
Figure 5
Figure 5. Figure 5: Alignment is critical. We add a random jitter in the range [0, jitter] to the aligned input samples. We train on LibriBrain with a vocabulary size of 250 and quote word classification accuracy [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]
Figure 6
Figure 6. Figure 6: Speech decoding is data-limited. The plot shows top-10 word classification accuracy on the test set as we scale up the volume of data in the train set of LibriBrain and Armeni. To scale up, we randomly sample, without replacement, increasing numbers of samples from all…
Figure 7
Figure 7. Figure 7: Generalization with data splits containing overlapping vs independent sessions. Refer to [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: Scaling vocabulary size on the LibriBrain dataset. The left plot shows how much of the story a vocabulary of a certain size will cover. The right plot shows the top-10 word classification accuracy as we scale the vocabulary. The plot on the right supports a similar res…
Figure 9
Figure 9. Figure 9: Edit distance matrix with and without <UNK> for a toy example. E Vocabulary Scaling and Coverage In [PITH_FULL_IMAGE:figures/full_fig_p017_9.png]
Figure 10
Figure 10. Figure 10: Score histograms. The scores are specific to the test set of LibriBrain for the experimental setup described in Appendix M. 27 [PITH_FULL_IMAGE:figures/full_fig_p027_10.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. Physiological Noise Augmentation Improves Non-Invasive Brain-to-Speech

    cs.LG 2026-07 conditional novelty 6.0 of 10

    PNA decomposes MEG recordings via ICA, isolates artifact components using EOG/ECG references, and re-injects scaled artifacts into clean data to train decoders that are invariant to physiological noise, improving imag...

  2. The 2025 PNPL Competition: Speech Detection and Phoneme Classification in the LibriBrain Dataset

    cs.LG 2025-06 conditional novelty 6.0 of 10

    The 2025 PNPL competition presents over 50 hours of within-subject MEG data, defines speech detection and phoneme classification benchmarks with F1-macro scoring, and reports reference baselines of 68.04% and 60.39%.

Reference graph

Works this paper leans on

46 extracted references · 31 canonical work pages · cited by 2 Pith papers

  1. [1]

    Neuroprosthesis for decoding speech in a paralyzed person with anarthria.New England Journal of Medicine, 385(3):217–227, 2021

    David A Moses, Sean L Metzger, Jessie R Liu, Gopala K Anumanchipalli, Joseph G Makin, Pengfei F Sun, Josh Chartier, Maximilian E Dougherty, Patricia M Liu, Gary M Abrams, et al. Neuroprosthesis for decoding speech in a paralyzed person with anarthria.New England Journal of Medicine, 385(3):217–227, 2021

    2021

  2. [2]

    A high-performance neuroprosthesis for speech decoding and avatar control

    Sean L Metzger, Kaylo T Littlejohn, Alexander B Silva, David A Moses, Margaret P Seaton, Ran Wang, Maximilian E Dougherty, Jessie R Liu, Peter Wu, Michael A Berger, et al. A high-performance neuroprosthesis for speech decoding and avatar control. Nature, 620(7976): 1037–1046, 2023

    work page 2023

  3. [3]

    A high-performance speech neuroprosthesis

    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, 2023

    2023

  4. [4]

    An accurate and rapidly calibrating speech neuroprosthesis

    Nicholas S Card, Maitreyee Wairagkar, Carrina Iacobacci, Xianda Hou, Tyler Singer-Clark, Francis R Willett, Erin M Kunz, Chaofei Fan, Maryam Vahdati Nia, Darrel R Deo, et al. An accurate and rapidly calibrating speech neuroprosthesis. New England Journal of Medicine, 391 (7):609–618, 2024

    work page 2024

  5. [5]

    A streaming brain-to-voice neuroprosthesis to restore naturalistic communication

    Kaylo T Littlejohn, Cheol Jun Cho, Jessie R Liu, Alexander B Silva, Bohan Yu, Vanessa R Anderson, Cady M Kurtz-Miott, Samantha Brosler, Anshul P Kashyap, Irina P Hallinan, et al. A streaming brain-to-voice neuroprosthesis to restore naturalistic communication. Nature Neuroscience, pages 1–11, 2025

    work page 2025

  6. [6]

    Signal quality of simultaneously recorded invasive and non-invasive EEG

    Tonio Ball, Markus Kern, Isabella Mutschler, Ad Aertsen, and Andreas Schulze-Bonhage. Signal quality of simultaneously recorded invasive and non-invasive EEG. Neuroimage, 46(3): 708–716, 2009

    work page 2009

  7. [7]

    The Brain’s Bitter Lesson: Scaling speech decoding with self-supervised learning

    Dulhan Jayalath, Gilad Landau, Brendan Shillingford, Mark Woolrich, and Oiwi Parker Jones. The Brain’s Bitter Lesson: Scaling speech decoding with self-supervised learning. In Forty- second International Conference on Machine Learning, ICML 2025, Vancouver, Canada, July 13-19, 2025. PMLR, 2025

    work page 2025

  8. [8]

    Decoding speech perception from non-invasive brain recordings

    Alexandre Défossez, Charlotte Caucheteux, Jérémy Rapin, Ori Kabeli, and Jean-Rémi King. Decoding speech perception from non-invasive brain recordings. Nature Machine Intelligence, 5(10):1097–1107, 2023

    2023

Show all 46 references
  1. [9]

    Decoding individual words from non-invasive brain recordings across 723 participants

    Stéphane d’Ascoli, Corentin Bel, Jérémy Rapin, Hubert Banville, Yohann Benchetrit, Christophe Pallier, and Jean-Rémi King. Decoding individual words from non-invasive brain recordings across 723 participants. arXiv preprint arXiv:2412.17829, 2024

    2024 arXiv

  2. [10]

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

    Zhenhailong Wang and Heng Ji. Open vocabulary electroencephalography-to-text decoding and zero-shot sentiment classification. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 36, pages 5350–5358, 2022. 10

    2022

  3. [11]

    Dewave: Discrete encoding of EEG waves for EEG to text translation.Advances in Neural Information Processing Systems, 36:9907–9918, 2023

    Yiqun Duan, Jinzhao Zhou, Zhen Wang, Yu-Kai Wang, and Chin-teng Lin. Dewave: Discrete encoding of EEG waves for EEG to text translation.Advances in Neural Information Processing Systems, 36:9907–9918, 2023

    2023

  4. [12]

    NeuSpeech: Decode neural signal as speech

    Yiqian Yang, Yiqun Duan, Qiang Zhang, Hyejeong Jo, Jinni Zhou, Won Hee Lee, Renjing Xu, and Hui Xiong. NeuSpeech: Decode neural signal as speech. arXiv preprint arXiv:2403.01748, 2024

    2024 arXiv

  5. [13]

    MAD: Multi-alignment MEG-to-text decoding

    Yiqian Yang, Hyejeong Jo, Yiqun Duan, Qiang Zhang, Jinni Zhou, Won Hee Lee, Ren- jing Xu, and Hui Xiong. MAD: Multi-alignment MEG-to-text decoding. arXiv preprint arXiv:2406.01512, 2024

    2024

  6. [14]

    NeuGPT: Unified multi-modal neural GPT

    Yiqian Yang, Yiqun Duan, Hyejeong Jo, Qiang Zhang, Renjing Xu, Oiwi Parker Jones, Xuming Hu, Chin-teng Lin, and Hui Xiong. NeuGPT: Unified multi-modal neural GPT. arXiv preprint arXiv:2410.20916, 2024

    2024 arXiv

  7. [15]

    Are EEG- to-text models working? The Fourth International Workshop on Human Brain and Artificial Intelligence at IJCAI, 2024

    Hyejeong Jo, Yiqian Yang, Juhyeok Han, Yiqun Duan, Hui Xiong, and Won Hee Lee. Are EEG- to-text models working? The Fourth International Workshop on Human Brain and Artificial Intelligence at IJCAI, 2024

    2024

  8. [16]

    Non-invasive neural decoding in source reconstructed brain space

    Yonatan Gideoni, Ryan Charles Timms, and Oiwi Parker Jones. Non-invasive neural decoding in source reconstructed brain space. arXiv preprint arXiv:2410.19838, 2024

    2024 arXiv

  9. [17]

    Resolving domain shift for representations of speech in non-invasive brain recordings

    Jeremiah Ridge and Oiwi Parker Jones. Resolving domain shift for representations of speech in non-invasive brain recordings. arXiv preprint arXiv:2410.19986, 2024

    2024 arXiv

  10. [18]

    Brant: Foundation model for intracranial neural signal

    Daoze Zhang, Zhizhang Yuan, Yang Yang, Junru Chen, Jingjing Wang, and Yafeng Li. Brant: Foundation model for intracranial neural signal. Advances in Neural Information Processing Systems, 36:26304–26321, 2023

    2023

  11. [19]

    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, Vienna, Austria, May 7-11, 2024, 2024

    2024

  12. [20]

    EEGPT: Pretrained transformer for universal and reliable representation of EEG signals

    Guangyu Wang, Wenchao Liu, Yuhong He, Cong Xu, Lin Ma, and Haifeng Li. EEGPT: Pretrained transformer for universal and reliable representation of EEG signals. Advances in Neural Information Processing Systems, 37:39249–39280, 2024

    2024

  13. [21]

    Semantic reconstruction of continuous language from non-invasive brain recordings

    Jerry Tang, Amanda LeBel, Shailee Jain, and Alexander G Huth. Semantic reconstruction of continuous language from non-invasive brain recordings. Nature Neuroscience, 26(5):858–866, 2023

    2023

  14. [22]

    emg2qwerty: A large dataset with baselines for touch typing using surface electromyography

    Viswanath Sivakumar, Jeffrey Seely, Alan Du, Sean Bittner, Adam Berenzweig, Anuoluwapo Bolarinwa, Alex Gramfort, and Michael Mandel. emg2qwerty: A large dataset with baselines for touch typing using surface electromyography. Advances in Neural Information Processing Systems, 3...

    2024

  15. [23]

    Brain-to-text decoding: A non-invasive approach via typing.arXiv preprint arXiv:2502.17480, 2025

    Jarod Lévy, Mingfang Zhang, Svetlana Pinet, Jérémy Rapin, Hubert Banville, Stéphane d’Ascoli, and Jean-Rémi King. Brain-to-text decoding: A non-invasive approach via typing.arXiv preprint arXiv:2502.17480, 2025

    2025 arXiv

  16. [24]

    BIOT: Biosignal transformer for cross-data learning in the wild

    Chaoqi Yang, M Westover, and Jimeng Sun. BIOT: Biosignal transformer for cross-data learning in the wild. Advances in Neural Information Processing Systems, 36:78240–78260, 2023

    2023

  17. [25]

    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, 15:653659, 2021

    2021

  18. [26]

    Neural data transformer 2: multi- context pretraining for neural spiking activity

    Joel Ye, Jennifer Collinger, Leila Wehbe, and Robert Gaunt. Neural data transformer 2: multi- context pretraining for neural spiking activity. Advances in Neural Information Processing Systems, 36:80352–80374, 2023

    2023

  19. [27]

    Learning topology-agnostic EEG represen- tations with geometry-aware modeling

    Ke Yi, Yansen Wang, Kan Ren, and Dongsheng Li. Learning topology-agnostic EEG represen- tations with geometry-aware modeling. Advances in Neural Information Processing Systems, 36:53875–53891, 2023. 11

    2023

  20. [28]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. Advances in neural information processing systems, 30, 2017

    2017

  21. [29]

    Exploring the limits of transfer learning with a unified text-to-text transformer

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of machine learning research, 21(140):1–67, 2020

    2020

  22. [30]

    meta-llama/Llama-3.2-1B

    Meta Llama Team. meta-llama/Llama-3.2-1B. https://huggingface.co/meta-llama/ Llama-3.2-1B. Accessed: 2025-04-26

    2025

  23. [31]

    A. Viterbi. Error bounds for convolutional codes and an asymptotically optimum decoding algorithm. IEEE Transactions on Information Theory, 13(2):260–269, 1967. doi: 10.1109/TIT. 1967.1054010

    1967

  24. [32]

    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 , pages 4171...

    2019

  25. [33]

    Claude 3.7 Sonnet and Claude Code

    Anthropic. Claude 3.7 Sonnet and Claude Code. https://www.anthropic.com/news/ claude-3-7-sonnet . Accessed: 2025-04-22

    2025

  26. [34]

    The 2025 PNPL competition: Speech detection and phoneme classification in the LibriBrain dataset

    Gilad Landau, Miran Özdogan, Gereon Elvers, Francesco Mantegna, Pratik Somaiya, Dulhan Jayalath, Luisa Kurth, Teyun Kwon, Brendan Shillingford, Greg Farquhar, Minqi Jiang, Karim Jerbi, Hamza Abdelhedi, Yorguin Mantilla Ramos, Caglar Gulcehre, Mark Woolrich, Natalie V oets, and...

    2025

  27. [35]

    LibriBrain: Over 50 hours of within-subject MEG to improve speech decoding methods at scale

    Miran Özdogan, Gilad Landau, Gereon Elvers, Dulhan Jayalath, Pratik Somaiya, Francesco Mantegna, Mark Woolrich, and Oiwi Parker Jones. LibriBrain: Over 50 hours of within-subject MEG to improve speech decoding methods at scale. arXiv preprint, 2025. Licensed under CC-BY-NC

    2025

  28. [36]

    A 10-hour within-participant magnetoencephalography narrative dataset to test models of language com- prehension

    Kristijan Armeni, Umut Güçlü, Marcel van Gerven, and Jan-Mathijs Schoffelen. A 10-hour within-participant magnetoencephalography narrative dataset to test models of language com- prehension. Scientific Data, 9(1):278, 2022. Licensed under CC-BY-4.0

    2022

  29. [37]

    Introducing MEG-MASC a high-quality magneto-encephalography dataset for evaluating natural speech processing

    Laura Gwilliams, Graham Flick, Alec Marantz, Liina Pylkkänen, David Poeppel, and Jean-Rémi King. Introducing MEG-MASC a high-quality magneto-encephalography dataset for evaluating natural speech processing. Scientific data, 10(1):862, 2023. Licensed under CC0 1.0 Universal

    2023

  30. [38]

    Electrophysiological correlates of semantic dissimilarity reflect the compre- hension of natural, narrative speech

    Michael P Broderick, Andrew J Anderson, Giovanni M Di Liberto, Michael J Crosse, and Edmund C Lalor. Electrophysiological correlates of semantic dissimilarity reflect the compre- hension of natural, narrative speech. Current Biology, 28(5):803–809, 2018. Licensed under CC0 1.0...

    2018

  31. [39]

    BLEU: A method for automatic evaluation of machine translation

    Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. BLEU: A method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, pages 311–318, 2002

    2002

  32. [40]

    ROUGE: A package for automatic evaluation of summaries

    Chin-Yew Lin. ROUGE: A package for automatic evaluation of summaries. In Text summariza- tion branches out, pages 74–81, 2004

    2004

  33. [41]

    METEOR: An automatic metric for MT evaluation with improved correlation with human judgments

    Satanjeev Banerjee and Alon Lavie. METEOR: An automatic metric for MT evaluation with improved correlation with human judgments. In Proceedings of the ACL workshop on intrinsic and extrinsic evaluation measures for machine translation and/or summarization, pages 65–72, 2005

    2005

  34. [42]

    Weinberger, and Yoav Artzi

    Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q. Weinberger, and Yoav Artzi. Bertscore: Eval- uating text generation with BERT. In8th International Conference on Learning Representations, ICLR, 2020. 12

    2020

  35. [43]

    Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks

    Alex Graves, Santiago Fernández, Faustino Gomez, and Jürgen Schmidhuber. Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks. In Proceedings of the 23rd international conference on Machine learning, pages 369–376, 2006

    2006

  36. [44]

    Nonsense correlations in neuroscience

    Kenneth D Harris. Nonsense correlations in neuroscience. biorxiv, pages 2020–11, 2020

    2020

  37. [45]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In 7th International Conference on Learning Representations, ICLR, 2019

    2019

  38. [46]

    don’t", 1:

    Tianqi Chen and Carlos Guestrin. XGBoost: A scalable tree boosting system. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 785–794, 2016. 13 A Sensor Positions and Dataset Pooling A key concern in pooling data is tha...

    2016

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.